White Paper ‑ PetaJakarta.org:
Assessing the Role of Social Media for Civic Co‑Management During Monsoon Flooding
in Jakarta, Indonesia

Tomas Holderness & Etienne Turpin

Download Bahasa Indonesian | English

Press Release

Available from the University of Wollongong.

Contributing Researchers

Matthew Berryman, Rodney Clarke, Sara Dean, Yantri Dewi, Olivia Dun, Ben Jones, Mohammad Kamil, Robert Ogie, Rhys Powell, Mary O’Malley, Milly Matthews-Mulroy, Alifa Rachmadia Putri, Widya Ramadhani, Frank Sedlar, Ariel Shepherd, Fitria Sudirman, Rohan Wickramasuriya and Albert Yang.

Supporting Agencies

PetaJakarta.org is a research project led by the SMART Infrastructure Facility, University of Wollongong in collaboration with the Jakarta Emergency Management Agency (BPBD DKI Jakarta) and Twitter Inc. The Joint Pilot Study for the project was operationally active from December 2014 to March 2015; during this time, the project enabled Jakarta’s citizens to report the locations of flood events using the social media network Twitter, thereby contributing to a publicly accessible real-time map of flood conditions at PetaJakarta.org. These data were used by BPBD DKI Jakarta to cross-validate formal reports of flooding from traditional data sources, supporting the creation of information for flood assessment, response, and management in real-time. This white paper provides a description of the PetaJakarta.org research project from both citizen and government perspectives and reports on the results of the Joint Pilot Study to evaluate of the project. At the time of publication, the project has entered its second phase—Operational Integration and Technology Transfer—as part of the ongoing collaboration between the SMART Infrastructure Facility, BPBD DKI Jakarta, and Twitter Inc.

Photo 01. BPBD DKI Jakarta during collaborative workshop at the SMART Infrastructure Facility, University of Wollongong, October 2014.

Photo 01. BPBD DKI Jakarta during collaborative workshop at the SMART Infrastructure Facility, University of Wollongong, October 2014.

Through its integration in BPBD DKI Jakarta’s existing disaster risk management (DRM) information ecosystem, the PetaJakarta.org project has proven the value and utility of social media as a mega-city methodology for crowd-sourcing relevant situational information to aid in decision-making and response coordination during extreme weather events. PetaJakarta.org is a promising evolution within the DRM information ecosystem because it leverages both the inherent capabilities within ubiquitous mobile devices (i.e. GNSS-enabled messaging) and the network capabilities of social media through free and open source software (FOSS) to provide validated and actionable information for citizens and government agencies, thereby improving situational knowledge and increasing response times in disaster scenarios. To maintain resilience within this information ecosystem, the study emphasizes the need for access to data via open application programming interfaces (APIs), which enable the integration of vulnerability information and potential hazard exposure to facilitate integrated risk evaluation and assessment.

For residents of Jakarta, PetaJakarta.org enabled autonomous users to make independant decisions on safey and navigation in response to the flood in real-time, thereby helping increase the resilience of the city’s residents to flooding and its attendant difficulties. Critically, this outcome was achieved using the same data and map that was used by the government; designing the platform to meet the needs of citizen-users and government agencies enables and promotes civic co-management as a strategy for climate adaptation. This social strategy is underpinned by the open source software CogniCity, a framework for urban data that anyone is free to inspect, download, and redesign; this open source ethos and the transparency it facilitates was critical to the success of the Joint Pilot Study.

PetaJakarta.org has demonstrated social media’s valuable niche within the disaster risk management information ecosystem; as an operational tool capable of providing decision support at the various spatial and temporal scales required by the different actors within city, PetaJakarta.org offers an innovative and inexpensive method for the crowdsourcing of time-critical situational information in disaster scenarios.

In this section, we provide a general overview of the PetaJakarta.org Joint Pilot Study and offer preliminary explanations of the method, software, and relevant infrastructure in Jakarta, Indonesia. While Jakarta is not unique in facing severe flooding during the seasonal monsoon in Asia, it is an ideal case study for the development of open source software to support crowdsourcing of time-critical situational information during extreme weather events. First, Jakarta’s government has embraced social media as a means to communicate with residents; in fact, the Jakarta Emergency Management Agency has a strong mandate to improve communication with the public through social media. Second, the Jakarta government has already actively endorsed and supported the development of open data projects (such as the extensive district mapping conducted by the Humanitarian OpenStreetMap Team, discussed further in Section 4.1.3). Third, Jakarta has an extremely high proportion of mobile phone users and one of the highest concentrations of Twitter users in the world; because the PetaJakarta.org project relies on densely populated urban environments with high proportions of social media users, Jakarta provides an especially compelling environment for understanding the value of social media in a disaster risk management context. The scalability and transferability of lessons learned from Jakarta is discussed in further detail below (see especially Sections 6.1 and 6.2).

Photo 02. PetaJakarta.org researchers Ariel Shepherd and Yantri Dewi system testing in Bukit Duri, East Jakarta; February 2014.

Photo 02. PetaJakarta.org researchers Ariel Shepherd and Yantri Dewi system testing in Bukit Duri, East Jakarta; February 2014.

1.1 Project Summary

1.1.1 PetaJakarta.org

PetaJakarta.org is a world-first Joint Pilot Study initiated by the SMART Infrastructure Facility, University of Wollongong, in collaboration with the Jakarta Emergency Management Agency (BPBD DKI Jakarta) and Twitter Inc. The study aims to research and develop open source software (OSS) for the integrated management of social media and API-sourced data1 in order to make risk information open, accessible, and actionable by residents, government agencies, NGOs, and private sector developers. PetaJakarta.org runs on the OSS CogniCity, a framework for urban data collection and management developed by the SMART Infrastructure Facility and the SMART OSGeo Lab; released as OSS, CogniCity can readily be deployed in other urban environments in Indonesia and Southeast Asia facing similar challenges. As rates of urbanization continue to increase in Asia [Fig. 01], it is imperative to develop tools that can facilitate the co-management of increasingly complex urban environments. With more than fourteen megacities located on deltas in the region [Fig. 02], hundreds of millions of urban residents will face difficult challenges due to extreme weather events as a result of climate change (IPCC 2014); using the power of social media to network and validate time-critical situational information offers a key element for resilience to extreme weather events. PetaJakarta.org is the first pilot study to prove the value of integrated social media and API-sourced data for the civic co-management of disasters resulting from extreme weather events in a major Southeast Asian megacity.

1API refers to Application Programming Interface, or the means by which ‘machine-to-machine’ communication occurs between computer programs automatically, without the aid of human agents.

Fig. 01. Map showing Asia's population.

Fig. 01. Map showing Asia's population.

The pilot project for CogniCity was sited in Jakarta for several key reasons. Jakarta, with its surrounding conurbation of Jabodatabek, has the highest rate of urbanization in the world and comprises the second-largest contiguous settlement on earth. With a greater metropolitan area hosting 13 rivers, 1100 kilometers of canals, seasonal monsoon flood events, and over 28 million residents, Jakarta is a key case study for the development of improved risk management through new tools and open source software (Turpin et al. 2013; de Wall 2014). Open source software solutions are critical in Jakarta because cost-prohibitive private products remain unrealistic as solutions under current budget constraints. In Jakarta, risk information and coordination through open data protocols is critical to support decision-making about disaster response, emergency planning, and community resilience. Furthermore, rich suites of open and accessible geospatial risk data generate activity in NGOs and the private sector, especially for longer term disaster risk management planning tools, such as InaSAFE (developed by the Australia-Indonesia Facility for Disaster Reduction), and economic calculators such as JakSAFE (developed by the World Bank). The development of social media and application-driven data collection via mobile devices allows for unprecedented data collection capacities; in order to be effective, these technologies require coordination through robust, enterprise-grade open source software.

BPBD DKI Jakarta is regularly faced with the difficult challenge of anticipating and responding to floods hazards and related extreme weather events in Jakarta. As a research partner, the organization allowed PetaJakarta.org invaluable access to their operations through a process of co-research and institutional ethnography. By carefully studying the operational procedures, concerns, and ambitions of the Agency, the project developed tools that could be integrated effectively into the existing structure of the organization and its various data flowlines, and transferred incrementally over the course of the research collaboration, thereby ensuring an efficient strategy for both development and implementation.

Twitter was selected as the ideal social media platform for the Joint Pilot Study because Jakarta has one of the highest concentrations of Twitter users in the world. Additionally, a Twitter #DataGrant gave the research team unprecedented access to a suite of historical data about flooding in Jakarta, thus allowing the system to be developed and calibrated with large data sets, thereby ensuring its operative functionality during the actual monsoon season in 2014-2015. Finally, by working in close collaboration with Twitter, the project benefitted from the advice and mentorship of a group of media experts, engineers, and advocates who were all critical to the overall success of the project.

Fig. 02. Map showing Megacities in Southeast Asia.

Fig. 02. Map showing Megacities in Southeast Asia.

1.1.2 GeoSocial Intelligence Framework

While social media has heralded a number of studies regarding its potential application for scientific research and civic governance, the movement from social media to “GeoSocial Intelligence” (Holderness 2014) relies on a more comprehensive understanding of the information ecosystem within which social media operates. According to Easterling (2014: 17), “Between 2000 and 2013, the global number of cell phone subscriptions went from 740 million to 6.8 billion phones with over three-quarters of phones in the developing world.” An ever-increasing number of these mobile devices are built with Global Navigation Satellite System (GNSS)-functionalities (e.g. GPS), thus allowing for the scientific study of GNSS-located data, frequently shared via social media platforms. The “intelligence” in such a configuration is not immediate; extracting knowledge from the “noise” of social media requires designed engagement and filtering processes to eliminate unwanted information, reward valuable reports, and display useful data in a manner that further enables users, governments, or other agencies to make non-trivial, actionable decisions in a time-critical manner. Thus, a GeoSocial Intelligence Framework depends on four basic principles: 1) develop reliable, free and open source software (FOSS) that enables the gathering, sorting, and display of useful disaster-related information; 2) move from passive spatial and temporal data mining techniques to “big crowdsourcing,” wherein users on a social media platform are actively encouraged to share information relevant to a given situation or anticipated scenario; 3) promote an ethic of co-management that values the peer-to-peer sharing of situational information within the same platform that is used by government agencies and first responders, who can transparently monitor and cross-check the data being shared; 4) make the data open, so that all users can inspect the software, review the system, and develop complementary tools and technologies that further enhance resilience within the information ecosystem.

The value of social media as a critical tool in the disaster emergency management toolkit has be proven by a number of previous studies (Meier 2014); the move from social media to GeoSocial Intelligence is an especially promising evolution of this prior research and software development. While the utility of passively-mined social media data can offer insights for offline analytics and derivative studies for future planning scenarios, the critical issue for frontline emergency responders is the organization and coordination of actionable, real-time data related to disaster situations. From this perspective, GeoSocial Intelligence can be reasonably understood as a promising evolution within the disaster risk management information ecosystem because it leverages both the inherent capabilities within ubiquitous mobile devices (i.e. GNSS-enabled messaging) and the network capabilities of social media, through OSS to provide validated and actionable information for citizens and government agencies, thereby improving situational knowledge and increasing response times in disaster scenarios.

1.1.3 CogniCity

PetaJakarta.org is the pilot project for the OSS CogniCity, developed at the SMART Infrastructure Facility as an open framework for urban data collection and analysis. The objective of this OSS is to remove redundancies in the study and promotion of urban resilience to extreme weather events as a result of climate change; the elimination of redundancies occurs because CogniCity can be used simultaneously to study patterns and practices of resilience among users and government agencies, and to enable more effective sharing of information between citizens, government agencies, and NGOs. At the time of writing, CogniCity is the only fully developed and tested OSS platform that uses a GeoSocial Intelligence Approach to urban data collection, analysis, and sharing. Such a system has a wide variety of applications for civic co-management as a strategy for adaptation to climate change, not least because the OSS is designed from first principles to be transferable among various domains of application (such as other weather-related hazards, waste management, or crime) and other geographies and languages.

1.2 Fluid Geographies of the Coastal Megacity

1.2.1 Flooding in Jakarta

The coastal megacity of Jakarta experiences annual flooding during the seasonal monsoon which seriously affects human life, property, and urban infrastructure (Hartono et al. 2010). The city is situated on a coastal deltaic plain served by 13 rivers which flow from the mountains in the south, northwards through the city to the Java sea [Fig. 03]. Over forty percent of the city is at or below sea level, necessitating a complex system of hydraulic and hydrological infrastructure to manage the movement of water through the city (Hartono et al. 2010). The rapid urbanisation and population explosion seen in Jakarta over the past decade (Li 2003) has diminished natural surface infiltration, increased ground subsidence, and added additional stress on the deteriorating hydrological infrastructure, increasing the city’s exposure and vulnerability to flooding (Hallegatte et al. 2013).

Importantly, however, flooding can no longer be seen as a “natural” event; as anthropogenic climate change interacts with the urban environment in increasingly unpredictable ways, the complex urban environment and its technical systems for the management of weather-related hazards is as much a part of the monsoon ecology as the precipitation itself. According to the renown philosopher of technology Jean-Luc Nancy (2015: 3-4), “From now on there is an interconnection, an intertwining, even a symbiosis of technologies, exchanges, movements, which makes it so that a flood—for instance—wherever it may occur, must necessarily involve relationships with any number of technical, social, economic, political intricacies that keep us from regarding it simply as a misadventure or a misfortune whose consequences can be more or less easily circumscribed.” Understanding these changing dynamics may mean giving up on expensive, “blackbox” (Townsend 2013: 88) predictive flood models, which have proven ineffective as tools for preparedness and response.

Fig. 03. Map showing Jakarta and PetaJakarta.org Study Area.

Fig. 03. Map showing Jakarta and PetaJakarta.org Study Area.

1.2.2 Regional Perspectives on Flood Management

Cities are constantly evolving into more complex systems of interconnected infrastructure and people; as a consequence, many coastal cities become increasingly vulnerable to the impacts of severe weather events and global climate change (Walsh et al. 2011). Insight into the resilience of these urban systems and the effects of failing infrastructure in such events is crucial to the understanding of a city’s ability to adapt to climate change (Rosenzweig 2011). This knowledge is particularly significant in the developing nations of Southeast Asia, which are expected to experience the greatest impacts of climate change (IPCC 2014), yet are often unable to make informed, evidence-based decisions concerning infrastructure and urban planning due to a lack of resources and an absence of reliable or consistent data (Paar & Rekittke 2011).

1.2.3 Regional Perspectives on Social Media

The rise of social media platforms has taken city planners by surprise because these technologies not only provide unprecedented volumes of data relevant to the analysis of urban systems, but, more fundamentally, they change the way residents interact with each other and with the city at large (Townsend 2013). As first adopters continue to expedite the uptake of social media in Asian megacities, it is imperative to understand how to best harness the power of networked communication technologies to overcome urban challenges that result from climate change. Because geospatial information derived from social media is especially valuable and actionable within dense urban environments with a high proportion of networked users, Asian megacities facing the combined challenge of rapid development and climate change are critical sites for the development of new OSS that can turn the noise of social media into intelligent, actionable insights for users and governments alike.

1.3 Infrastructure Overview in Indonesia

1.3.1 Jakarta’s Flood Infrastructure: A General History

Seasonal flooding has been a part of life in Jakarta since the seventeenth century. By 1850, the Dutch colonial government finally took action and established the Burgelijke Openbare Werke, a task force responsible for flooding in the colonial city of Batavia. Following the major flood of 1918, an integrated master plan was designed to prevent further flooding; the West Flood Canal was built under the supervision of Prof. Dr. Herman van Breen. Later, under Soekarno’s administration, Waduk Pluit, Waduk Setiabudi, Waduk Tomang and Waduk Grogol were all successfully constructed as part of Kopro Banjir project. Following these infrastructural measures to combat flooding, Soeharto then planned to widen the canal in 1973, but the project failed to launch. Instead, the central government and local governments built the Cengkareng Drainage System as a flood control network that was finished a decade later. Despite these efforts, the flood infrastructure was unable to protect Jakarta as major floods paralyzed the city in 1979, 1996, 1999 and 2002. The Indonesian government, under Megawati’s administration, initiated the East Flood Canal as an extension of the original design by van Breen; its construction was completed under Susilo Bambang Yudhoyono’s government, costing over 4.9 trillion Rupiah (US$ 373 million). Notably, as in many other countries hit by the 1997 financial crisis, Indonesia’s infrastructure has also suffered from underinvestment for over a decade; flood management infrastructure has been seriously affected.

