|dc.description.abstract||Water is essential for human life. Due to the increasing imbalance in freshwater availability and water demand, water scarcity is becoming a big problem. Irrigation accounts for 69% of the global water withdrawal and, therefore, global hydrological models are widely used to map water demand for irrigation. These models assume that surface water is taken from its nearest source and if there is no surface water available, it is taken from groundwater. Due to the absence of irrigation networks in these models, there is no explicit spatial link between irrigated areas and their surface water source. The aim of this study was to extract irrigation networks from OpenStreetMap (OSM) data and a global digital elevation model to create the link between rivers and irrigated areas. The method included three main steps: (1) processing the OSM water data (waterways and open water shapefiles), whereupon the OSM data was linked to a river, (2) subtracting urban areas and (3) excluding higher located areas with an elevation threshold. Thereafter individual irrigation networks were created. These networks were created on a 30 arcseconds resolution. Four different study areas were described, which are the Netherlands, the Nile delta, the Indus basin and the Citarum basin. Furthermore, the definition of the river source has been analysed. The OSM based irrigation networks were created on three minimum discharges, 100, 50 or 5 m3/s, and the effects on the networks were analysed. Moreover, the irrigation networks were compared to the area equipped for irrigation reported by the Global Map of Irrigated Areas (GMIA). Finally, the networks were created for inclusion in a global hydrological model and therefore the hydrological effects of the irrigation networks were evaluated. The created networks were compared to raster-based irrigation networks with 60’, 30’ and 5‘ irrigated areas and evaluated for the water demand, actual evapotranspiration, discharges at the river mouth and water allocation fractions simulated by the global hydrological model PCR-GLOBWB. These water allocation fractions were compared to the fractions reported by the GMIA.
This specific method has successfully created irrigation networks. OSM data seems to be a good source for constructing irrigation networks. The definition of the river discharge as source affected the created irrigation networks. Smaller discharges created more and smaller networks. The choice of an irrigation network influenced several hydrological parameters. Total water withdrawal increased for the OSM and 60’networks compared to the 30’ and 5’ networks due to the increasing ability of allocating surface water. This indicated that the water demands for the 30’ and 5’ networks were not met. Actual evapotranspiration showed a more reliable spatial distribution for the OSM based irrigation networks compared to the 30’ and 5’ irrigated areas. This was best seen in arid regions, such as the Nile delta. Discharge at the river mouth was highly overestimated using the 30’ and 5’ networks and slightly overestimated using the 60’ and OSM based irrigation networks. In general, water allocations fractions per water source were probably overestimated using the 60’ networks due to the cell size of the 60’ cells and were better computed using the OSM data, except for the Indus. The simulated water allocation fractions were spatially in line with the fractions reported by the GMIA. However, the validation statistics (Pearson correlation coefficient, Nash-Sutcliffe efficiency and the Percentage Bias) were in severe disagreement. This is probably due to the variability in irrigation extent between our irrigation networks and the ones created by the GMIA. So, OSM data can be used to create irrigation networks in future studies, however the OSM data needs to be improved.||