Bangladesh is a disaster-prone country, with floods being the most frequent and severe danger.

 

Cyclones also constitute a major risk. Bangladesh is also susceptible to other dangers, including drought, earthquakes, river bank erosion, tsunamis, arsenic contamination of groundwater, soil salinity, and health pandemics. Because Bangladesh is a densely populated, low-income country, natural calamities cause disproportionate and adverse impacts to its population and economy. The government has made substantial efforts in disaster risk management and instituted national disaster management structures and legislation both nationally and sub-nationally. However, the application of financial protection in disaster risk management is still limited. 

​

ADB has initiated a project concerning capacity building for Disaster Risk Finance (DRF) to enhance disaster risk preparedness of Bangladesh by developing public and private capacity for DRF solutions and fostering linkages between public and private sector in disaster risk mitigation. The project will develop a disaster risk model that requires geo-information as input and plans to use additional disaster risk information from other sources, such as in-situ data on physical data. 

​

The EO support project linked to the ADB activity by providing datasets that support assessment of nature of disaster risk profiling: probability of occurrence, and the potential loss of, or impact on, life and physical capital. The delivered information included land use and land cover, a DEM, flood history as well as disaster risk data and maps for floods and cyclones (e.g. flood damage calculations based on land use, flood depths and duration). The areas of interest were mainly Dhaka and Chittagong. 

​

Urban land cover mapping and flood occurrence monitoring represent two of the most common and useful applications of EO. Satellite-based land cover information when coupled with actual historical inundation and flooding detected by EO and hydrological models can help to analyse the existing situation as well as projected climate change, flood impacts, and developments to provide essential information for better decision making and planning of disaster response and finance projects. For example, such analysis can serve to highlight the areas at highest risk of floods and (riverbank) erosion, as well as key areas where critical urban assets (e.g. densely populated, commercial or transport) and infrastructure such as roads and bridges is at highest risk of being affected.

​

Moreover, the EO-based information can help calculate scenarios with associated damage in financial terms. Finally, actual flooding can be mapped and monitored systematically to assess the impact and affected areas as floods happen. 

​

Quantitative map accuracy assessment was applied using internationally accepted standard methods. Reference data was generated by means of expert visual interpretation of very high resolution satellite imagery (0.6— 2.5 m spatial resolution) available from e.g. Google Earth. A minimum of 20 sample points were interpreted for each individual land cover class. 

All products are visualised in an online geodata platform available at https:// bangladesh.lizard.net. 

​

Land use / land cover maps and Digital Elevation Model (DEM) 

​

Land use / land cover maps were created for Dhaka and Chittagong at 15 m spatial resolution, based on a combination of imagery from the Landsat-8, Sentinel-1 and Sentinel-2 satellites, with SPOT 5 data being used as reference. The service meets the standard product specifications defined for the European Commission’s Urban Atlas. 

​

As other sources of very high resolution elevation data were not available over large areas during the project, the SRTM DEM dataset available for the entire country of Bangladesh at a resolution of 30 m was used for the assignment.

Historical flood mapping 

​

Historical flood inundation maps of Bangladesh were derived from MODIS and Sentinel-1 satellite data period 2003–2015. The dataset is presented as weekly maps with a spatial resolution of 250 m. In each map, the pixel count represents the number of wet observations measured during that week (observations twice daily for MODIS for the entire period and ). Flood occurrence has been mapped using long time series of MODIS (all twice daily observations between the entire period 2003-2015) and Sentinel-1 (observations at approximately 20 m made near-weekly not affected by clouds, between October 2015 and March 2016). 

​

As an example of systematic flood monitoring, after 300 m of river embankment collapsed in the region of Bogra, radar satellite imagery could help to verify that over 100 villages were flooded and over 40,000 hectares of crops were affected. Reportedly, during this and other events in the region that left 3 people dead and 1 missing, more than 129,000 farmers had been affected and 1.7 million Euro in crop damage occurred.

Disaster risk mapping 

​

This service provided maps related to flood, cyclone and earthquake risk for the entire national territory of Bangladesh. 

​

Maps of both EO-based actual inundation and 3D model-based flood risk were produced for areas around Dhaka and Chittagong. The result is a temporal stack of maps with water depths. With the calculated water depths and durations the potential damage (€/m2) was also calculated for the two cities. Graphs are available at any location and can be used to determine e.g. at what time the flood waters will reach the location, but also how fast the water depth increases or how the damage increases with depth. 

​

Tropical cyclone tracks were acquired from the US National Centers for Environmental Information (NOAA NCEI, http://www.ncdc.noaa.gov). Earthquake data was used from USGS (http://earthquake.usgs.gov). One of the goals of the exercise was to translate the available information into easy-to-use statistics, with the ability to explore and aggregate data spatially and temporally in the simple online analysis tool. For instance, one can easily filter and extract the earthquakes that occurred in the year 2011.

Limitations and constraints 

​

The urban land cover maps were produced at lower resolution due to nonavailability of recent very high resolution data (2.5 m or better). Such information can be programmed in operational service provision scenarios. 

​

The production of a spatially explicit historical flood occurrence database is limited by the availability and quality of existing satellite data. Due to gaps in time and space (no satellite overpass or observations in cloudy conditions) the historical flood event data should be used with caution. Especially the detection of (historical) flash floods with short duration remains problematic, such cases may likely not be recorded by the satellite data. Combining constellations of radar and optical satellites are able increasingly able to deal with this problem. 

​

Moreover, further work is needed to separate flood occurrences into ‘good’ inundation (e.g. normal inundation of agricultural fields) from ‘bad’ inundation (e.g. disastrous flooding). For this reason, in an additional processing step, the nation-wide historical inundated areas map 2003–2015 should be interpreted by specialists. 

The flood damage that was calculated for Dhaka and Chittagong could be created in much more detail. The more detailed the Digital Elevation Model and land cover map, the more precisely the damage can be calculated. At the time of project implementation very high resolution satellite (e.g. WorldDEM) or LiDAR datasets were not available. 

​

Effective characterisation of flood risk in Bangladesh requires the combination of a range of different datasets, including coastal conditions, river hydrology and urban development in a consistent manner. In addition, costeffective monitoring of changes that impact on flood risk (e.g. movement of channels in the delta system, changes in mangrove cover, evolution in the urban land cover) can be achieved for the entire area only through the use of EO-derived information. The possibility of combining EO-derived flood risk information generated in the present project together with the cyclonerelated risk characterisation performed by the ADB contractor and generated in cooperation with this project also represents an interesting possibility for providing a methodology for integrated risk assessment. Thanks to the longterm availability of data from the Sentinel-1 and Sentinel-2 satellite missions, the historic datasets generated here can be updated on a regular basis to ensure a robust, continuous assessment of the evolution in exposure to floodrelated risk. In addition, this capability can be extended geographically to neighbouring India and Myanmar. Given the development investments planned for the region in relation to risk reduction, there is strong potential for wider use of this type of information in the near future.