On Friday 25 August 2017, the insurgent group the Arakan Rohingya Salvation Army carried out a series of attacks against an army base and several police posts in Rakhine State in Myanmar. The military’s response rapidly escalated into a crackdown. By the end of October, over 700,000 Rohingya had fled Rakhine; a year later, the number of Rohingya refugees in Bangladesh, including previous influxes, is estimated at more than 900,000. The vast majority arrived at and have stayed in the Cox’s Bazar district, along the border between Myanmar and Bangladesh. The events following the August crackdown will be remembered for many reasons – not least the unprecedented levels of violence. For humanitarian actors, the crisis marked a chain of events that put both response and coordination to the test. Among these actors was REACH, which worked to provide data on the rapidly growing refugee camps.
Data cannot wait — neither can coordination
The early stage of the crisis was characterised by a daily influx of refugees of up to 12,500 arrivals per day.+The first Humanitarian Response Plan (October 2017) reports 509,000 arrivals from 25 August to 2 October 2017, which represents a daily average of about 12,500 new arrivals a day. The rate of arrivals resulted in the rapid expansion of new camps across Cox’s Bazar, and the area’s previously green hills were quickly cleared of vegetation to make way for temporary shelters and homes.
As the need for aid mounted, humanitarian actors scaled up their life-saving interventions to provide basic services. In-kind distributions, building new emergency infrastructure and strengthening inter agency coordination platforms, were all response priorities. Another was the need for data. First, humanitarian actors had to define and regularly update information on the scale and extent of the crisis. Second, actors had to record and share information on new emergency infrastructure being built by response actors working both within and outside of standard coordination structures. Both tiers of data were key for planning, targeting and implementing an efficient response.
Filling the gaps
REACH arrived in Cox’s Bazar in September, less than a month after the escalation of the crisis. In light of the continuous expansion of the sites and the challenging terrain, REACH collaborated with the UN Operational Satellite Applications Programme (UNOSAT) to acquire satellite imagery of the border areas of Myanmar and Bangladesh. Using this imagery, the team digitised tens of thousands of shelter footprints.
The use of satellite imagery allowed for a better understanding of the extent of the camps across the region and the accurate definition of camp boundaries, and contributed to assessments of population estimates, spatial patterns and settlement trends. Later on in the crisis, satellite imagery proved critical in identifying infrastructure built in locations at risk of flooding and landslides ahead of the monsoon season.
Satellite imagery provided a broad canvas of information, but it did not answer questions related to the conditions, characteristics, availability and access to infrastructure. In order to fill the gaps satellites could not, REACH teams set out on foot.
Everything begins with a pilot
Before the launch of systematic data collection in the megacamps of Cox’s Bazar, REACH conducted a pilot infrastructure survey developed in collaboration with the water, sanitation and health (WASH) sector and carried out with the help of an opensource mobile data collection platform, Kobo Toolbox. The pilot began by mapping WASH infrastructure (mainly handpumps and latrines) in Hakimpara, Jamtoli and Baggoha-Potibonia camps, which together had a total population of 100,000. The relatively small size of these camps enabled a quick assessment of the feasibility of the data collection methodology and the appropriateness of the tools to be used in camps and extension sites more than six times the size of the pilot settlements.
The pilot was straightforward. On reaching WASH structures, the type and GPS coordinates of the structure were noted, and a photograph was taken in order to enable cross-checking during data cleaning. The camps were mapped in a matter of days. The pilot quickly revealed that, despite having detailed static reference maps, the complexity of the settlements and the hilly terrain called for more dynamic means of data collection and navigation. A GPS-enabled, offline base-map and a numbered grid system were developed and loaded on mobile devices. Each team of enumerators was assigned a series of grid cells to maintain their bearings and collect infrastructure data.
With time, the deployment developed into a process of data collection and infrastructure and functionality monitoring in all the camps in Cox’s Bazar, conducted on a daily basis. At the end of each monthly data collection period, REACH was able to provide the humanitarian community with a comprehensive dataset of camp infrastructure, along with a number of related static maps and factsheets.