Photo 03. Children use the pumps in North Jakarta as a playground as water from Jakarta’s rivers is pumped over the seawall into Java Bay; March 2014.

Photo 03. Children use the pumps in North Jakarta as a playground as water from Jakarta’s rivers is pumped over the seawall into Java Bay; March 2014.

The most recent strategy employed to fight flooding, initiated by the central government, is the National Capital Integrated Coastal Development program (NCICD). NCICD is supported by the government of The Netherlands, DKI Jakarta, and the government of Indonesia. The plan is to create an outer sea wall, the investment costs for which will be recovered by selling properties of a new waterfront city built on reclaimed land (NCICD 2014). The NCICD project, also known colloquially as Great Garuda Sea Wall, has attracted various controversies (Noviansyah 2015). Some see this solution as too profit-oriented and argue that the sea wall would lead to a “giant sewerage system” given that rivers and canals are highly polluted and would effectively drain the sewage of millions of residents into a giant retention pond on the coast, behind the proposed sea wall. There are also criticisms regarding the possible damage to the local marine ecology and fisheries, as seen during the development of the Saemangeum Sea Wall in South Korea (Lighthouse Foundation 2006). In addition, many in the public doubt the feasibility of the project as Public Private Partnership (PPP) because such schemes are relatively new in Indonesia, and because the country has seen numerous infrastructure failures due to poor project execution and corruption (Noviansyah 2015).

The Jakarta government is also currently undertaking a “Normalization” project aimed at increasing the drainage capacity of waterways, such as Ciliwung River. The project will displace over 34,000 families living along the river bank and widen Ciliwung river to almost double its natural width (Rujak Center 2008). The measurement of sedimentation in surrounding channels and the Ciliwung River suggests that normalization could damage ecology and fisheries in ways similar to that of the Saemangeum Sea Wall in South Korea (Smith 2015).

As the population of Jakarta, and the surrounding area of Jabodetabek, continues to expand, the impacts of flooding are significantly heightened. The 1996 flood forced 30,000 people to evacuate; the flood in 2007 inundated houses of 320,000 residents and claimed the lives of 80 people (Taufik 2014). At present, 72.7% of Jakarta is prone to flooding (Rostanti 2014), threatening the lives of over 983,399 residents, or 10.2% of Jakarta’s population, with residents in North Jakarta facing the highest risk (Rostanti 2014).

Although Jakarta is the main urban area affected by flooding, the hydrological problem can be traced all the way to Bogor, 60 kilometers south of Jakarta. There are 13 major rivers running through Jakarta; most of these rivers originate in the mountainous areas of Bogor. Regrettably, the natural catchment areas of Bogor are decreasing due to weak land use policy enforcement. For example, based on Regional Regulation No. 19 / 2008, Kampung Sukatani is a conservation area and therefore construction in this area is prohibited; however, Kampung Sukatani has been transformed into an area of extremely luxurious residences. The villas were forcefully demolished by the Bogor government (with funding from the Jakarta government) in 2013, only to be rebuilt again two years later (Rahmawati 2015). Examples like these show the need for an enduring commitment on the part of the governments of Bogor and Jakarta to manage the root causes of flooding upstream, where land use changes have serious effects on the ability of the downstream hydrological system to cope with the annual monsoon rains.

1.3.2 The Future of Flood Infrastructure in Jakarta

The history of investment in flood infrastructure in Jakarta tends to suggest that new infrastructure rarely guarantees a reasonable cost-benefit analysis; research also suggests that the government should commit to a thorough analysis of all existing facilities and assets to develop an operational overview, in addition to investing in new infrastructure and additional maintenance. In the 2014 Regional Budget, the government allocated as little as 0.0153%, or 750 billion IDR, for flood control systems and drainage maintenance. With such a modest investment in the maintenance of critical flood infrastructure, it is not surprising that before monsoon season of 2012, for example, it was discovered that 141 out of 555 pumps in Jakarta were not operational (Kompas 2014). While infrastructure asset analysis is a relatively novel science, it is nevertheless critical for the Jakarta government to devote resources to assessments and maintenance in order to determine the current state of flood control infrastructure and to make evidence-based, targeted investments in new facilities that can reduce risk in the weakest aspects of the current system. Because infrastructure facilities do not exist individually and are highly interdependent (Ebrahimy 2014; Tran et al. 2014), functional failures in one flood management facility can make any attempt to add new components to the system difficult to plan. Decisions regarding critical flood infrastructure investment should be evidence-based and rely on an understanding of critical interdependencies, energy demands, and an integrated assessment of needs in all related sectors because infrastructure interdependencies bring with them layers of complexity, uncertainty, and risk to urban planning and design (Tran et al. 2014).

From one administration to another, the most common approach toward flood infrastructure in Jakarta is to focus on new physical construction. Not only are these ventures capital intensive, but projects with large construction budgets are also prone to corruption (Yanto, 2010). Currently, information and communication technologies (ICT) are a neglected dimension of flood management infrastructure. With such a high rate of mobile phone penetration, and exceedingly high rates of social media usage, the Jakarta government is well-positioned to encourage greater public participation in infrastructure monitoring and flood event reporting; these elements of civic co-management could help alleviate much of the burden currently carried borne solely by the government. However, such an approach requires a new vision of infrastructure that goes beyond canals and pumps, and begins to approach the question of flood management from a holistic view that includes ICT and public utilities for flood reporting.

As a compliment to increased ICT investment, open data allows universal participation where data is available as a whole and and at no more than a reasonable reproduction cost. By providing data under terms that permit reuse and redistribution, including intermixing with other datasets, citizen-users, NGOs, student and advocacy groups, and private developers can generate new tools for information management within the application economy. One excellent example of a government innovation to promote open data is the recently released SmartCity Jakarta website, with its applications Qlue (public) and Crop (government use only). Further investment in projects like SmartCity Jakarta, and the API-enabled data they produce, would be a welcome element to the DRM information ecosystem.

Jakarta has one of the highest concentration of Twitter users worldwide (Semiocast, 2012), and experiences severe seasonal flooding (Hartono et al., 2010). The tacit knowledge of local communities, government agencies, and first responders in Jakarta, as well as the dense network of mobile sensors connected via social media, provides a data source of unprecedented resolution for mitigating urban risk. In the context of DRM, the challenge for information and communication technologies is not to develop new sensors or additional applications for crowd-sourcing data collection, but instead to seed the evolution of social media networks as a mega-city methodology for resilience to extreme weather events and climate change. This section explores the development of PetaJakarta.org and its underlying software CogniCity as a Geosocial Inteligence Framework for civic co-management during periods of flooding in Jakarta throughout the 2014-2015 monsoon season.

Photo 04. Resident in East Jakarta using a mobile device during monsoon flooding; January 2014. Photography by Ariel Shepherd.

Photo 04. Resident in East Jakarta using a mobile device during monsoon flooding; January 2014. Photography by Ariel Shepherd.

2.1 CogniCity: A GeoSocial Intelligence Framework for Urban Data

CogniCity is an open source framework for urban data, which harnesses the power of social media by gathering, sorting and displaying real-time situational reports from urgent infrastructure issues such as flooding. The CogniCity toolset, first trialed through the PetaJakarta.org pilot study, makes crowd-sourced information available in real-time to citizens in need of daily information, and to government agencies to support decision making using a strategy of civic co-management. CogniCity builds on Geographical Information Systems (GIS) theory, to gather citizen reports from the social media network Twitter and create geospatial visualisations of this information [Fig. 04]. However, CogniCity extends the traditional GIS paradigm to facilitate the process of citizen Twitter reporting by programmatically sending “invitation tweets” to users in the city who use the keyword “banjir” (flood). Reports are collected in a centralised geospatial database, and served via a data API to a client-side rendered map showing activity across the city in real-time, or as a data layer within external organisation geographical information systems [Fig. 04].

Fig. 04. CogniCity System Architecture.

Fig. 04. CogniCity System Architecture.

2.2 CogniCity: System Requirements & Design

CogniCity was conceived in response to system testing during a flood event in Jakarta on 5 February 2014, which captured 150,000 tweets in Jakarta related to flooding in a 24 hour period [Table 01]. This test was conducted using a proof-of-concept, rapid prototype NodeJS application, connected to the public Twitter Stream application programming interface (API), filtering tweets by location and keyword. The keywords included: banjir (‘flood’), tinggi (‘high’), genangan (‘pool’) and terendam (‘submerged’). The captured data revealed a user/tweet ratio of an average 1.5 per tweets per user, suggesting a wide distribution of users in conversation during this period [Table 01]. Furthermore the test revealed that a little over three percent of the captured tweets had precise coordinate-level geolocation metadata attached.

Total number of Tweets 150,000
Number of users 100,000
Tweets with precision geo-location 5,000
Average tweet frequency 100 tweets / 60s

The results of the system testing in February 2014 suggested that there was significant opportunity to crowd-source reports of flooding from Jakarta’s citizens via Twitter. The tests also showed that the use of words other than ‘flood’ or ‘banjir’ included a high number of tweets not relevant to flooding. As a result, for the 2014/2015 monsoon season, it was decided to only use these two keywords.

The findings from the system testing were subsequently verified by the award of a Twitter #DataGrant, which made it possible to demonstrate both the spatial and temporal coverage of Twitter activity in relation to flooding in the city of Jakarta. Fig. 05 shows geolocated tweets from the Twitter #DataGrant archive, awarded to the University of Wollongong, related to flooding over the entire 2013/2014 monsoon season. The spatial continuity and geographical coverage of Twitter data during flooding was revealed by the #DataGrant, with over 150,000 geolocated Twitter conversations observed in Jakarta during the 2013/2014 monsoon season. The #DataGrant highlighted the heterogeneity of Jakarta’s online citizens and supported the development of a real-time platform to map reports of flooding using Twitter. It is interesting to note that Fig. 05 also shows underlying urban infrastructure (roads, transport interchanges, airports) derived solely from the locations of Jakarta’s citizens Tweeting from their mobile devices. The geographical extents for PetaJakarta.org were based on the #DataGrant [Fig. 05], and further design parameters for CogniCity were subsequently derived in an empirical manner based on the results of the system test and Twitter #DataGrant analysis. Each of the key design requirements is discussed below.

Fig. 05. Geolocated Tweets Related to Flooding in Jakarta over the 2013/2014 Monsoon Season. Data source: Twitter #DataGrant.

Fig. 05. Geolocated Tweets Related to Flooding in Jakarta over the 2013/2014 Monsoon Season. Data source: Twitter #DataGrant.

2.2.1 Networked Interfaces for Civic Co-Management

One of the challenges facing decision makers using citizen reports for DRM is the verification of submitted information, particularly if this information has been harvested from social media in a passive manner (i.e. without direct contributions from users). Previously verification has been undertaken manually, a time consuming and labour intensive task. In situations where crowd-sourced reports are the primary source of information during a disaster, the classification of reports as verified or not is of critical importance to ensure that decision makers and the public are only interpreting data which is directly relevant to the situation. Filtering approaches are often adapted to a DRM context from existing practices in information theory, where outlying data is removed as a matter of routine (Medina 2011).

To address these issues, previous DRM projects have used online communities to outsource human analysis of reports in order to classify their validity, and even rank their importance, forming online international communities of digital humanitarians (Meier 2014). The large volume of data captured during the system testing, and corroborated by the Twitter #DataGrant, necessitated that PetaJakarta.org develop a new solution to collect verified reports of flooding directly from Jakarta’s citizens, reducing the requirements for manual verification or intensive computation analysis through machine-learning methods. Only in this way was it possible to provide a situational overview for decision makers and the public in a real-time manner. This approach builds on previous crowdsourcing efforts in response to natural disasters that relied on a people-centric approach to gather local knowledge, translate, and classify information (Holderness 2014; Meier 2014), tasks which are challenging to automate and computerize.

Through the PetaJakarta.org platform, the OSS CogniCity used the social media platform Twitter to engage citizens and form a network of ground truth reports about the flood condition. In this “people-as-sensors” paradigm, Twitter users within the system parameters (e.g. geographic location, discussing flooding) were invited to confirm whether flooding was taking place at their location. Users could then see the results of their contributions, as well as those of other citizens, visualised on a map linked to their network (Twitter), in the public domain and in real-time. The map used by both the public and government agencies thus created a two-way communication interface between users, PetaJakarta.org, and the government. Importantly, users were offered the opportunity to report by text or media in an unrestricted manner (within the confines of Twitter platform) on the given situation, a method which has been historically proven as more effective at providing reliable data than crowd-sourced requests for highly structured reports (Coen 2012). In this manner, PetaJakarta.org formed a self-evolving network of reporters, who through their engagement with the project, the platform, and each other, shared information and refined their reports during the course of monsoon season. This process is evident in the typologies of reports generated by users, developed through a process of communicative evolution, which is further described in Section 3.3.

The process of soliciting active validation from citizen reports was a two-step process; first, users in the city who included the keyword banjir (or ‘flood’ in English) in a public conversation on Twitter were sent an invitation to verify whether they were experiencing flooding. Second, the user was able to verify flooding simply by replying to the invitation—that is, by sending a Tweet to the @petajkt account containing the keyword ‘flood’ or ‘banjir,’ and including geolocation metadata enabled by the Twitter application on mobile devices.

PetaJakarta.org became one of the first projects in the world to programatically invite users to participate in a crowd-sourcing effort using Twitter. Whilst existing infrastructure exists within the Twitter platform to disseminate specific messages to a targeted audience (e.g. through promoted tweets), through this world-first collaboration with Twitter, PetaJakarta.org was able to send every user in Jakarta who mentioned the keyword ‘flood’ or ‘banjir’ during the monsoon season an invititation to confirm the flood situation via a Twitter; these reports were automatically added to the publically available PetaJakarta.org map. The design of the automated invitation and related outreach campaign are discussed in detail in Section 3. Over the course of the 2014/2015 monsoon, PetaJakarata.org sent 89,000 invitations to citizens in Jakarta as a call to action to confirm flood conditions, gaining over two million Twitter impressions over the same period as a result. The results and an evaluation of this process are discussed further in Section 5.1 below.

2.2.2 Real-Time Multi-stable Platform

The key requirement of the interface for the flood reports was the provision of data in a manner accessible to the project stakeholders: the general public and the government of Jakarta. The interface needed to provide actionable information both for decision makers operating at the city scale (e.g. BPBD DKI; Office of the Governor DKI Jakarta, SmartCity Jakarta) and for the public at neighbourhood and street scales. Importantly, these outputs could not be divided into separate products; to allow for transparency between decision makers and users, both groups needed to reference the same map with the same data. To overcome this challenge, the web-map at PetaJakarta.org drew on web-based GIS and web design development to create one interface which was optimized depending on the client device used [Fig. 06]. When viewed on a desktop computer, the web-application scaled the map to show a situational overview of the city. In this mode, CogniCity generates an on-the-fly thematic choropleth representation of the number of reports per administrative area. These choropleth layers were available at one, three, or six hour intervals to match the temporality of the government’s existing internal reporting structure and decision-making processes. Section 4 discusses BPBD DKI Jakarta’s existing practices and information structures relevant to PetaJakarta.org.

In contrast, when viewed on a mobile device, the application scaled the web-map to show the reports at the neighbourhood scale over the past hour only [Fig. 06]. If provided with geolocation data from the user's’ mobile device, the map orientates itself to the user’s location and shows the reports in the immediate vicinity. Based on Tobler’s first law of geography (Tobler 1970), it was assumed that in a DRM context the closest and most recent reports would be of most relevance to the user, enabling the map itself to act as a spatio-temporal filter. In effect, the goal was to distil the Tweet reports on a per-user basis so as to minimise the possibility of overwhelming the user with a larger number of reports while still providing relevant and actionable information.

