TIGER - Water Resources Management in Africa

Reliable access to water, managing the spatial and temporal variability of water availability, ensuring the quality of freshwater and responding to climatological changes in the hydrological cycle are prerequisites for the development of countries in Africa. Water being an essential input for biomass growth and for renewable energy production plays an integral part in ensuring food and energy security. Water, as a source of safe drinking water, is furthermore the basis for ensuring the health of citizens and plays an important role in urban sanitation. The recently announced Sustainable Development Goals (SDG) of the United Nations include for the first time a dedicated goal (6th SDG) on "Ensure water availability and sustainable management of water for all". Earth Observation can support the assessment and monitoring of several targets and indicators asscociated to the 6th SDG. 

The concept of Integrated Water Resource Management (IWRM) is seen as an opportunity to help manage water variability and the wide spread water scarcity in Africa. One key component missing from IWRM in Africa is the limited knowledge of the available extent and quality of water resources at basin level. ESA’s TIGER initiative aims at enabling African water authorities to fill this information gap by monitoring water resources at adequate temporal and spatial scales based on Earth Observation (EO) technology.

In direct collaboration with African Water authorities the Water Observation and Information System (WOIS) has been developed in TIGER as an open source software tool for monitoring, assessing and inventorying water resources using EO data. The WOIS offers more than 28 EO products for IWRM tasks from watershed to transboundary basin levels. Resulting EO information products cover basin-wide characterization of land and water resources (e.g. small water bodies), lake water quality monitoring, hydrological modeling, flood forecasting and mapping. This contribution will present WOIS use-cases validated and demonstrated with major river basin authorities (Nile Basin Initiative, Lake Chad Basin Commission, Zambezi Watercourse Commission, Volta Basin Authority) and national water authorities (Department of Water Affairs South Africa; Namibian Ministry of Agriculture, Water and Forestry; Zambian Ministry of Mines, Energy and Water Development).

Water resources management has become a challenging problem worldwide, especially in developing countries. The lack of information on land cover monitoring has a huge impact on the water resources management, since it may hamper the collection, treatment and distribution of water for human consumption and agricultural development. To fulfil this demand, Earth Observation data has been used in African countries to map land cover of large and often inaccessible areas, providing useful information on qualitative and quantitative land cover changes. Although optical data methods for land cover classification are well established and almost operational, these data are not applicable to regions where the cloud coverage is frequent. In these regions, the use of Synthetic aperture radar (SAR) data is an alternative due to its ability to acquire data regardless of weather conditions and day/night cycle. The aim of this study is to assess the complementarity and interoperability of optical and polarimetric SAR data to map and monitor crops of an agricultural area in Wako Kungo, Cela municipality, South Kwanza province, Angola. For this purpose, 28 SPOT 5-Take 5 images (April, 10 to September, 12), 9 Sentinel-1 dual polarisation (VV+VH) images (March, 26 to October, 10) and field data (April, 15-30) acquired during the 2015 growing season are used. SPOT5 Take 5 experiment images are used, as a proxy of Sentinel-2 data, to evaluate the potential of its enhanced temporal resolution for agriculture applications. SPOT Normalized Difference Vegetation Index (NDVI) and VV and VH polarization backscattering time series were plotted, for the crops parcels identified in the test area, to evaluate the discrimination among the different crops. The field data collection and classification focused on the main crops grown in the region, which include: maize, soybean, bean and pastures. Average NDVI values are also used to compute the basal crop coefficients (K_cb) for each crop growth stage and to estimate the respective length of each phenological growth stage. Both are then used to compute the crop evapotranspiration and subsequently to estimate the crop irrigation requirements based on a soil water balance model. The integration of optical and SAR data is assessed by comparing the classification results from different algorithms under 2 different scenarios: SPOT time series and mixed SPOT-Sentinel 1 time series. The SAR inputs include VV and VH backscatter intensity channels, VV and VH ratios and VV and VH differences. Preliminary results show that the combination of images from different sources provides the best information to map agricultural areas and also that the use of multi-temporal data can successfully classify crops due to spectral information for the complete growing season. The study was developed in the scope of the ESA Alcantara initiative project (Ref: 14-P13) and Spot-take 5 project ID: 29142. Publication supported by FCT through project UID/GEO/50019/2013 - Instituto Dom Luiz, University of Lisbon.