Data collection was not done in parallel to existing operational coordination structures, but designed and implemented in close collaboration with existing coordination platforms. The strengthened partnership between coordination and data collection enabled the utilisation of established networks and available information, and increased ownership of the data by the humanitarian community. This was a necessity, as only through ownership would the data be used to influence decision-making and lead to evidence-based response.
Data alone has little value
At the most basic level, REACH’s infrastructure mapping provided a ‘good enough’ set of maps and data during the early phase of the response. By providing information on the basic layout of the rapidly established camps, and the functionality and coverage of services, the data provided by REACH served as a basis for humanitarian planning.
Providing information to enable comparisons of the usability, access and infrastructure of services in the camps was a valuable tool for response actors in identifying gaps in WASH provision. Regular updates in turn allowed gaps to be monitored and addressed over time. Spatial data on infrastructure and functionality allowed partners to plan and respond at the very local level, be it in terms of coverage, maintenance or ensuring sufficient distance between, for example, latrines and drinking water sources. Since the start of the crisis, the infrastructure data provided by REACH has been used consistently by both the site management+Site Management Sector Cox’s Bazar, ‘Site Management Site-level Services Monitoring Approach’ (Cox’s Bazar, April 2018). and the WASH sectors.+WASH Sector Cox’s Bazar, ‘WASH Sector Strategy for Rohingyas Influx March to December 2018’ (Cox’s Bazar, March 2018), p. 15.
Figure 1. The rapid expansion of refugee camps has led to a drastic decrease of vegetation in the historic forests of Bangladesh’s Cox’s Bazar district
Aside from providing information on basic services, the infrastructure data set was critical in supporting response-wide emergency preparedness planning. With Cox’s Bazar experiencing some of the most intense monsoon rainfall in the country, it was only a question of time before landslides and flooding threatened camps spread over steep hills and narrow valleys. To enable targeted prevention, REACH identified structures that were likely to be at risk or unusable once the monsoon season started. REACH also identified structures that could be suitable for community shelters in the case of relocations. This resulted in the creation of a list of sites in need of further scrutiny, ground-truthing or reinforcement, saving time, resources and – potentially – lives.
The devil’s in the detail — unique identification codes
As the months passed and the response stabilised, the Inter Sector Coordination Group, sector working groups and implementing agencies substantially strengthened. New actors such as Open Street Map were beginning to catch up and provide full coverage of shelter footprints, making it unnecessary for REACH to continue its early response analysis. In this changed environment, a series of issues began to gain prominence, including the lack of universal unique identification codes for infrastructure, limitations in GPS accuracy and the proliferation of multiple overlapping datasets.
Under REACH’s census approach to infrastructure mapping, each round of data collection produced a new infrastructure dataset from scratch. Enumerators would return to the field at the start of each round, record what they found and a new infrastructure list would be generated based on the findings. With infrastructure still in flux, REACH did not attempt to send enumerators back to the same points each month to check whether the structures were still there. The fear was that mapping infrastructure in such a rapidly changing environment without unique identifiers would mean repeatedly mapping the same structure, with slightly different coding. Each round of data collection thus had to be treated as a new exercise, the risk of discrepancies and gaps notwithstanding.
Although enumerators attempted to record identification codes on infrastructure, it was impossible to use this data due to a proliferation of various approaches to identification by implementing agencies. Codes were often entirely absent or written in a bewildering array of different formats. In the course of the crisis it became evident that building infrastructure was not enough. Creating and coordinating an infrastructure coding mechanism for data collection was necessary too.
During the early stages of the response, wiping the slate clean with a new set of data was not an issue given the number of changes taking place, with new infrastructure being added as first-phase and temporary infrastructure was taken down. However, as the situation stabilised the lack of consistency created significant confusion.
Due to the labyrinthine nature of the camps, REACH enumerators were never able to ensure that they were assessing every infrastructure point with each round. This meant that, even if the number of points for a given category of infrastructure had not changed month on month, the number of points recorded in REACH’s databases could fluctuate – sometimes by up to 50% in a single camp. This was especially problematic for hard-to-categorise infrastructure, such as mental health spaces, community spaces and women’s centres or centres with multiple purposes depending on the time of day, week or month.