Fig. 06. PetaJakarta.org: A Multi-stable Cartographic Interface.

Fig. 06. PetaJakarta.org: A Multi-stable Cartographic Interface.

PetaJakarta.org created a user-centric, multi-stable cartographic interface with the aim of providing actionable information to both the government and public. These groups operate at different spatial and temporal scales. However, in the interest of transparency, the same interface needed to be accessible by both user groups. Furthermore, the development of the map through a web-interface suitable for both desktop and mobile devices was key to ensuring the widest possible accessibility and reducing the adoption thresholds to PetaJakarta.org, in contrast to platform specific applications (‘apps’). The effectiveness of a user-centric, multi-stable cartographic interface to the data is discussed further in Section 5.1.

2.2.3 Scalable Geospatial Information System

CogniCity operates a web-based geographic information system using a client-server model. A request is issued by the client device for relevant data, such as the number of tweets per municipal area for the past hour. This request is received by the CogniCity server module, computed in real-time, and returned to the client (i.e. PetaJakarta.org). Rendering of the data on the map interface is the completed on the client device. This structure of the IT architecture was adopted for two reasons; first, so that CogniCity can provide streams of geospatial data in real-time which can be used by any client, not just PetaJakarta.org. Second, rendering data on the client device reduces the computational overhead of server processes, and negates the need for the server to generate image based map tiles, which would be computationally-intensive to achieve in near-real time. In this configuration, CogniCity becomes extremely scaleable and affords the opportunity to use cloud services to perform load balancing and on-the-fly server creation to help distribute the load of large numbers of client requests.

CogniCity provides data to the web-interface by means of a geospatial Application Programming Interface (API), operating in a RESTFul manner. The API provides Uniform Resource Locator end-points to current and archived locations of flood reports, aggregate counts of activity over time, and hydrological infrastructure layers. Client-side JavaScript is then used to connect to and request data from these endpoints when the user loads the map. Data is delivered using the GeoJSON format, or optionally the compressed TopoJSON format to reduce the volumes of data from CogniCity server to client. One of the key requirements of structuring the architecture in this way—using a client-server model—was to support the consumption of the data within the DRM information ecosystem, beyond of the standard PetaJakarta.org interface. The open source Quantum GIS Platform is widely used to assist DRM in Southeast Asia and supports loading of geospatial data using web protocols in the GeoJSON format. As such, data stored within CogniCity for the PetaJakarta.org project can be loaded into QGIS, helping to support the use of this data by decision makers (e.g. BPBD DKI) within their existing GIS infrastructure. The technical specification of the CogniCity API is discussed below in Section 2.3.3.

CogniCity is capable of collecting data from either the Twitter Public Stream API or the commercially-available Twitter PowerTrack product [Fig. 02]. The Public Stream API is suitable for lower volume applications, which expect a fewer number of Tweets. However, in relation to PetaJakarta.org, the advantages of using PowerTrack over the public API were threefold: 1) the PowerTrack API provides access to all relevant tweets and is not subject to rate limiting, whereas the public API is limited; 2) PowerTrack hosts advanced filtering functionality, to search for keyword terms (e.g. ‘banjir’), reducing computational overhead for local filtering of tweets, as compared to using the Public Stream API; 3) The PowerTrack API includes advanced connection monitoring and technical support, helping to reduce local monitoring requirements and facilitating a stable connection and data-flow.

In contrast to the incoming data, programmatic invitation tweets are sent out via CogniCity using the public Twitter REST API [Fig. 07]. The connection to this API was perfomed using the credentials of the verified @petajkt Twitter account, with support from Twitter, who removed the limits on the number of tweets which could be sent from the account to enable the automated delivery of high-volumes of invitation messages.

In reviewing the proposed design following the system testing in January 2014, user anonymity in the reporting process was embedded within CogniCity and PetaJakarta.org system. Whilst the data produced by Twitter reports of flooding is in the public domain (albeit with the restriction of registering as a Twitter user), the objective of PetaJakarta.org (and CogniCity) was not to create an archive of users who submitted potentially sensitive reports about flooding events, outside of the Twitter platform. In addition to privacy issues related to the collecting and storing information about individual users, the Twitter development guidelines specifically call for respect of the user’s privacy. For these reasons, CogniCity was designed, as part of its original specification, to anonymise reports collected by separating reports from their respective users. Furthermore, the text content of tweets is only stored when the report is confirmed, that is, when the user has opted to send a message to the @petajkt account to describe their situation. Similarly, when usernames are stored, they are encrypted using a one-way hash function. The open source nature of CogniCity is a key design factor in this respect. Publicly sharing the source code means that anyone can examine the software to see how unconfirmed reports are discarded and how the one-way encryption process of usernames operates. Through an interrogation of the code, it is possible to see CogniCity’s objective of creating derivative geospatial products based on Twitter data, which contain actionable information for the public and government in response to flooding.

2.3 Technical Specifications

CogniCity is organised into four components [Fig. 07]; the ‘Reports’ module is responsible for collecting data from Twitter and sending invitation tweets. The ‘Server’ module provides the real-time data API, and serves the third component, the ‘Client Interface’ (PetaJakarta.org), which provides access and visualisation to the data. The final component is the CogniCity database, which underpins the entire CogniCity software stack. The database runs on the open source PostgreSQL object-relational database management system and uses the PostGIS extension to support geospatial data, including the locations of Tweet reports. For the 2014/2015 monsoon season, the reports module, the server module, and the database were deployed operationally in the cloud using Amazon Web Services. All of the cloud instances were located in Amazon’s Singapore datacentre to reduce latency for the target audience located in Jakarta. The following section describes each of the modules and their technical specifications. Following this, Section 2.3.5 provides an overview of the corresponding metadata catalogue, and the PetaJakarta.org Major Open Data Collection from the data collected during the 2014/2015 monsoon season. References for CogniCity’s source code and documentation are included in Appendix I.

Fig. 07. Detailed CogniCity System Architecture.

Fig. 07. Detailed CogniCity System Architecture.

2.3.1 CogniCity Database

The CogniCity database is comprised of 14 database tables [summarised in Appendix II], populated by the flood reports created from the tweets received by the CogniCity Reports Module [Fig. 05]. All tables are independent of each other, except for the ‘all_users’ table which is populated by a trigger function each time an entry is added to the ‘tweet_invitees,’ ‘tweet_users,’ or ‘nonspatial_tweet_users’ tables by CogniCity Reports. The entry of data into the database by the Reports module, and retrieval of information by the Server module, is discussed below in Sections 2.3.2 and 2.3.3, respectively.

2.3.2 CogniCity Reports Module

The Reports module is CogniCity’s interface to Twitter. The module collects relevant tweets, adding them to the database as flood reports, and sends tweets to users inviting them to confirm the current flood condition. CogniCity divides Tweets about flooding into two categories; unconfirmed reports are Tweets from Jakarta, with geo-location metadata, that contain the specific keyword ‘flood’ or ‘banjir.’ Keywords can be either a hashtag (e.g. #banjir), or part of the tweet text (e.g. ‘Jakarta is flooding’). These reports are added to the database table ‘tweet_reports_unconfirmed’ [Appendix II], and subsequently added to the map. However, as these reports are unconfirmed (e.g. the tweet could refer to a different point in time or location), only the geolocation metadata of the Tweet is stored in the database. Upon receipt of an unconfirmed report, the Reports module programmatically sends the user a Tweet in reply, inviting them to confirm the situation. Once a user has been sent an invitation, their username is hashed using a one way encryption function and added to the ‘tweet_invitees’ database table. This ensures that a user never receives more than one programmatically-generated invitation Tweet; before each invitation is sent, the username is compared against the existing list in the database.

Fig. 08 shows an unconfirmed report, detected by the CogniCity Reports module, which then issues an automated invitation Tweet to the user. The invitation message aims to inform the user about the project and provide instructions on how to submit a confirmed report. The invitation Tweet is limited to 109 characters to allow for Twitter usernames in the reply; however, by using Twitter Cards, the automated Tweet could also contain a short embedded video which explained the project and how to participate. The design of the video is discussed further in Section 3.2.2. Invitation Tweets were either sent in English or Bahasa Indonesian depending on the language as defined in the metadata of the tweet (unconfirmed report) received by CogniCity:

“Flooding? Enable geo-location, tweet @petajkt #banjir and check petajakarta.org”

CogniCity Invitation Tweet sent from @petajkt in English

“Kena banjir? Aktifkan geolokasi. Laporkan ke @petajkt #banjir. Cek di petajakarta.org"

CogniCity Invitation Tweet sent from @petajkt in Bahasa Indonesian

Fig. 08 also shows the user’s response to the programmatic invitation Tweet, in the form of a confirmed report containing a photo of the flood condition, approximate height of the flood waters, and user’s location by ‘Rukun-Warga’ (RW) and ‘Rukun-Warga’ (RT) municipal boundary numbers. Upon receipt of a confirmed report as shown in Fig. 08, the Reports module adds the Tweet, text, and geolocation metadata to the ‘tweet_reports’ database table, enabling the report to be plotted on the map at PetaJakarta.org. The hashed version of the username is then added to ‘tweet_users’ table. Finally, the CogniCity generates a thank you message confirming the receipt of the user’s report and directing them to PetaJakarta.org to see their contribution to the map. Fig. 09 shows the process of CogniCity managing incoming Tweets, classifying them as confirmed or unconfirmed, and issuing the required response.

Fig. 08. An Example User Interaction with CogniCity.

Fig. 08. An Example User Interaction with CogniCity; (1) unconfirmed report, (2) programmatic invitation from PetaJakarta.org, (3) confirmed report in reply including photo, (4) programmatic thank you message from PetaJakarta.org.

In addition to confirmed and unconfirmed reports, two other conditions may arise on receipt of tweets by the Reports module. If a confirmed report (i.e. sent to ‘@petajkt’) is received but without the required geo-location metadata to be added to the map it is entered into the ‘nonspatial_tweet_reports’ database table, and a count of users is kept in the ‘non_spatial_tweet_users’ table [Appendix II]. Upon receipt of such a Tweet, CogniCity programmatically sends the user a reminder to enable geolocation and try submitting the report again [Fig. 09]. Unlike unconfirmed Tweets, no username check is carried out prior to sending a geolocation reminder, as the user has opted to include the PetaJakarta.org account ‘@petajkt’ in their Tweet. This strategy was designed so as to not limit the number of reminders sent to help increase the use of geolocation, and as such the number of confirmed reports received. The CogniCity Reports module stores a blacklist of official user accounts such as BPBD DKI Jakarta, who regularly send tweets without geolocation metadata (i.e. from their Incident Control Room), to prevent these accounts from receiving continuous reminders to enable geolocation on their Tweets.

Fig. 09. CogniCity Process of Classifying and Responding to Tweets.

Fig. 09. CogniCity Process of Classifying and Responding to Tweets.

Fig. 09 also shows shows the process when an unconfirmed tweet is received without geo-location data. As the system testing suggested that this would represent the majority of tweets [Table 01], before being discarded CogniCity checks the location parameter of the user profile metadata in the tweet for the key word ‘Jakarta,’ suggesting the user is a resident in the city. If present, the user receives a programmatic invite as per an unconfirmed invitation; if not, then the Tweet is disregarded by CogniCity.

As described in Section 2.2 above, the system testing from February 2014 showed 5,000 tweets over a single day which could potentially be classified as unconfirmed reports, albeit with the six keywords, instead of only “flood” and “banjir” as used for the 2014/2015 monsoon season. To assess the CogniCity Reports module prior to deployment, the system was load-tested using both data from the Twitter #DataGrant and Twitter PowerTrack. At maximum throughput, CogniCity is capable of handling up to 250 tweets per second, with the complete processing of each tweet including classification, writing to the database, and sending of automated replies taking 30 milliseconds, although these values are subject to the platform on which CogniCity is deployed operationally. During operation, errors such as disconnections from the PowerTrack API or the database server are logged and the module attempts to reconnect at specified intervals. Within a five minute window, any missing data is recovered using the PowerTrack backfill functionality, and loss of connection to the CogniCity database results in internal caching in CogniCity Reports until the database connection is restored. Any exceptions which go unhandled result in a notification Tweet being sent to the CogniCity system administrators.

2.3.3 CogniCity Server Module and Database Design

The server module is a custom NodeJS application tasked with providing the API-based data on the reports stored in the CogniCity database. The server module also serves the static components of the client interface (i.e. the PetaJakarta.org website), using the Express framework within NodeJS.

The server module provides access to three groups of data across nine API endpoints [Appendix III]. The reports endpoint provides real-time (i.e. last hour) listing of confirmed and unconfirmed reports, and their geolocation metadata. The aggregates endpoint provides real-time counts of the sum of confirmed and unconfirmed reports from the past one, three, six or 24 hour periods at three different municipal scales. The municipal scale is specified by the client when requesting data and includes the city scale (i.e. the five cities of Jakarta), sub-district scale (i.e. ‘Kecamatan’), village scale (i.e. ‘Kelurahan’), and RW scale (‘Rukun-Warga’). Lastly, the infrastructure endpoint provides access to data representing the location of waterways, pumps, and floodgates in Jakarta, used by PetaJakarta.org as contextual support for the real-time flood map. CogniCity serves the endpoints under the domain of the web interface (i.e. PetaJakarta.org). For example, to access counts of reports from the last hour, the client makes a request to the following URL https://petajakarta.org/banjir/data/api/v1/aggregates/live. For references to complete API documentation for each of the endpoints, consult Appendix III.

Apart from the report-count and report-time series endpoints, all endpoints provide data in the GeoJSON format by default. This provides the data in a format which is readable by the PetaJakarta.org map library LeafletJS and by the open source GIS platform QGIS used by BPBD DKI Jakarta. Optionally, the client can request the data in the TopoJSON format, which is a compressed representation of GeoJSON data. TopoJSON reduces data volume by employing a topological representation to minimise the number of geometric vertices and edges which are required to represent vector geometries. PetaJakarta.org used the TopoJSON library to compress geometries served by the data API, which are then uncompressed (converted back to GeoJSON) on the client. During testing, the TopoJSON format reduced data volumes by 85% and associated data transfer speeds by a third, helping to improve performance of PetaJakarta.org, especially on mobile devices.

To further improve responsiveness of the map when loading data, each of the endpoint requests is cached in computer memory by the CogniCity server application for sixty seconds. This ensures that users are provided with data in real-time, which is no more than sixty seconds old, but reduces the number of individual database queries executed if a large number of users logon to PetaJakarta.org simultaneously. This architecture is particularly relevant for the aggregates layers, which are computed on-the-fly after a request to an aggregates endpoint is received in order to reduce database volume. Whilst this creates increased workload for the database, it reduces database size, removes the need for periodic calculations of aggregates, and ensures that the aggregates data is real-time. The spatial computation of how many tweets are within a specific municipal district at a given time is achieved using a spatial SQL query embedded within the server module. The query uses the PostGIS function ‘ST_Within’ to aggregate and count tweets in the reports and unconfirmed reports tables, with the polygon table of the municipal district scale requested [see Appendix II for listing of CogniCity Database tables]. The process is optimised by employing a Generalized Search Tree (“GIST”) Index on the geometry column for all spatial tables in the database. Furthermore, the unique ID for each polygon geometry at the RW scale is the same primary key system used by BPBD DKI Jakarta, ensuring inter-operability for PetaJakarta.org data with applications within the existing DRM information ecosystem in Jakarta, such as the InaSAFE planning tool, built by AIFDR. As such, the PetaJakarta.org aggregate layers, generated by CogniCity, could be used as hazard exposure layers for future flood mitigation and adaption planning in Jakarta.

2.3.4 PetaJakarta.org Map Interface

The interactive map at PetaJakarta.org was built using the Leaflet JavaScript library with real-time data provided from the CogniCity API. Through the development of the client-side ‘map.js’ JavaScript module, as part of CogniCity, PetaJakarta.org provided a multi-stable cartographic interface for flood reports. Users are divided into two groups: devices with a touch screen are classed as ‘mobile’ (i.e. smartphones and tablets), and those without are referred to as ‘desktop’ devices (i.e. laptops and desktop computers).