More than 80% of farms are operated by smallholder farmers in rural Sub-Saharan Africa with 70% of them being women (AGRA, 2014). Again a large percentage (more than 70%) relies on rainfed farming which makes their work risky with a changing and uncertain climate. The few who use small reservoirs for irrigation as a means of increasing their resilience to climate induced stresses are not always sure about the quantity of water they have and how many acreages could be cultivated with it at any given time. Since 2002, the small reservoirs project (www.smallreservoirs.org) has been developing tools and methods to estimate the volume of water stored in these reservoirs using a combination of bathymetric analyses and remotely sensed surface areas of the reservoirs. In 2014, the Small Reservoirs Team developed a Bayesian Classifier (Eilander et al., 2014)  for improved classification of water bodies for flood mapping, runoff estimation and reservoir water balance assessment. These activities took place in the framework of the European Space Agency’s (ESA) TIGER and Alcantara projects, and the Canadian Space Agency’s SOAR Africa project using Radarsat-2 Images. There is the continuous need for a dedicated effort of using a combination of in-situ data and new Earth Observation (EO) systems, such as the Sentinels, to monitor small inland water bodies used for irrigation. Through the Spot5-Take5 initiative by ESA a time series of 27 Level-2A products was developed over a pilot area in Ghana (Upper East Region). These Spot5 images were acquired from 11-April 2015 to 13-September 2015. These images had a 10m and 5-day spatial and temporal resolutions for monitoring small reservoirs dynamics. We show in this research that Sentinel-1 images (acquired from 01-April 2015 to 06-August 2015) and Sentinel-2 here simulated with the Spot5 images are great additions to the satellite constellations available for near-real-time water monitoring for operational water management (flood & drought) in ungauged basins. Establishing these sound scientific time series and trends is required for reducing the risks associated with farming by developing effective adaptation and mitigation measures in agriculture to alleviate rural Sub-Saharan African farmers (especially women) from poverty.

Caprivi basin in Namibia has been affected by severe flooding in the past years resulting in deaths, displacements and destruction of infrastructure. The negative consequences of these floods have emphasized the need for timely, accurate and objective information about the extent of affected areas. The aim of this work is to explore the performance in mapping flooding areas by thresholding methods using high and low resolution for both active microwave synthetic aperture radar (SAR) data and multispectral data. Recent developments in remote sensing technology have resulted in substantial improvements in the spatial and temporal resolution of SAR (Sentinel 1) and optical data (SPOT 6/7).  However, actually, optical and SAR data at lower resolution than SPOT and Sentinel-1 have a good temporal and spatial coverage in the study area and were used to monitor flooding in the period from 2008 to 2011.

Thresholds on radar backscatter were defined in order to map flooding areas in the Caprivi basin using ENVISAT ASAR data acquired during 2008-2011. The mean and standard deviation thresholds were used on a series of Envisat ASAR data. A thresholding on NDWI index was applied to map flooding events using available multispectral Landsat data and compared with maps obtained by ENVISAT ASAR. Change detection method has been used to monitor flooding events in the entire period. The improvement in the delineation of flooding extent using different spatial resolution data has been also explored by comparing flood maps obtained by ENVISAT ASAR at 1 km and at 30 m resolution SAR data. The same comparison was done using thresholding method on ENVISAT multispectral data at 1 km resolution and MERIS data at 300 km resolution.

Depending on flooding month, temporal variations of the radar backscatter have shown better performance in flood monitoring than thresholding methods using multispectral data. The maximum extent for flooding in Caprivi basin has been found in 2011 where flooded vegetated areas were small with respect to total mapped flooded area.  Future improvements in monitoring and delineation of floods using sentinel data are discussed.

Water management is an essential issue in arid and semi-arid lands, especially where there are extensive irrigated areas such as in the Doukkala area (Western Morocco). Optimization of irrigation water use, by giving only what is actually needed by plants, is one of the effective ways to save water. In order to monitor crop water requirements (CWR) over large areas, remote sensing is a very interesting tool that gives synoptic information about crop status.
A time series of high spatial resolution satellite images (SPOT 5- take5), acquired at high revisit frequency (5 days) from April to September 2015, were processed by Harmonic ANalysis of Time Series (HANTS) method to remove and reconstruct cloudy observations.  Normalized Difference Vegetation Index (NDVI) temporal signals were generated for each pixel and used as input for unsupervised multi-temporal classification (ISODATA), to identify the dominant crop types (Wheat, Sugar beet, Alfalfa and Maize). The generated crop classes were used for CWR-estimates using simple and (semi) empirical methods: the first one is called Kc-NDVI method, based on the correlation between NDVI and the crop coefficient (Kc); the second one is the analytical approach based on the direct application of the Penman-Monteith equation.
Periodical field campaigns were carried out in the study area in the same date of SPOT5 acquisition to assess the accuracy of the generated landuse and to validate satellites estimates of CWR which shows a good agreement with ground-based data. The assessment of adequacy of water allocations to water requirement showed that CWR were much larger than allocated surface water for the entire area, with this difference being small at the beginning of the growing season. Smaller differences were observed between surface water allocations and Irrigation Water Requirements (IWR), taking into account precipitation, throughout the irrigation season.
 In order to highlight the necessity of crop map for each method used and taking into account the complexity of crop mapping, we have compared the generic and crop specific CWR-estimates. Accordingly, the Kc-NDVI is crop dependant method unlike the analytical approach that gave similar generic and crop specific CWR-estimates.