As sector-level information management focal points began receiving GPS coordinates from partners, sectors were able to maintain and update their own infrastructure database. However, the divergences between REACH data and their own information further aggravated the confusion over which figures and locations to use. The lack of a common identifier system coupled with limitations in the accuracy of smartphone GPS meant that cross-verifying different datasets proved challenging and, in the case of small, numerous infrastructure such as latrines, impossible. It was soon clear that sector data coverage and quality were simply not consistent. Compounding these factors, high levels of turnover and coordination bottlenecks throughout the response meant that understanding of the purposes and limitations of REACH’s infrastructure mapping was becoming increasingly unclear. Fresh approaches to infrastructure monitoring were required.
As the Rohingya response reaches the one-year mark, and as it transitions towards more mid- and long-term approaches, it is time to assess the lessons emerging from a crisis that in many ways has been unprecedented. From the viewpoint of data and data collection, two lessons stand out.
In the course of the response, it became abundantly clear that data or information is not valuable in and of itself. Information that is not gathered in a well-coordinated way loses its usability. Coordination of data collection needs to be built in as part of the wider coordination system. Any such exercise must be firmly owned by the coordination structure in charge of the response, alongside the close involvement of sector working groups.
Standardisation of a phased approach
Recognising that there is no established blueprint for a humanitarian response matching the magnitude of the Rohingya refugee crisis, there is an obvious need to adapt data collection and methodologies as the response changes over time. The standardisation of a phased approach could offer clear benefits here. Intrinsically linked to the architecture of data coordination, such an approach could comprise an initial phase and a stabilisation phase, each with a specifically tailored methodology and clearly defined roles for sector actors and partners. A system for coding unique identifiers or IDs, for example, can be built into a project from the start and triggered once the stabilisation phase is imminent, if not earlier. Having a phase-out strategy from day one, supported by a sequenced handover over a set timeframe, could allow key tasks to be transferred to local actors. This would further provide space for new approaches, activities and programmes designed, developed and implemented by international actors.
Conclusions and next steps
Overall, the REACH response in Cox’s Bazar succeeded in filling critical information gaps. It demonstrated that large-scale infrastructure monitoring in exceedingly complex settlements is not only possible, but also sustainable. However, much of REACH’s work to date has been tailored to the initial phase of the response. REACH Bangladesh is working to overcome some of the early data coordination challenges and adapt to a stabilisation phase.
With the development and roll-out of a long-term strategy in mind, REACH has worked with the Inter Sector Coordination Group and sector lead agencies to hand over responsibility for collecting infrastructure data to each corresponding sector, offering capacity-building support where needed. Recognising that data collection is not just the responsibility of data collectors, all actors engaged in the response are critical in ensuring that the foundation for reliable data is there.
In order to address the most significant obstacle affecting REACH’s ability to meet the data needs of its partners, REACH is now exploring the possibility of designing a system for creating labels with unique identification codes that can be applied to all infrastructure in the camps. Building on this, it is worth noting that learning early lessons during the Rohingya crisis should lead to more effective and robust processes for monitoring infrastructure. If applied efficiently, these lessons can easily become the new standard in future emergency contexts globally.
Matthew Wencel Senior GIS Manager, Vincent Annoni REACH Global Coordinator, Augusto Comé Assessment Specialist of the Global Wash Cluster, Jeremy Wetterwald Country Coordinator, and Oliver Lough Assessment Manager.
REACH activities during the first three months were funded by the Global WASH Cluster (GWC) and implemented through a series of surge deployments by the REACH global team. GWC seed funds were complemented by the UNICEF country office, which significantly expanded the assessment portfolio to meet growing information needs. Besides camp mapping and infrastructure monitoring, REACH’s activities currently include camp-level household surveys, advanced analysis in support of preparedness efforts for the monsoon season and assessment capacity-building for partners.