For mobile users, confirmed and unconfirmed reports of flooding were added to the map in point form. This cartographic symbology simplified the interface on small screens and reduced the volume of data required when loading the map. Additionally, mobile users only saw the reports from within the last hour, whereas desktop users had the option of seeing aggregate data from the past one, three, or six hours.

Mobile device users were asked to share their location with PetaJakarta.org, which, if granted, zooms the map to the device’s location, showing the reports in the user’s immediate vicinity. Fig. 10 shows a screen capture of the map on a mobile device in November 2014; the user’s location is denoted by the blue pin, and five unconfirmed reports are visible on the map.

Fig. 10. Screen capture of PetaJakarta.org Mobile Interface, November 2014.

Fig. 10. Screen capture of PetaJakarta.org Mobile Interface, November 2014.

Clicking on a confirmed report allowed the user to see the contents of the original tweet, and any included media are hyperlinked back to the Twitter platform. The text from unconfirmed reports is not retained by CogniCity (see Section 2.2.3); instead, the popup for unconfirmed reports contained instructions on how to send a confirmed report via Twitter. If geolocation was not available, or the user was outside of Jakarta, then the map rendered a view of reports at the city scale as shown in Fig. 11. Additionally, to help with user navigation of the data, the map also included a reports tool which shows a listing of confirmed reports ordered by the most recent first. Referring to Fig. 11, one sees that the reports module is notifying the user that there are currently four confirmed reports across the city, marked as blue dots on the map.

Fig. 11. Screen capture of PetaJakarta.org Mobile Interface at City Scale November 2014.

Fig. 11. Screen capture of PetaJakarta.org Mobile Interface at City Scale November 2014.

Rendering of all data is carried out by the Leaflet library on the client. To facilitate a smooth interactive experience when using the map, all current confirmed and unconfirmed reports of flooding in Jakarta are downloaded onto the user’s device when the map is loaded. This enables the user to pan and zoom around the city to see reports beyond their immediate vicinity. The loading of large numbers of points (up to 1,000 during the 2014/2015 season) on mobile devices, over cellular data networks, was facilitated by using the TopoJSON library for geospatial data compression, and HTML5 Canvas layers as the default model for point rendering. Additionally, asynchronous loading of data from the CogniCity API means that the map is interactive before all the data have completed loading. As a result of these intergrated components, PetaJakarta.org provided a useable map, with reasonable performance, even when rending hundreds of reports on low-end touch-screen devices using a cellular data connection.

Fig. 12 (a). Visualisation of the PetaJakarta.org Desktop User Interface at the City Scale.
Fig. 12 (b). Visualisation of the PetaJakarta.org Desktop User Interface at the District Scale.
Fig. 12 (c). Visualisation of the PetaJakarta.org Desktop User Interface at the Neighbourhood Scale.
Fig. 12 (d). Visualisation of the PetaJakarta.org Desktop User Interface at the Street Scale.

Fig. 12. Visualisation of the PetaJakarta.org Desktop User Interface, Showing the Map at the City, District, Neighbourhood and Street Scale.

Desktop users viewed the same map, but instead of point representations of reports, users saw aggregate counts of reports in municipal areas across the entire city. This thematic choropleth symbology showed variations in intensity of twitter activity across the city. Primarily aimed at decision-makers such as BPBD DKI Jakarta, the aggregate interface also allowed users to drill down through the different spatial scales of municipal boundaries, with each layer showing the count of tweets at a finer grain resolution, until reaching street level where individual reports were displayed in a manner similar to the mobile user interface [Fig. 12]. In this way, starting at the city-scale overview, a user could quickly identified hot-spots of activity with high numbers of reports. Fig. 13 shows a screen capture of PetaJakarta.org from flooding in January 2015. Importantly, the desktop user could view the same data as the mobile user, but data were represented first in aggregate form to support decision-making at greater spatial and temporal scales than required by the general public. The only discrepancy between the two interfaces is that the bounding box of tweet capture in the CogniCity Reports module [see Fig. 06] extends beyond that of the Jakarta boundary, meaning that the mobile user saw reports from a larger area than represented by the aggregate layers. However, both interfaces also include overlays of rivers, pumps and floodgates in Jakarta. These layers, which were optionally enabled on the map, enhance spatial navigation of the city by representing the urban infrastructure which makes up Jakarta’s hydrological and hydraulic network. It is anticipated that in the future this network data could be supplemented, for example, to tie reports of flooding to failures of specific infrastructure assets, thereby further enhancing the information available for decision-makers.

Fig. 13. PetaJakarta.org Desktop User Interface.

Fig. 13. PetaJakarta.org Desktop User Interface.

2.3.5 Metadata Records and the PetaJakarta.org Major Open Data Collection

The PetaJakarta.org Major Open Data Collection (MODC) is an initiative supported by the Australian National Data Service (ANDS). In addition to provision of the data via the CogniCity API at PetaJakarta.org, an archive of data from the 2014/2015 monsoon season is available under an open license to support further research. The MODC compliments the open source ethos of CogniCity and includes the aggregate counts of confirmed and unconfirmed reports at the ‘Rukun-Warga’ (RW) scale in hourly intervals, as well as the layers representing Jakarta’s hydrological network. As shown in Fig. 05, these data and their metadata are held in the University of Wollongong institutional repository, referenced by unique digital object identifiers (DOIs). Corresponding records for the data are kept in the National Library of Australia and the Research Data Australia Portal. Appendix IV contains further details of the metadata and links to their records.

2.4 Summary & Recommendations

In summary, through the PetaJakarta.org Joint Pilot Study, CogniCity has demonstrated the feasibility and utility of a web-based, open source, Geosocial Intelligence Framework as a tool to crowd-source geospatial information about flood activity in Jakarta. Using the NodeJS framework on top of a PostgreSQL database, it was possible to build a rapid prototype of a scalable geographic information system capable of capturing Twitter data and serving them through a web-map interface as confirmed and unconfirmed flood reports in real-time. NodeJS, combined with a scaleable cloud deployment proved to be a robust solution for such a system, capable of handing more than 3,000 user requests on PetaJakarta.org within an hour, and consuming upto 240 incoming tweets per second. Furthermore, the use of JavaScript to build the client-side interface using the Leaflet library combined with GeoJSON and TopoJSON data from the CogniCity API suggest that JavaScript is capable of acting as a unifying language for the rapid prototyping of high-volume, low-latency, web-based geospatial applications.

PetaJakarta.org was used by residents of Jakarta to share flood information, call for support, and navigate the city during the seasonal monsoon. The following section provides a summary of the design parameters for the development of PetaJakarta.org as a user-centric platform, detailing the methodology and approach of designed engagement; the communication patterns and typology of reports is subsequently examined in detail; an overall summary and recommendations conclude the review of civic co-management from the citizen-user perspective.

Fig. 14. PetaJakarta.org Launch Event Invitation.

Fig. 14. PetaJakarta.org Launch Event Invitation.

3.1 User Experience & Interface

3.1.1 Branding

In developing the PetaJakarta.org brand as the public face of the project, it was important to ensure that the interface and map were presented as community-owned, rather than as a government product or academic research tool. This concern is directly related to the open source and open data ethos of the project; moreover, because the platform relied on public adoption, it needed to not only operate as a transparent, community-centric information sharing platform, but had to also feel like a community platform [Figs. 14 & 15].

Fig. 15. PetaJakarta.org Launch Poster.

Fig. 15. PetaJakarta.org Launch Poster.

The design of community inclusion can be summarized effectively by three key objectives: (1) draw on existing platforms with similar crowd-sourced city engagement; (2) make compatible with the technology, OS, and resolutions most frequently used for web access; (3) enhance and highlight the crowd-sourced ‘ask’ for the users, and their incentives for participation.

  1. From a user-experience perspective, commercial, crowd-sourced traffic applications (eg. Waze) share a similar logic of public participation; these applications studied in order to understand the expectations of features and language of map-based crowd-sourced tools.
  2. As many of the users were connecting via mobile devices to access both the Twitter platform and the website—with the emphasis on those using mobile GNSS to geo-locate flood reports and check nearby flood reports—the site and its graphic elements were designed to load quickly and legibly at low bandwidth and low resolution.
  3. The community-participation aspect of the project was communicated in a concise, action-oriented form in all of the media produced for the public. “See a flood? Tell us,” is the basic message used to indicate how the public should best engage with the project. The ‘big crowdsourcing’ element of the project was also highlighted in language and visuals (eg. “The more people use Peta Jakarta, the better the map will be”), indicating a user-centric incentive that also encouraged users to share information about the project with their own social media network by retweeting messages from the @petajkt account.

This branding strategy towards community inclusion can also be traced through individual design decisions, including the use of language, color, graphics, and the map interface itself. Aiming to appeal to first adopters—the young, tech-savvy Twitter-public of Jakarta—the language used in all the outreach materials (Twitter replies, the outreach video, graphics, and print advertisements) was intentionally casual and concise. In these communications, a direct action was always emphasised, for example, ‘turn on your phone’s geolocation,’ ‘tell us about it,’ or ‘report a flood’. Because of the repeated recurrence of flood events during the monsoon, and the continuation of daily activities around and through these flood events, the messages were intentionally designed to be more like normal twitter chatter and less like public service announcements. Geolocation pins with speech bubbles were used in advertisements to indicate the connection between the map and community expression of concern. The graphic style and color choices were similarly light and youthful, connecting visually to sites used as part of everyday life rather than as emergency response. [Fig. 15] Graphics and colors were also selected to ensure they were easily identifiable even on a small mobile device screen. A clean, black and white map background with brightly colored report pins was selected as the map interface so that its legibility was optimized on all screen resolutions.

3.1.2 Website

Although much of the public engagement with the project occurs through the Twitter platform, the PetaJakarta.org website hosts the map interface and information about the project. As explained in Section 2.2.2, the map is both a tool for government and public access to real-time flood reports. From the perspective of the citizen-user, the emphasis was to provide easy access to the map and direct information on how to engage with the project. Thus, the public engagement with the project is divided into two actions: first, “share flood information” and second, “view the map.” The website indicates these two actions on the homepage [Fig. 16], allowing the user to easily navigate to via either of these concise instructions to learn how to share information or proceed directly to the map. The PetaJakarta.org home page also includes an embedded copy of the Outreach Video in Bahasa Indonesian and English. The default language of the website is Bahasa Indonesian, allow a main navigation tab allows users to switch the platform to English.

Fig. 16. Screenshot of user instructions from PetaJakarta.org.

Fig. 16. Screenshot of user instructions from PetaJakarta.org.

3.2 Methods of Engagement

3.2.1 Twitter as a Platform

Jakarta has one of the highest number of Twitter users of any city in the world (Semiocast 2012). Because of this high adoption rate, the platform was selected as the primary method of public engagement for the project. Twitter allows users to passively listen to other users, directly engage with single users, and broadcast to a set of self-selected users (followers). Twitter also enables additional functionality for public address, including Twitter Cards (programmatically attaching predetermined media based on terms included in a tweet) and Twitter Video Cards (programmatically attaching predetermined video based on terms included in a tweet). Promoted Tweets can also be used to purchase space on users feeds, although this feature was not used by PetaJakarta.org during the Joint Pilot Study. PetaJakarta.org used a variety of methods of public engagement to notify and incorporate the citizen-users into the project.

3.2.2 Soliciting Civic Participation

Public participation is critical to the project. There were several challenges to encouraging participation within Jakarta’s Twitter community; first, the flood events in the monsoon season are multiple and scattered throughout the city; there is not a single opportunity in the season when public attention is entirely focused on the floods. Because of this, developing strategies for capture public attention on Twitter at the time of impact was important. Second, because PetaJakarta is not a passive social-media scraping system, the public was asked to take action in order to participate. This call to action had to be made clear, yet appear unintrusive and incentivized.

In response to these challenges, a series of low-contact strategies were used solicit public participation in the project. An Outreach Video was attached, via Twitter Video Cards, to any tweet containing the term ‘petajakarta.org.’ And, a series of low-threshold, opt-in requests were delivered via programmatic reply whenever users tweeted keywords (as described in 2.3.2).

The Outreach Video (Video 01) was the main method of public promotion. It is a simple, concise, one-minute video that describes the user experience of PetaJakarta.org and asks the public to participate. The video focuses on the call to action: tweet @petajkt with details when you see a flood, and the aggregate effect of this action: ‘working together we can help everyone bypass flooded areas, saving time and avoiding danger.’ The video uses direct, clear language that is action-oriented. Since the video was accessed through Twitter, and potentially through the Twitter app, it was designed for maximum legibility on a small mobile screen and in noisy urban conditions. The graphics use flat, bright colors on a white background; there is no background music under the calm, direct spoken narrative.

It was important to design the user interaction with PetaJakarta.org to create a user experience that highlighted the community resource element of the project (again, similar to the Waze traffic app), rather than an emergency or information service. With this aim in mind, the graphics and language are casual and light in tone. In the video, auto-replies, and print advertisements, PetaJakarta.org never used alarmist or moralizing language; instead, the graphic identity is one of casual, opt-in, community participation. That is, the communication strategy is not to insist that users should participate, but instead offer users the opportunity to share information through a non-moralizing, opt-in approach.

3.3 Communication Patterns: Overview

During 2014-15 monsoon season, Jakarta residents participated in sharing real-time information about flooding in the city via @petajkt Twitter account, which was co-operated by the CogniCity Reports module and a human project coordinator. Users immediately took @petajkt seriously (even before the account was Verified by Twitter) and started reporting about flood conditions around their homes, offices, and in traffic situations. The account gained significant number of impressions during flood events. In fact, it is clear when the flood events occurred just from viewing at the analytics bar of @petajkt Twitter account. Below is a summary of the tweet activity of @petajkt in the 90-day period from 2 December 2014 (public launch of PetaJakarta.org and the @petajkt Twitter account) to 1 March 2015.

Fig. 17. Chart of Five Flood Main Events for 2014-15 Monsoon.

Fig. 17. Chart of Five Flood Main Events for 2014-15 Monsoon.

As shown in Fig. 17, the highest bars indicate major flood events. The first event is 27 December 27 2014, which marks the first major flood in Jakarta during 2014-15 monsoon season [Fig. 18]. There were 100,494 organic impressions on that day.2 The second flood event is 23 January 2015 [Fig. 19], an event that gained 83,748 organic impressions. The third event is 9 February 2015, which marked the peak of the flood events during 2014-15 monsoon season [Fig. 20]; on that day there were 541,754 impressions and the @petajkt Twitter account gained nearly 5,000 new followers. Flood events continued for several days and there was still a high number of impressions: on 10 February 2015 [Fig. 21] there were 331,151 impressions, and on 11 February 11 2015 [Fig. 22] there were an additional 132,683 impressions.

2 Twitter defines impressions as: “Times a user is served a Tweet in timeline or search results.” https://support.twitter.com/articles/320043-tweet-activity-dashboard.

Fig 18. ‘Banjir’ tweets / Kelurhan on 27 December 2014.
Fig 19. ‘Banjir’ tweets / Kelurhan on 23 January 2015.
Fig 20. ‘Banjir’ tweets / Kelurhan on 9 February 2015.
Fig 21. ‘Banjir’ tweets / Kelurhan on 10 February 2015.
Fig 22. ‘Banjir’ tweets / Kelurhan on 11 February 2015.

Figs. 18-22. ‘Banjir’ tweets / Kelurahan on 27 December 2014 and 23 January, 9 February, 10 February, and 11 February 2015.

3.4 Tweet Typology

There are several important considerations when evaluating the tweets received by @petajkt during the 2014-15 monsoon. First, many users sent tweets without geo-location information on during flood events. However, these tweets nevertheless had valuable information about flood situations and there was location information in the the body of their tweets. Typically, tweets with valuable information such as these were manually re-tweeted because sharing this detailed, time-critical information was deemed valuable to other users, although it lacked precise geo-location data. Additionally, CogniCity would issue a tweet to the user which used a Twitter Media Card to include a picture explaining how to activate the geo-location function for the next report.

Second, due to high number of Twitter users mentioning the term ‘banjir’ in the Jakarta bounding box, the @petajkt account sent a large volume of programmatic replies containing the invitation to tweet @petajkt about flood reports. As noted in 2.0, the invitation is only sent once per user. As a result of these programmatic invitations, @petajkt hit the Twitter rate limit on several occasions. The first time was on 27 December 2014, during a period of heavy precipitation; the second occurrence was on 23 January 2015. Consequently, on both occasions the @petajkt account needed to wait for one hour to be able to send addition tweets or re-tweets. However, once @petajkt was granted the status of a Verified Twitter account, this problem did not occur again.

There were various types of flood reports addressed to @petajkt; in what follows, the three types of tweets received by @petajkt are classified, and this classification is clarified with example tweets.

3.4.1 Tweet Type A—Flood Reports

Flood reports came in several forms. First, flood reports with photos and detailed information, such as the name of the street, village, RT, RW, and the height of the flood, were received [Fig. 23]. In this type of report, flood height information was provided in meters, centimeters, or by the reference to the human body, such knee high, chest high, etc.

Fig. 23. Sample tweet from user @OpickNiy.

Fig. 23. Sample tweet from user @OpickNiy.

The @petajkt account also received flood reports with photos and information, such as the name of the street, village, RT, RW, and the height of the flood, as well as additional requests for needed evacuation supplies, such as portable pumps, boats, or food supplies [Fig. 24].

Fig. 24. Sample tweet from user @SuryaGuang.

Fig. 24. Sample tweet from user @SuryaGuang.

Additionally, the @petajkt account received flood reports without photos, but with useful information such as the name of the street, village, RT, RW, and the height of the flood [Figs. 25 & 26] This type of flood report sometimes came from individual residents, as well as from organized communities who helped residents without cell phones or Twitter accounts to deliver flood information.

Fig. 25. Sample tweet from user @MuktaPutri.

Fig. 25. Sample tweet from user @MuktaPutri.

Fig. 26. Sample tweet from user @gugunmuhammad1.

Fig. 26. Sample tweet from user @gugunmuhammad1.

The @petajkt account also received flood reports from outside the bounding box of Jakarta [Figs. 27 & 28].

Fig. 27. Sample tweet from user @AzizahKurniawa2.

Fig. 27. Sample tweet from user @AzizahKurniawa2.

Fig. 28. Sample tweet from user @bayuhartono289.

Fig. 28. Sample tweet from user @bayuhartono289.

3.4.2 Tweet Type B—Help & Evacuation

The @petajkt account also received requests for help and evacuation support. These messages sometimes originated from government agencies and were addressed to affected residents. Usually, Dinas Sosial DKI (Social Services Agency) tweeted @petajkt to share what they were doing to help flood victims, such as delivering food, building common kitchens, etc. [Fig. 29]

Fig. 29. Sample tweet from user @DinsosDKI1.

Fig. 29. Sample tweet from user @DinsosDKI1.

The @petajkt account also received from organized communities to residents [Figs. 30 & 31].

Fig. 30. Sample tweet from user @inforelawan.

Fig. 30. Sample tweet from user @inforelawan.

Fig. 31. Sample tweet from user @urbanpoor.

Fig. 31. Sample tweet from user @urbanpoor.

Finally, the @petajkt account also helped to relay peer to peer information being shared public on Twitter platform by re-broadcasting relevant information once it was confirmed [Figs. 32 & 33].

Fig. 32. Sample tweets from users @petajkt and @monictika.

Fig. 32. Sample tweets from users @petajkt and @monictika.

Fig. 33. Sample tweet from user @petajkt to rebroadcast requests for shelter information.

Fig. 33. Sample tweet from user @petajkt to rebroadcast requests for shelter information.

3.4.3 Tweet Type C—Reviews

Users also sent reviews and feedback regarding @petajkt and PetaJakarta.org, which helped to spread the word about the project to their followers and alert us to any problematic reports [Figs. 34 & 35].

Fig. 34. Sample tweet from user @satyawinnie.

Fig. 34. Sample tweet from user @satyawinnie.

Fig. 35. Sample tweet from user @elisa_jkt.

Fig. 35. Sample tweet from user @elisa_jkt.

3.5 Summary & Recommendations

3.5.1 Frequently Asked Questions

The most frequent question directed to @petajkt on Twitter was about how to activate the geo-location function for tweets. So far, this question has been addressed manually by sending a reply tweet with a graphic instruction describing how to activate geo-location functionality.

Another question that is frequently asked, often outside the Twitter platform, is about information accuracy and validation. Although ensuring the information sent by Twitter users is challenging, there a number of ways PetaJakarta.org addresses this issue. First, during a flood event, users continuously report areas that are flooded. In one scroll of the mention tab, the PetaJakarta.org Social Media Coordinator can easily see the areas where reports are concentrated. This allows PetaJakarta.org to manually assess the validity by cross-checking other reports from the same district. Second, PetaJakarta.org monitors Twitter accounts that crowdsource and share info about flooding, such as @BPBDJakarta (official Twitter account of Jakarta Emergency Management Agency), @TMCPoldaMetro (official Twitter account of Traffic Management Center Jakarta Metropolitan Police), and @RadioElshinta (official Twitter account of Radio Elshinta, a radio news station) to detect areas experiencing are floods. By monitoring these accounts, additional cross-checking of crowdsourced @petajkt reports is possible. Third, PetaJakarta.org monitors other electronic media, especially television and internet news sites, to follow coverage of flooded areas and cross-check reports. The final aspect for verifying reports is the recognition of active users that frequently tweet to @petajkt with reliable information; as PetaJakarta.org receives valuable reports from Twitter users who actively tweet about floods with right format (using #banjir and enabling their geo-location), the Social Media Coordinator can ‘follow’ these users, and re-tweet their reports, thereby validating their participation and encouraging other followers to make similar reports.

An additional remark on verification is critical. While the @petajkt account sends a high volume of programmatic messages to Twitter users who have triggered replies through their use of keywords, the aim of this ‘big crowdsourcing’ approach is not to generate an equal volume of replies. According to Harlan Hale of USAID (Hale 2015), the most critical resource in disaster scenarios is information; therefore, it is important to encourage users to share flood data that is relevant, well-formatted, and accurate. In the context of social media, this can occur through the mechanisms described above; however, there is also a naturally-occurring filter of self-selection. That is, users who do not have valuable information, or who are not willing to make a detailed report (albeit one not exceeding 140 characters), are disinclined to send replies. In this way, PetaJakarta.org has found that the ratio of invites to confirmed reports is less significant for disaster risk management than is the ratio of high quality reports to low quality reports. Importantly, the overall aim of sending programmatic messages is not to simply solicit a high volume of replies, but to reach active, committed citizen-users willing to participate in civic co-management by sharing nontrivial data that can benefit other users and government agencies in decision-making during disaster scenarios.

3.5.2 Programmatic Reply Feature

The most frequent complaint to the @petajkt account was regarding the programmatic reply feature. The existence of a so-called ‘bot’ on the Twitter platform led to various responses. Many replies to @petajkt about the programmatic replies used humor or ironic criticism to explain that they were not, in fact, in the vicinity of a flood event. However, @petajkt also received a number of positive responses and messages of gratitude from users who were not affected, but who still saw the value of the programmatic messaging.

The programmatic reply sent by CogniCity helpd to awareness about the existence of PetaJakarta.org to Twitter users, and was therefore a valuable asset to the project. Many people discovered PetaJakarta.org because they received a programmatic reply, and many of them tweeted back to report flood conditions after receiving an invitation. Importantly, the programmatic replies also allowed PetaJakarta.org to contact many more Twitter users than any human administrator could hope to reach. During the 2014-15 monsoon season, the @petajkt sent approximately 90,000 programmatic replies.

Unfortunately, there were also several negative aspects that resulted from using the programmatic reply functionality. First, the function sent replies to all Twitter users within the Jakarta bounding box who mentioned the word “banjir” or “flood,” even though these terms may not be related to a DRM context or a flood situation. Because it is not uncommon for Indonesians to use the word banjir figuratively, for example banjir air mata (flood of tears), banjir kenangan (flood of memories), etc., occasionally users received a programmatic reply even though they were not discussing a real flood event.

The second main weakness that was created by using a programmatic reply function has to do with usability on the Twitter platform itself. On mobile devices, auto-replies occupy the Tweets tab because tweets and replies are listed in a single tab. Because most users use their mobile device to check the @petajkt account, they can become frustrated that most of @petajkt tweets in the tab are programmatic invitations, not flood information. However, this is only the case for users who have not ‘followed’ @petajkt; for ‘followers,’ auto-replies are less significant because they would receive tweets from @petajkt in their timeline, which would exclude auto-reply messages. Notably, this is not a problem when people view @petajkt’s timeline from a desktop computer; however, this user experience is an important consideration in the development of the system.

3.5.3 Additional Improvements for Citizen-Users

Several critical improvements for citizen-users are anticipated for the next phase of software and platform development. PetaJakarta.org has already developed a graphic to explain to users how to activate geo-location on their mobile device; improving user understanding of how to activate this functionality is key to improving the geospatial data collected. Second, citizen-users have expressed their desire to see their photographs directly embedded on the PetaJakarta.org map, either as an embedded image or a pop-up that occurs when the cursor hovers over the tweet; during the operation phase in the 2014-15 monsoon, only the tweet text appeared on the map, even if users had included an image in their report. To see the complete tweet, users needed to click on the tweet, which re-directed them to the web-based tweet; however, to reach this image, users had to click through a “Download Twitter” pop-up ad, and usually another generic ad. Because of this interruptive advertising content, it is clear that embedding images and tweets directly would greatly improve the user experience and the efficiency for citizen-users trying to view multiple tweets.

Regarding the PetaJakarta.org website, one improvement would involve adding the date and time stamp on the flood map to assist users who take screenshots of the map and then post them on Twitter or other social media. Tweets with screenshots of the current situation received the highest number of impressions in February 2015, suggesting that users like this type of tweet, and that adding date and time details would help encourage further circulation of the images. Another improvement anticipated for future development is the addition of a time slider bar that would allow users to go back in time and view the map historically to see changes over time; this functionally would be difficult to achieve on a mobile device, although it can be quickly developed for desktop viewing. Finally, many users have asked for more detailed flood height information; such data can we integrated in several ways, including through the additional integration of a Digital Elevation Model in the base layers of the map (see Section 6.1.2), and by pushing flood-affected area data to the web from the Open API of the Disaster Information Management System (DIMS), currently used by BPBD DKI Jakarta (see Section 6.1.1). Until May 2015, DIMS did not provide an open API; however, the Informatics Division of BPBD DKI Jakarta successfully advocated for this functionality, which is anticipated to dramatically improve data-sharing within the DRM information ecosystem in the coming 2015-16 monsoon season and beyond.

A critical element of the PetaJakarta.org Joint Pilot Study was the sustained engagement with the Jakarta Emergency Management Agency (BPBD DKI Jakarta), who not only visited the SMART Infrastructure Facility at the University of Wollongong for an intensive workshop and schematic prototyping sessions, but also hosted two groups of University of Wollongong Student Flood Support Teams (in November 2014 and January 2015, respectively). The collaboration with BPBD DKI Jakarta allowed the PetaJakarta.org platform to be developed in response to institutional goals, operational requirements, and the existing flowline for data management and information dissemination, thus avoiding an intrusive technological imposition and alleviating the difficulties of adopting and integrating external platforms within an already complex information ecosystem. This section details the government experience of the PetaJakarta.org platform, the methods for integration, communication patterns, and recommendations for further research and development.

Photo 05. The Governor of Jakarta launching the Joint Pilot Study, December 2014.

Photo 05. The Governor of Jakarta launching the Joint Pilot Study, December 2014.

4.1 Government Experience

4.1.1 Global Challenges Workshop at University of Wollongong

While the collaboration between the SMART Infrastructure Facility at the University of Wollongong and BPBD DKI Jakarta was well-established before the Joint Pilot Study workshop, financial support from the University of Wollongong Global Challenges Program enabled a critical advancement of the research collaboration. In October 2014, eight members of BPBD DKI Jakarta travelled from Jakarta to Wollongong, Australia, to participate in a 10 day intensive research workshop at the SMART Infrastructure Facility. The objective of the workshop, co-organized by PetaJakarta.org Investigators Associate Professor Rodney Clarke and Dr Olivia Dun, was the detailed institutional ethnography of BPBD DKI Jakarta’s operational structure, data flowline, objectives and ambitions for the 2014-15 monsoon season. Not only did this workshop built trust between the members of the respective partner organizations, it also allowed for a more considered approach to the software and platform development. Specifically, the workshop allowed for a frank discussion of how and why data moved through the BPBD DKI Jakarta Disaster Information Management System (DIMS), and allowed the SMART team to better understand the operational objectives of integrating social media into the DRM context in Jakarta.

Photo 06. University of Wollongong Global Challenges-funded workshop with members of BPBD DKI Jakarta.

Photo 06. University of Wollongong Global Challenges-funded workshop with members of BPBD DKI Jakarta.

During the workshop, it became clear that social media-sourced data would be used to compliment traditional data sources, and therefore it needed to be formatted in such a way as to optimize this complimentary usage without creating excessive demands on the Agency’s human resources. It also highlighted three main areas of concern: 1) how to best encourage users to share time-critical and non-trivial data and how to ensure this data was accurate; 2) how to ensure the data could be processed in a timely manner; 3) how to sustain the crowdsourcing effort throughout the monsoon season, without losing user interest. The system architecture in Section 2 above details the technical solutions to these problems; however, it is important to stress that the technical resolutions did not develop in isolation (Douglas 1986, Medina 2011). The technical development of the system was a considered response to the operational concerns and existing logic of the Agency; without such an intensive institutional ethnography that included technical, social, and ethical dimensions, it is evident that the project could not have succeeded.

4.1.2 System Design as Collaborative Research & Development

The design of PetaJakarta.org and the CogniCity OSS that underpins the public platform system followed an integrated co-research and development methodology. To optimize the performance of the system, it was necessary to achieve a high degree of understanding regarding the existing operational procedures and protocols of BPBD DKI Jakarta. This understanding is ongoing, as the research phase of the project does not end when the development component begins; however, it should be stressed that the institutional ethnographic element of the research relied on recurrent semi-structured interviews, group discussions, and development meetings with the aim of maximizing feedback during the prototyping and development phases. At the same time, the development goals included an OSS that was highly transferable to other domains, geographies, and languages. These two objectives—building a robust OSS that suited the operational and institutional needs of BPBD DKI Jakarta that was also transferable beyond this immediate use case—worked to parameterize the design process and frame decisions during prototyping and development.

4.1.3 Humanitarian OpenStreetMap Team

The Joint Pilot Study used base maps of the city of Jakarta produced by the Humanitarian OpenStreetMap Team. This allowed PetaJakarta.org to build on a solid foundation of open data, and to work in an information ecosystem already familiar with the role and advantages of open data. The accuracy of the flood map, and the ability to aggregate reports at the various district levels in a multi-stable cartographic representation would not have been possible according to the timeline for development without this exemplary prior data collection and mapping work conducted by Jakarta’s Humanitarian OpenStreetMap Team.

4.1.4 Student Involvement

During the Joint Pilot Study, BPBD DKI Jakarta also hosted two groups of University of Wollongong Student Flood Support Teams, first in November 2014 and again in January 2015. Comprised of undergraduate students from engineering, geography, biology, and media studies, the UOW Student Flood Teams were supported financially by the New Colombo Plan of the Australian Department of Foreign Affairs and Trade, with matching funds made available by the UOW Faculty of Engineering and Information Sciences. The Student Flood Teams worked with BPBD DKI Jakarta, as well as with students from the Center for Applied Geography of the Universitas Indonesia, to collect data relevant to flooding, and survey key infrastructure assets throughout the watershed. As a field research experience, many students expressed sincere gratitude for learning and working with BPBD DKI Jakarta colleagues. Additionally, five undergraduate students have planned to return in anticipation of the 2015-16 monsoon to conduct thesis research that can be used by PetaJakarta.org and BPBD DKI Jakarta. These endeavours have clearly demonstrated the value of field-based learning and first-hand experience that, as one student remarked during the January 2015 visit:

"makes my learning from the classroom actually real, not just an abstract problem, but a real situation where lives are at stake.”

Photo 07. University of Wollongong student with BPBD DKI Jakarta colleague during the November 2014 Student Flood Team deployment.

Photo 07. University of Wollongong student with BPBD DKI Jakarta colleague during the November 2014 Student Flood Team deployment.

4.2 Operational Integration

4.2.1 Launch Event with DKI Jakarta Governor

Since the project’s official launch by Jakarta Governor Basuki Tjahaja Purnama in December 2014, thousands of people have reported flood problems to PetaJakarta.org via their mobile devices. At peak times, PetaJakarta.org handled more than 3,000 users per hour. Critical to the success of the project was its official public launch with, and promotion by, the Governor [Fig. 36]. This endorsement gave the platform very high visibility and increased legitimacy among other government agencies and public users; it also produced a very successful media event, which led substantial media coverage and subsequent public attention.

4.2.2 BPBD DKI Jakarta Operational Usage

The Joint Pilot Study was a positive interaction between PetaJakarta.org and BPBD DKI Jakarta, including the interaction between the Twitter accounts @petajkt and @BPBDJakarta. During the monsoon season, @petajkt helped @BPBDJakarta to disseminate key messages by re-tweeting information, such as water levels, rainfall forecasts, and flood-affected areas. Also, @petajkt retweeted information about the response and relief efforts of BPBD DKI Jakarta and other government agencies, including activities such as helping residents through rescue and relief efforts, draining flood water, installing medical shelters, and delivering food and other aid. In addition, @petajkt manually forwarded relevant reports about floods and evacuation requested to @BPBDJakarta, alerting them to urgent reports.

Ultimately, the sharing of flood information directly on the Twitter platform was somewhat ad hoc; it was difficult to anticipate developing protocols of this type of data sharing in advance of the Joint Pilot Study. Instead, what was critical was developing a rhythm of information sharing, and detailing the challenges and opportunities for new tools that could expedite this process without losing aspects of quality control that are gained by manual checks and human oversight.

Although the PetaJakarta.org API was designed to the specifications agreed upon by SMART and BPBD DKI Jakarta, the Agency determined during operations that the aggregation of the tweets (designed to match the spatio-temporal structure of flood reporting in their system) was still inadequate when looking at social media because it could result in their overlooking reports that occurred in areas of especially low Twitter activity. Importantly, the consumption of aggregate data in real-time via the API was insufficient to meet operational demands of BPBD DKI Jakarta. Instead, the Agency relied on the Twitter stream, and began to sketch out a plan for the development of a more robust decision-support dashboard in the form of a Risk Evaluation Matrix that could help manage this point data more effectively (see Section 6.1.2).

Through their internal analysis of aggregates during periods of flooding, BPBD DKI Jakarta identified that the spatial distribution of points (i.e. density) didn’t relate to the criticality of information in a DRM context. Instead, the Agency recognised that the point representation (i.e. interface as seen on mobile devices) is more useful; nevertheless, such a strategy would still require some representation of change over time to handle the dynamic nature nature of incoming data as well as a valid methodology for the verification of reports. These lessons are critical for the subsequent development of the Risk Evaluation Matrix, discussed in Section 6.1.2.

4.2.3 SmartCity Jakarta

SmartCity Jakarta was developed by Dinas Komunikasi dan Informatika Provinsi DKI Jakarta as a one-stop hub with data feeds on traffic, flood, violation reports and even restaurant locations. Some of the information is interactive and is extracted from third party applications, such as flood information from the PetaJakarta.org open API. This kind of real-time input is not only useful for government coordination and management, but also for urban residents (Townsend 2013). SmartCity Jakarta is an example of public sector innovation wherein the government aims to provide better service delivery through digital applications; making this platform open, sharing the data collected, and fostering further innovations through an open API can call support additional improvements to service delivery.

4.3 Recommendations

4.3.1 Year 2—Operational Integration and Knowledge Transfer

The success of the Joint Pilot Study has led to the extension of the collaboration into a second and final phase—Operational Integration and Knowledge Transfer—to be completed in May 2016. This phase includes three primary components: first, the additional prototyping and development of the CogniCity platform to integrate new features, and to develop a new Risk Evaluation Matrix (REM) as an extension of the CogniCity OSS. Second, the integration of additional API-sourced data consumed from other applications (external to the data gathered from the Twitter PowerTrack API), discussed further in Section 6.1.2. Third, following the successful deployment of PetaJakarta.org 2.0 as an operationally-integrated platform with BPBD DKI Jakarta in the 2015-16 monsoon season, the project completion will necessarily include the complete transfer of the software and project documentation to enable BPBD DKI Jakarta to continue to use and develop the platform for future use regarding flooding, and potentially for other types of deployment.

4.3.2 API-sourced Data Integration

Following the Joint Pilot Study, it is especially clear that one of the most critical elements for success in the DRM sector is the integration of API-sourced data streams (see Sections 4.3.3 and 6.1.2). The data rich environment of Jakarta includes a variety of programs and applications that could share data through open APIs. Critical among these sources are the Disaster Emergency Management System of BPBD DKI Jakarta, as well as the ‘Siaga Level’ aggregation, which includes water level information throughout the flood management infrastructure. Beyond the immediate data flowline of BPBD DKI Jakarta, it is also important to enable the integration of other data sources, including public ‘apps,’ like SmartCity Jakarta’s Qlue and Crop apps, and NGOs applications such as KoBo Collect, developed by UNOCHA for various types of disaster reports. Finally, the integration of additional private sector platforms, such as Detik.com, would allow for a great blend of data sources in a single map and enable more efficient cross-checking across heterogeneous platforms that can validate information of the core data provider, the Twitter PowerTrack API.

As a compliment to increased ICT investment by the Jakarta government, open data allows universal participation where data is available as a whole and and at no more than a reasonable reproduction cost. By providing data under terms that permit reuse and redistribution, including intermixing with other datasets, citizen-users, NGOs, student and advocacy groups, and private developers can generate new tools for information management within the application economy. One excellent example of a government innovation to promote open data is the recently released SmartCity Jakarta website, with its applications Qlue (public) and Crop (government use only). Further investment in projects like SmartCity Jakarta, and the API-enabled data they produce, would also be a welcome element to the DRM information ecosystem.

4.3.3 Decision-Support System Matrix

The Joint Pilot Study, as well as ongoing interviews and meetings between PetaJakarta.org researchers, BPBD DKI Jakarta stakeholders, and other colleagues in the DRM sector have led to the proposal for the development of a more robust Risk Evaluation Martrix (REM) that would enable CogniCity OSS to serve a wider community of users and optimize the data colleciton process (see 6.1.2).

Fig. 36. Launch tweet sent by user @basuki_btp (Governor of Jakarta) to user @jokowi_do2 (President of Indonesia).

Fig. 36. Launch tweet sent by user @basuki_btp (Governor of Jakarta) to user @jokowi_do2 (President of Indonesia).

For the purposes of data collection and evaluation of PetaJakarta.org the monsoon season was defined as the period between 24th December 2014 and the 1st of March 2015. This period included five key flood events, manually identified by observations on the ground in Jakarta. These events corresponded with increased Twitter and PetaJakarta.org activity as shown in Fig. 15. This section provides an overview of the response and use of PetaJakarta.org by the public and the Jakarta government over this period, focusing on the five floods. Public interaction is reviewed using metrics from Twitter for the PetaJakarta.org account @petajkt, the CogniCity database, and from Google Analytics for PetaJakarta.org. Where possible erroneous analytic measurements from the media coverage following the project launch in Jakarta on 2 December 2014 (which led to numerous tweets to the PetaJakarta.org twitter account), and from ongoing system testing and calibration during the monsoon have been removed. The discussion of system use by BPBD is based on follow up discussions and interviews with colleagues at BPBD DKI Jakarta.

Photo 08. PetaJakarta.org Team, including University of Wollongong Student Flood Support Team and members of BPBD DKI Jakarta and Twitter, following the launch at City Hall, December 2014.

Photo 08. PetaJakarta.org Team, including University of Wollongong Student Flood Support Team and members of BPBD DKI Jakarta and Twitter, following the launch at City Hall, December 2014.

5.1 Critical Successes

In what follows, we provide further details regarding the elements of the Joint Pilot Study that were deemed most successful by the research team and the project stakeholders. The aim of this section is not, however, to list project achievements, but to highlight ‘what works’ for further research, development, and deployment in the DRM sector. Below we divide these remarks by focusing first on the social media platform Twitter, then on the web-based cartographic dimension of the project, and finally by analyzing how the Jakarta government valued the project for operational decision-support.

5.1.1 @petajkt Twitter Account

Throughout the extent of the monsoon season the PetaJakarta.org project recieved and mapped 1,119 confirmed reports of flooding. These reports were formed by 877 users, indicating an average tweet to user ratio of 1.27 tweets per user. A further 2,091 confirmed reports were received without the required geolocation metadata to be mapped, highlighting the value of the programmatic geo-location ‘reminders’ sent by the CogniCity Reports module as discussed in Section 2.3.2. With regard to unconfirmed reports, PetaJakarta.org recorded and mapped a total of 25,584 over the course of the monsoon. The difference between the number of unconfirmed and confirmed reports highlights the importance, and validity, of the user-based confirmation system to filter out Twitter conversations which are not indicative of flooding at the user’s location. Confirmed reports were largely generated in response to the programmatic invitations messages sent by CogniCity. A total of 89,000 invitation messages were sent (one message per user), generating over 2.2 million twitter impressions as a result. Note that the number of invitations is significantly higher than the number of unconfirmed reports, because CogniCity issues a programmatic invitation for unconfirmed reports for tweets without geolocational information but does not record them [Fig. 07].

In association with Fig. 14 the results in Table 02 show that Twitter activity corresponded with the peak inundation events on the 9th and 10th of February, with the public contributing 385 and 204 confirmed reports to the map on these days respectively. These dates also saw the highest number of users engaging with tweets from the PetaJakarta.org account, clicking on the embedded video card and re-tweeting the message to their followers on the network [Table 02]. In contrast overall engagement rate appears to reduce over the five events, although there are two few data to support a trend in this instance. In analysis of these data over the monsoon, the relationship between Twitter activity related to flooding, as measured by twitter analytics metrics in Table 02, and the number of unconfirmed, and confirmed reports about flooding received by PetaJakarta.org is also evident. While it is self evident that the number of impressions must be linked to the number of programmatic invitations sent (of which unconfirmed reports are a subset), it would appear that number of confirmed reports is proportional to the the number of impressions, and related sharing metrics such as retweets [Table 02].

The numbers in Table 02 suggest that the programmatic invitation process performed by the CogniCity Reports module is capable of reaching large numbers of users, who are sending flood related tweets in Jakarta (as measured by overall impressions for the five flood events in Table 02). However, it is also clear that the process successfully generates a filtered subset of useful and actionable reports as shown in the previous examples [e.g. Fig. 06] that are added to the map, and which are proportional to the scale of the event’s coverage on social media. This is significant given the number of impressions and total unconfirmed reports over the monsoon would suggest the potential for the map to be overwhelmed by user conversations not relevant to the current flood situation. One critique of the aggregate layer in this regard is that at times the thematic scales of color on the aggregate map interface, capped at ‘30+’, were overwhelmed by the count of unconfirmed and confirmed reports during these key flood events [Fig. 14], suggesting that the scale need be extended to accommodate greater numbers. The ratio of filtering between unconfirmed and confirmed reports is approximately consistent over all five events, with between 20 and 30 unconfirmed reports for every one confirmed [Table 02].

Source Metric 27-Dec-14 23-Jan-15 9-Feb-15 10-Feb-15 11-Feb-15
Twitter Analytics (@petajkt account) Impressions 100,494 83,748 541,754 331,151 132,683
Engagement Rate 4.60% 4.00% 4.10% 3.20% 3.10%
Link Clicks 922 738 4,700 2,000 697
Retweets 456 266 1,800 922 358
PetaJakarta.org Unconfirmed Reports 1,228 2,513 7,971 5,293 1,868
Confirmed Reports 66 118 385 204 64

5.1.2 PetaJakarta.org Website

The PetaJakarta.org website was monitored using the Google Analytics suite. Over the course of the 2014/2015 monsoon season from December 2014 to March 2015 the PetaJakarta.org Google Analytics reported that the website was visited more than 97,900 times. Over the four month period new users consistently made up approximately 70% of these visits (approximately 68,530), with the remaining 30% of visits from returning users. These figures would suggest that user outreach predominantly achieved through the programmatic Twitter invites and embedded PetaJakarta.org video in response to unconfirmed reports (i.e. conversations about flooding on Twitter) was succesful in reaching different groups of new users as different geographic regions of the city were innundated [Fig. 14]. This is supported by correspondence between number of pageviews and key flood events.

For the majority of the monsoon season the number of website visits was less than 500 per day, however on the days of 9th and 10th February [Fig. 15] the number of views increased by over 4,200% (9th: ~21,200 visits, 10th: ~21,300 visits). In relation to user location Google Analytics only received location information from approximately 50% of users, with between 31 and 37% of this cohort accessing the website from within Jakarta. Importantly, more than half of users were recorded as accessing the website from a mobile device (60% of users). Demographic information, created independently by Google based on user web-traffic showed that the gender of PetaJakarta.org users was approximately equal, with estimated gender divisions of users for females and males at 45 and 55% respectively.

5.1.3 Government Usage: BPBD DKI Jakarta

As noted in 4.2.1, although the PetaJakarta.org API was designed to the specifications agreed upon by SMART and BPBD DKI Jakarta, the Agency determined during operations that the aggregation of the tweets (designed to match the spatio-temporal structure of flood reporting in their system) was still inadequate when looking at social media because it could result in their overlooking reports that occurred in areas of especially low Twitter activity. Instead, the Agency used the @petajkt Twitter stream to direct their use of the map and to verify and cross-check information about flood-affected areas in real-time. While this use of social media was productive overall, the findings from the Joint Pilot Study have led to the proposal for the development of a more robust Risk Evaluation Matrix (REM) that would enable CogniCity OSS to serve a wider community of users and optimize the data collection process through an open API (see 6.1.2).

In this regard, it is disconcerting that a recent report prepared by the Humanitarian OpenStreetMap Team and commissioned by the Australia-Indonesia Facility for Disaster Reduction (Reeves 2015) confuses the use of Twitter-sourced data as a research tool with its existing integration into the DRM information ecosystem. The confusion in the report appears to derive from a misunderstanding of how the Twitter-sourced data is used operationally by BPBD DKI Jakarta, as well as a misperception about the role of open API-sourced data integration as a goal of the Agency. While reports about the potential value of social media-derived data can provide insights for innovation, it is critical that studies provide accurate information and precise detail to avoid spreading misinformation within the DRM sector. Because Reeves (2015) fails to grasp broader institutional goals and partnerships—no doubt because the report was prepared at such a distance from the real, on the ground concerns of the Agency—he creates a false impression of the operational mandate and current developments within the Agency. Such errors can be avoided by undertaking analysis of the actual agencies and collaborations involved in DRM projects; the total absence of this consideration in Reeves (2015) implies that flood response is simply a technical problem, thereby reifying an antiquated, “technology-centric” approach to DRM that undervalues substantial efforts on behalf of the agencies and their partners, who have already committed to developing a more holistic approach to flood detection, management, and response.

5.2 Improvements

In this section, the objective to is provide more detail regarding the four key aspects of the project that could be dramatically improved by rather straightforward emendations within the existing DRM information ecosystem. In Section 6.0, we provide a more comprehensive report on the lessons learned, as well as the most critical areas of research and development for advancing the use of social media for DRM via OSS.

5.2.1 Open Data

With the support of the Australian National Data Service (ANDS), PetaJakarta.org undertook a Major Open Data Collection during the 2014-15 monsoon season in Jakarta. As noted in 2.3.5, the MODC enabled the provision of the data via the CogniCity API at PetaJakarta.org, and an archive of data from the 2014-15 monsoon season is available under an open license to support further research. The MODC compliments the open source ethos of CogniCity and includes the aggregate counts of confirmed and unconfirmed reports various scales in hourly intervals, as well as the layers representing Jakarta’s hydrological network.

5.2.2 Application Programming Interface

While the PetaJakarta.org open API was used successfully by BPBD DKI Jakarta and the SmartCity Jakarta project, it is critical to the advancement of the project to continue to develop an ethic of open source and open data within the DRM sector in Indonesia; such an approach would necessarily include the continued sharing of collected data, including social media derived data and flood infrastructure layers. A major improvement in the next phase of the project will be the availability of an open API from the Disaster Information Management System (DIMS), currently used by BPBD DKI Jakarta (but until May 2015 without an open API). This will facilitate greater information sharing and data coordination among PetaJakarta, the Agency, and the public.

5.2.3 Embedded Images

As noted in Section 3.5.3, citizen-users support the system development of directly embedding photographs on the PetaJakarta.org map, either as an embedded images or pop-ups that occur when the cursor hovers over the tweet; we noted above that during the operation phase in the 2014-15 monsoon, only the tweet text appeared on the map, even if users had included an image in their report. To see the complete tweet, users needed to click on the tweet, which re-directed them to the web-based tweet; however, to reach this image, users had to click through a “Download Twitter” pop-up ad, and usually another generic ad. Because of this interruptive ad content, embedding images and tweets directly will greatly improve the user experience and the efficiency for citizen-users trying to view multiple tweets from a mobile device.

5.3 Research & Development

In this section, we highlight research and development goals for the project, moving from the platform of the Joint Pilot Study, to the OSS which underpinned the platform, and finally to the methodological approach which motivated the research undertaken during the study.

5.3.1 PetaJakarta.org

A world-first collaboration initiated by the SMART Infrastructure Facility, University of Wollongong, with BPBD DKI Jakarta and Twitter Inc., PetaJakarta.org has been used effectively by citizen-users and government agencies for flood risk identification and response. PetaJakarta.org also received a substantial degree of Indonesian and international media attention, and attracted a number of additional partner organizations, NGOs and community advocates. The overall goal for the next phase of the project—Operational Integration and Technology Transfer—is to complete the OSS development for PetaJakarta.org 2.0 and effecively deploy the system with the continued support of the project partners during the 2015-16 monsoon season in Jakarta. Following this deployment, it is the objective of the research team to also complete the transfer of the OSS and attendant tools and documentation to the Agency for future use, deployment and development by May 2016.

5.3.2 CogniCity

The use of mobile devices for identifying risk and coordinating disaster response is well accepted and has been scientifically proven as a critical element in DRM. As new tools, applications, and software are adopted by municipal governments and NGOs for the identification and management of urban risk, the need for greater integration of the various data they aid in collecting becomes acute. While the challenge of integrated data management is substantial, it is aided by the fact that many new tools have been developed to include an Application Programming Interface (API), which allows the machine-to-machine (i.e. automated) sharing of open data. While some proprietary platforms for the management of urban data are currently available, they are extremely costly and very limited in terms of data inputs; to date there are no OSS tools for the integrated management of various API sources. Therefore, a key extension of the CogniCity OSS for improving disaster risk identification and management is the development of an integrated Risk Evaluation Matrix (REM) that enables: 1) automated integration of multiple API sources into a low cost, user-oriented dashboard; 2) backend database and software design for the Risk Matrix that enables data sources to be parameterized and interrogated; 3) the development of an output API stream that allows additional secondary applications to optimize their evaluations and analyses through open access to critical risk information.

5.3.3 GeoSocial Intelligence Framework

As noted in 1.1.2, GeoSocial Intelligence can be understood as an evolution within the DRM information ecosystem because it leverages both the inherent capabilities within ubiquitous mobile devices (i.e. GNSS-enabled messaging) and the network capabilities of social media through CogniCity OSS to provide validated and actionable information for citizens and government agencies, thereby improving situational knowledge and increasing response times in disaster scenarios. Further research and development of the GeoSocial Intelligence Framework will require additional testing and operational deployments in Jakarta, but also in other megacities in the region, as well as other coastal or deltaic cities with high flood risk.

The Joint Pilot Study for the PetaJakarta.org project was operationally active from December 2014 to March 2015; during this time, the project enabled Jakarta’s citizens to report the locations of flood events using the social media network Twitter, thereby contributing to a web-based, publicly accessible, real-time map of flood conditions at PetaJakarta.org. These data were used by BPBD DKI Jakarta to cross-validate formal reports of flooding from traditional data sources, supporting the creation of information for flood assessment, response, and management in real-time. The findings of the PetaJakarta.org Joint Pilot Study offer scientific evidence that prove the value and utility of social media as a mega-city methodology for crowd-sourcing relevant situational information to support decision-making and response coordination during extreme weather events. The key lessons from the Joint Pilot Study are summarized here for decision makers; following this summary, we provide several additional recommendations for more specific research and development that is anticipated to occur in the final year of the project, in preparation for the 2015-16 monsoon season.

6.1. Recommendations

6.1.1 Recommendations at a Glance

In order to provide the most clear and precise recommendations derived from the PetaJakarta.org Joint Pilot Study for using social media in the DRM context, here we offer a list of critical recommendations for stakeholders.

  • DRM projects seeking to engage social media platforms and their users should avoid proprietary (“blackbox”) software packages; instead, these projects should support Open Source Software (OSS), provide open data, and offer open Application Programming Interface (API) streams whenever possible to maintain transparency and foster further development and innovation of the DRM software ecosystem.

  • In this context, OSS should be designed for scalability and transferability with respect to the domain of application, the location, and the language of the users. Tools and platforms built for single-use applications are both costly and inefficient; for the DRM sector to advance social media usage and crowd-sourcing potential most effectively, investments should target scalable and transferable OSS platforms that can be brought up to enterprise-grade performance in a timely manner.

  • As the leading social media platform for real-time information sharing, Twitter offers a variety of functional elements that should be more thoroughly leveraged in the DRM sector; these functionalities include account verification, ‘retweet validation’ of citizen reports, Twitter Cards, programmatic reply functionalities, and the PowerTrack API Connection. While Twitter has been used effectively to analyse disasters offline after the event, and to source data in an ad hoc manner during disaster events, importantly, decision makers and DRM agencies do not need to wait for disasters to afflict their cities before developing these elements to support their existing DRM information ecosystem. Building robust social media strategies as an element of preparedness is key to using social media during response, rescue, and recovery efforts.

  • DRM OSS for social media integration should be built with the aim of complimenting existing institutional frameworks and offer an open API for further integration into DRM information ecosystems; when possible, the storage of social media-sourced data should adapt to standard metadata formats such as the Common Alerting Protocol (CAP).

  • DRM OSS should be developed incrementality through an iterative, co-research process that involves the widest variety of stakeholders, including DRM agencies, government managers, scientific researchers, industry partners, and citizen-user groups. By developing modular OSS components through co-research, user feedback can be effectively integrated into both prototyping and development processes, thereby optimizing user feedback as the driver of better design and functionality.

  • DRM OSS platforms for integrating social media and crowdsourced data should not be developed in isolation from other potential data sources; developing a more robust integration of social media data also means leveraging other potential data sets to increase the intelligence produced by the system through hybridity; these other sources could include, but are not limited to, government, private sector, and NGO applications (‘apps’) for on-the-ground data collection, LIDAR or UAV-sourced elevation data, and fixed ground control points with various types of sensor data. The “citizen-as-sensor” paradigm for urban data collection will advance most effectively if other types of sensors and their attendant data sources are developed in concert with social media sourced information.

  • DRM OSS cannot be developed solely as research tools; investments should target applied research projects that use scientific research to aid in operational development, and operational challenges to drive scientific innovation; such an approach removes the redundancies of research and development which can prevent the best tools from reaching the users who could most benefit from their deployment. By fostering co-research among industry, government, and academia, scientific research can advance apace with industrial innovations and the needs of both government agencies and citizen-users.

  • Public engagement and transparency are invaluable assets for the development of meaningful, effective, and sustainable DRM OSS platforms and tools; engagement must be nontrivial and demonstrate the value of opt-in projects for participating users. Social media platforms offer a ready-made site for the advancement of ‘big crowdsourcing,’ allowing DRM OSS to move beyond passive spatial and temporal data mining to active and responsive crowdsourcing and validation of user-generated data.

  • Because information is a valuable resource for both citizen-users and government agencies in disaster situations, the objective of DRM OSS data visualization should be the ease of use and clarity of information; an emphasis on useable design should be the principle concern for all components of data visualization.

  • Mapping initiatives that move beyond traditional sources are especially valuable for gathering critical data to aid first responders and rescue workers; supporting mapping projects in high risk communities is therefore a critical area of investment if DRM OSS projects will succeed in reducing risk over the long term. In this regard, further investment in the Humanitarian OpenStreetMap Team mapping efforts is an especially promising trajectory in the development of data-supported resilience within the DRM information ecosystem.

6.1.2 CogniCity in the DRM Information Ecosystem

As explained in greater detail above, CogniCity is a GeoSocial Intelligence Framework for urban data; more specifically, CogniCity is a geographical information system that allows collection and visualization of geospatial data on flood alerts (via Twitter) to support decision making for both government and citizen stakeholders.. To extend CogniCity’s capabilities beyond the Joint Pilot Study, the research team proposes to develop an Application Programming Interface (API)-enabled Risk Evaluation Matrix (REM). Through this additional OSS, developed in collaboration with BPBD DKI Jakarta, CogniCity will interface with various APIs, including UN OCHA’s KoBo Collect (currently in development for Rapid Assessment, Joint Needs Assessment, and Damage and Loss Assessments), the Jakarta Disaster Information Management System (DIMS), Detik.com, and Twitter, (notably, additional APIs can be added without additional development costs) to create an integrated, OSS REM with a web-based GIS interface. CogniCity will also push this vital data through an open API to various applications, including the Australia-Indonesia Facility for Disaster Reduction’s InaSAFE tool, as well as the World Bank-supported JakSAFE Damage and Loss Assessment Tool (currently in development) and the Pacific Disaster Center’s DisasterAWARE platform; access to this open data through the REM API will enable accurate, efficient, and cost-effective risk identification and analysis.

While the value of both risk identification and disaster risk management mobile applications is widely accepted, the variety of data sources, the variation of metadata structures, and the different configuration of data processing from such applications make the effective coordination of these data sources a key area for building greater resilience. To reap the benefits of accessible, low cost mobile devices and their various applications, BPBD DKI Jakarta require an OSS Risk Evaluation Matrix for the management of API-derived data. CogniCity is Geosocial Intelligence Framework currently used to deliver PetaJakarta.org—a decision support system developed by the SMART Infrastructure Facility, University of Wollongong, with BPBD DKI Jakarta based on citizen reporting of flood events via social media. By extending CogniCity to gather data from additional API sources so as to deliver field reports from other sources in a real-time manner, the system can further aid BPBD’s decision making during flood events.

Such a system would leverage CogniCity’s mapping capabilities to present additional reports via the existing map interface used by BPBD DKI Jakarta. In this way, CogniCity would enable critical data integration, by extending the information ecosystem to fuse data gathered from citizens via social media and formal reporting via API-derived sources into a single Risk Evaluation Matrix. Critically, the design of the OSS REM would also include the development of a robust output API for use by secondary users, as well as a feature to export the data in Common Alerting Protocol (CAP), used by the Indonesian National Emergency Management Agency (BNPB) through their InAWARE DRM software, developed by the Pacific Disaster Center.

6.1.3 Drone-sourced Digital Elevation Models

The rapid advancements in civilian unmanned aerial vehicles (UAVs or drones) has led to a proliferation of drones and amateur drone pilots in cities. From an urban management perspective, hundreds of thousands of images with exchangeable image file format (EXIF) metadata (including latitude, longitude and altitude) are being generated from these UAVs. By aggregating these images and processing them with a series of computer-vision algorithms across an entire city, digital elevation models (DEM) can be built, maintained, and constantly updated. DEM’s can then be combined with other information on urban topography, such as tweets containing information on flood heights, to construct additional datasets. For example, when 3-dimensional elevation data is overlain with 2-dimensional tweets of flooding, flooding extents and depths can be quickly determined. A civilian drone-generated DEM thereby offers an entirely novel urban dataset and an innovative means for developing civic co-management.

6.1.4 Location-based Citizen Alerting

More than half of the users engaging with PetaJakarta.org used an internet-enabled mobile device (see Section 5.0). This figure is expected to grow with the continued development of mobile technology and ubiquitous computing in Southeast Asian markets. Importantly, in addition to an internet connection, most devices are also capable of determining the user’s geospatial location. This feature was harnessed by PetaJakarta.org to show the reports in the immediate vicinity of the user’s location, in effect creating a real-time spatial data filter. The opportunity exists to build on this feature and extend CogniCity to provide the user with real-time alerts based on their location, warning of them of recently reports of flood events. This is particularly relevant due to recent developments in wearable technology, such as smart watches, which can provide notifications at a glance, without the same level of engagement as is required for the user to study a map. Importantly, future research and development in this area should not focus on developing a single mobile device application, but should instead support the development of a range of applications within the civic co-management DRM ecosystem by extending the CogniCity data API with an alerts module. Such development would be supported by further integration of CogniCity with existing DRM processes in the Jakarta government, for example, by enabling BPBD DKI Jakarta to verify reports of flooding using the social media network Twitter (Section 6.1). This extension to the ecosystem would also enable third parties, including the government, the public, and private sector developers, to leverage the geosocial intelligence gathered by PetaJakarta.org to optimise user interaction and situational awareness in the city during disasters through the development of bespoke alert-based mobile applications.

6.2 Jakarta in the ASEAN DRM Community

As the largest capital city in the region, DKI Jakarta can play a significant leadership role in the DRM sector, both in Indonesia and in the wider ASEAN DRM community. Importantly, with the development of the ASEAN Coordinating Centre for Humanitarian Assistance (AHA Center) in Jakarta, there is an even greater potential for transferable research and development related to DRM that can be shared in the region. DKI Jakarta can increase its leadership role by demonstrating improved cooperation between the provincial (BPBD DKI Jakarta) and national (BNPB) emergency management agencies; as this type integration is a challenge throughout the region, Indonesia is poised to demonstrate how effective, applied co-research can help overcome these challenges and achieve greater efficiency through data coordination. Such an approach would promote the further development of data sharing within the sector and across departments through open data, common protocols, and open APIs. DKI Jakarta can also play an exemplary role by investing in in ICT and OSS management tools for DRM, thereby diversifying infrastructure investment to move beyond the merely physical elements of flood management to include the technical tools and platforms for effective planning, preparedness, response and recovery. This approach would necessarily mean moving away from investment in proprietary “blackbox” software packages and instead fostering an open thriving development culture using the best OSS for DRM in the region. Finally, by continuing to work with academic partners like the SMART Infrastructure Facility to develop long term research objectives, DKI Jakarta can begin to leverage the success of the Joint Pilot Study to promote an ethic of co-research that sustains evidence-based decision making for investment in the DRM sector.

The authors would like to acknowledge contributions from the following researchers to this publication: Alexandra Berceanu, Matthew Berryman, Rodney Clarke, Sara Dean, Yantri Dewi, Olivia Dun, Ben Jones, Mohammad Kamil, Robert Ogie, Rhys Powell, Mary O’Malley, Milly Matthews-Mulroy, Alifa Rachmadia Putri, Widya Ramadhani, Frank Sedlar, Ariel Shepherd, Fitria Sudirman, Rohan Wickramasuriya and Albert Yang. This formidable, committed team of researchers have all worked to ensure that PetaJakarta.org could improve safety and resilience for the residents of Jakarta during the monsoon. We are also grateful to the University of Wollongong whose financial support has made this project possible; this includes support from the Deputy Vice-Chancellor (Research) Professor Judy Raper, the UOW Global Challenges Fund, the Faculty of Engineering and Information Sciences, and the SMART Infrastructure Facility. For their generous financial support of the project, we would also like to thank the Australian National Data Service, the Australia- Indonesia Facility for Disaster Reduction, and the Australian Department of Foreign Affairs and Trade. A special thanks to SMART Chief Operating Officer Tania Brown and SMART Research Director Professor Pascal Perez for their continued leadership and support. We are also extremely grateful to the brilliant team at Selera Labs who assisted in the project software development.

We are also grateful for the advice, openness, and friendship of all our colleagues at BPBD DKI Jakarta, as well as allies and supporters in the DKI Jakarta government, the Jakarta SmartCity project, and the Jakarta Timur government; the success of the project is a measure of their commitment and dedication to the residents of Jakarta. At Twitter, we are grateful for support, advice, and our early #DataGrant which propelled the research; we are especially indebted to Twitter’s Director of Academic Partnerships, Mark Gillis, who has been a stalwart supporters since first learning about the project, and Data Strategist Jim Moffat, whose expertise and advice has been critical to the success of the project. We are indebted to the many organization who have supported our data collection efforts and enriched our understanding of Jakarta’s complexity, including the Urban Studies Graduate Program and the Center for Applied Geography at Universitas Indonesia, the Urban Poor Consortium, Ciliwung Institute, Ciliwung Merdeka, and the Humanitarian OpenStreeMap Team. Finally, and most sincerely, we are so grateful to all the users in Jakarta who shared their experiences, concerns, frustrations, and advice on our platform—it is to all these citizen reporters who helped make the city more resilient to flood events that we dedicate this publication.

Coen, D. R. 2012. The Earthquake Observers: Disaster Science from Lisbon to Richter. University of Chicago Press.

de Wall, M. 2014. The City as Interface: How New Media are Changing the City. Rotterdam. nai10 publishers.

Douglas, M. 1986. How Institutions Think. Syracuse. Syracuse University Press.

Ebrahimy, R. 2014. Investigating SCADA Failures in Interdependent Critical Infrastructure Systems. School of Computing Science, Newcastle University, Newcastle, UK.

Easterling, K. 2014. Extrastatecraft: The Power of Infrastructure Space. London and New York. Verso.

Hallegatte, S. Green, C. Nicholls, R. J. Corfee-Morlot, J. 2013. Future Losses in Major Coastal Cities. Nature Climate Change, 3, pp 802-806.

Holderness, T. 2014. Geosocial Intelligence. Technology and Society Magazine, IEEE, 33, pp 17-18.

Hartono, D. M. Novita, E. Gusniani, I. Oriza, I. I. D. 2010. The Role of Water Supply and Sanitation During Floods: Case Study of Flood Disaster in Five Regions of Jakarta. International Journal of Technology. 1. pp 29-37.

IPCC. 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland.

Kompas, 2014. Jakarta Siap Antisipasi Banjir, tetapi 141 Pompa Air Rusak, 24 November, online access: http://megapolitan.kompas.com/read/2014/11/24/11190751/Jakarta.Siap.Antisipasi.Banjir.tetapi.141.Pompa.Air.Rusak

Li, H. 2003. Management of Coastal Mega-Cities – a New Challenge in the 21st Century. Marine Policy. 27. 333-337.

Lighthouse Foundation, 2006. For A Sustainable Saemangeum Final Report, online access: http://www.lighthouse-foundation.org/fileadmin/LHF/PDF/FASS-final_report_KOREA_.pdf

Medina, E. 2011. Cybernetic Revolutionaries: Technology and Politics in Allende's Chile. MIT Press.

Meier, P. 2014. Digital Humanitarians: How Big Data Is Changing the Face of Humanitarian Response. CRC Press.

Noviansyah, A. 2015. Islands in the Sand. Tempo English Vol. 41, No. 15.

Nancy, J.-L. (2015) After Fukushima: The Equivalence of Catastrophes, trans. Charlotte Mandell. New York: Fordham University Press.

NCICD, 2014. Online access: http://en.ncicd.com/direct-for-solution/great-garuda/

Paar, P. and Rekittke, J. 2011. Low-Cost Mapping and Publishing Methods for Landscape Architectural Analysis and Design in Slum-Upgrading Projects. Future Internet, 3, 228-247

Rahmawati, L. 2015. Vila-vila bekas bongkar di Puncak dibangun kembali. Antara News, 23 April, accessed online: http://www.antaranews.com/berita/492388/vila-vila-bekas-bongkar-di-puncak-dibangun-kembali

Reeves, J. 2015. Crowdsourcing Tools for Flood Reporting in Jakarta. Report by Humanitarian OpenStreetMap Team to the Indonesia-Australia Facility for Disaster Reduction.

Rosenzweig, C. 2011. All Climate is Local. Scientific American. 305, pp 70-73.

Rostanti, Q. 2012. 13 Sungai di Jakarta Berpotensi Banjir, Republika Online, 19 November 2012, online access: http://www.republika.co.id/berita/nasional/jabodetabek-nasional/12/11/19/mdq562-13-sungai-di-jakarta-berpotensi-banjir

Rujak Center for Urban Studies, 2008. “Diskusi Normalisasi Kali Ciliwung,” accessed online: http://rujak.org/2012/08/diskusi-normalisasi-sungai-ciliwung/

Smith, Z., 2015. Personal correspondence with PetaJakarta.org research team.

Semiocast, 2012. Twitter reaches half a billion accounts [Access 20 April 2015]. Online: http://semiocast.com/en/publications/2012_07_30_Twitter_reaches_half_a_billion_accounts_140m_in_the_US

Taufik, M. 2014. Ini 5 banjir besar yang pernah melumpuhkan Jakarta, 16 January, online access: http://www.merdeka.com/peristiwa/ini-5-banjir-besar-yang-pernah-melumpuhkan-jakarta

Tobler, W. R.1970. A Computer Movie Simulating Urban Growth in the Detroit Region. Economic geography, 234-240.

Townsend, A.M. 2013. Smart Cities: Big Data, Civic Hackers, and the Quest for a New Utopia. New York: W.W. Norton and Co.

Tran M.; Hall J.; Hickford A.; Nicholls R.; Alderson D.; Barr S.; Baruah P.; Beavan R.; Birkin M.; Blainey S.; Byers E.; Chaudry M.; Curtis T.; Ebrahimy R.; Eyre N.; Hiteva R.; Jenkins N.; Jones C.; Kilsby C.; Leathard A.; Manning L.; Otto A.; Oughton E.; Powrie W.; Preston J.; Qadrdan M.; Thoung C.; Tyler P.; Watson J.; Watson, G. & Zuo, C. National infrastructure assessment: Analysis of options for infrastructure provision in Great Britain, Interim results. 2014. UK Infrastructure Transitions Research Consortium, Environmental Change Institute, University of Oxford.

Turpin, E., Bobbette, A., Miller, M., ed. 2013. Jakarta: Architecture + Adaptation. Jakarta. Universitas Indonesia Press.

Walsh, C. L.; Dawson, R. J.; Hall, J. W.; Barr, S. L.; Batty, M.; Bristow, A. L.; Carney, S.; Dagoumas, A. S.; Ford, A. C.; Harpham, C.; Tight, M. R.; Watters, H. & Zanni, A. M. 2011. Assessment of climate change mitigation and adaptation in cities. Proceedings of the ICE - Urban Design and Planning, 164, 75-84(9)

Yanto, O. 2010. Mafia hukum: membongkar konspirasi dan manipulasi hukum di Indonesia. Jakarta: Raih Asa Sukses.

Appendix I. CogniCity Software References

The table below provides links to the source code and documentation for each of the components of CogniCity, as deployed for the 2014/2015 monsoon season in Jakarta. For further information see: http://cognicity.info/cognicity/

Appendix II. Listing of Tables in CogniCity Database

Table Name Description
tweet_reports Confirmed tweet reports of flooding
tweet_reports_unconfirmed Unconfirmed tweet reports of flooding
nonspatial_tweet_reports Confirmed tweet reports of flooding missing geolocation metadata
all_users Encrypted hash of all related Twitter usernames
tweet_users Encrypted hash of user names who have submitted confirmed reports
tweet_invitees Encrypted hash of users have been been sent an invitation
nonspatial_tweet_users Encrypted hash of users who have submitted confirmed reports missing geolocation metadata
jkt_city_boundary Boundaries of Jakarta’s five municipalities
jkt_subdistrict_boundary Boundaries of Jakarta’s municipal sub-districts (‘Kecamatan’)
jkt_village_boundary Boundaries of Jakarta’s municipal villages (‘Kelurahan’)
jkt_rw_boundary Municipal boundaries of Jakarta’s municipal RW districts (‘Rukun-Warga’)
pumps Locations of water pumps in Jakarta
floodgates Locations of floodgates in Jakarta
waterways Locations of waterways in Jakarta
[1] Tables in the CogniCity Database

Appendix III. Listing of Data API Endpoints in CogniCity Server

CogniCity API Endpoint Description Data Temporal Extent Spatial Extent
reports/confirmed Real-time listing of confirmed flood reports Point geometries + message 1 hour Jakarta + surrounds
reports/unconfirmed Real-time listing of unconfirmed flood reports Point geometries 1 hour Jakarta + surrounds
reports/count Real-time sum count of all reports Count 1, 3, 6, or 24 hours n/a
reports/timeseries Real-time sum count of all reports at hourly intervals Timestamps + count 24 hours n/a
aggregates/live Real-time count of all reports by municipal area Polygon geometries + count 1, 3, 6, or 24 hours Jakarta
aggregates/archive Archive of previous counts of all reports by municipal area Polygon geometries + timestamp + count Archive extents in 1 hour blocks Jakarta
infrastructure/waterways Waterways in Jakarta Linestring geometry + name n/a Jakarta + surrounds
infrastructure/pumps Water pumps in Jakarta Point geometry + name n/a Jakarta + surrounds
infrastructure/floodgates Floodgates in Jakarta Point geometry + name n/a Jakarta + surrounds
[1] CogniCity Server Data API Endpoints

Appendix IV. PetaJakarta.org Major Open Data Collection References

The table below provides information on the PetaJakarta.org Major Open Data Collection (MODC). The PetaJakarta.org MODC project was supported by the Australian National Data Service (ANDS). ANDS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy Program. For further information see: https://petajakarta.org/banjir/en/data/

Site Reference
Australian National Data Service https://projects.ands.org.au/id/MODC15
Research Data Australia https://researchdata.ands.org.au/petajakartaorg/552178
[1] PetaJakarta.org MODC Project Information

Photo 09. Official launch of PetaJakarta.org at City Hall.

Photo 09. Official launch of PetaJakarta.org at City Hall; L to R: Twitter Director of Academic Partnerships Mark Gillis, Jakarta Governor Basuki Tjahaja Purnama, PetaJakarta.org Co-Principal Investigators Dr Tomas Holderness and Dr Etienne Turpin, December 2014. Photo courtesy of Tatyana Kusumo.

Dr Tomas Holderness is a Geomatics Research Fellow at the SMART Infrastructure Facility, University of Wollongong. His research focuses on developing new architectures for urban data collection, integration and analysis applied to urban infrastructure resilience and Earth systems engineering. In particular, he has developed new modes of inquiry into the response of megacities to climate change using informal and crowd-sourced data. Tomas is a Chartered Geographer,and Fellow of the Royal Geographic Society, and leader of the SMART Infrastructure Facility Open Source Geospatial Lab.

In collaboration with colleagues from Newcastle University (UK), Dr. Holderness developed a pioneering spatio-topological database schema for encoding, modeling and analysis of spatial infrastructure networks, and their inter-dependencies. He has successfully applied this framework to model sanitation networks in rapidly-urbanizing developing regions of Africa using crowd-sourced data and volunteer geographic information. This research provided a novel insight into the infrastructure challenges faced by developing nations, and allows the calculation of long term operating costs of different improved sanitation network options. This work, published in the Proceedings of the Institute of Civil Engineers “Municipal Engineer”, was awarded the James Hill Prize for best paper in journal in 2014.

Dr Holderness is also the co-creator of the PetaJakarta.org project: a world-first collaboration between a research institute, a government agency, and indistry leader. Through his work on ‘geosocial intelligence’ as a method to crowd-sourcing situational awareness using social media, Dr Holderness led the development of CogniCity, an open source geospatial intelligence framework which is the foundation of PetaJakarta.org. His research on PetaJakarta.org—to examine the application of ‘geosocial intelligence’ to create of structures of civic co-management as a method of adaptation to climate change—has been reported on by The Wall Street Journal, BBC Indonesia and National Geographic Indonesia.

Before joining the SMART Infrastructure Facility, as a spatial modeler for the Geospatial Engineering Research Group at Newcastle, Dr Holderness was responsible for the development of an open-source integrated modeling environment for urban systems research. Dr Holderness developed a prototype framework that allowed the integration of models to create processing flow-lines and statistical ensembles for land-use change modelling in response to different climate scenarios. Prior to this research Dr Holderness’ PhD thesis analysed long time series thermal Earth observation data to quantify intra-urban spatio- temporal temperature dynamics in Greater London.

email tomas@uow.edu.au
Twitter @iHolderness

Dr Etienne Turpin is a philosopher studying, designing, curating, and writing about complex urban systems, political economies of data and infrastructure, aesthetics and visual culture, and Southeast Asia colonial-scientific history. At the University of Wollongong, Australia, Etienne is a Vice-Chancellor’s Postdoctoral Research Fellow with the SMART Infrastructure Facility, Faculty of Engineering and Information Science, and an Associate Research Fellow with the Australian Center for Cultural Environmental Research, Department of Geography and Sustainable Communities. He is a member and former team leader of the SMART GeoSocial Intelligence Research Group and his research is supported by the SMART OSGeo Lab.

In Jakarta, Indonesia, Etienne is the founder and director of anexact office, a design research practice operating on the shifting conceptual and physical terrain of the Anthropocene through the interventive study of urbanization processes, knowledge infrastructures, and data polities; and the co-principal investigator, with Dr Tomas Holderness, of PetaJakarta.org, an applied research project utilizing a GeoSocial Intelligence Framework to study urban infrastructure. Through strategic community organizing, institutional ethnography, and novel approaches to social media platforms, data gathering, and designed engagement, Etienne’s research helps to produce new tools, techniques, and methods to help democratize processes of urban transformation by meaningfully engaging the concerns and capacities of the urban poor.

His urban research with Dr Holderness has been reported on by The Wall Street Journal, National Geographic, BBC International, and CNN, among others, and featured in numerous climate adaptation studies including reports by the International Federation of the Red Cross, Internews, and the Australian Department of Foreign Affairs and Trade; his research has also been supported by a number of prestigious grants, including those awarded by the Australian National Data Service, the New Colombo Plan of the Australian Department of Foreign Affairs and Trade, the Australia-Indonesia Facility for Disaster Reduction, and USAID, as well as Twitter's OSS #DataGrant.

Etienne is also a member of the SYNAPSE International Curators’ Network of the Haus der Kulturen der Welt in Berlin, Germany where he is the co-editor, with Anna-Sophie Springer, of the intercalations: paginated exhibition series as a part of Das Anthropozän-Projekt. He is also the editor of Architecture in the Anthropocene: Encounters Among Design, Deep Time, Science and Philosophy (Open Humanities Press, 2013) and co-editor of Art in the Anthropocene: Encounters Among Aesthetics, Politics, Environments & Epistemologies (Open Humanities Press, 2015), Fantasies of the Library (K. Verlag, 2015), Land & Animal & Nonanimal (K. Verlag, 2015), and Jakarta: Architecture + Adaptation (Universitas Indonesia Press, 2013).

Prior to his work in Australia and Indonesia, Etienne was a Research Fellow at the Center for Southeast Asian Studies, University of Michigan, where he also taught advanced design research, architecture theory, and coordinated international research-based design studios for the Taubman College of Architecture and Urban Planning. He has also taught advanced design research and design theory at the College of Environmental Design, University of California Berkeley, and the Daniels Faculty of Architecture, Landscape, and Design, University of Toronto.

email eturpin@uow.edu.au
Twitter @turpin_etienne