Big Earth Data And Its Applications For Water Management And Land Degradation In The Nile Basin

Sustainable use of natural resources is fundamental to human development,and therefore land and water resource management from plot to basin-scale has become even more critical.Increasing perturbation of land and water resources by human activities has led to an imbalance between the supply capacity of the Earth system and human expectations,especially in water-limited basins such as the Nile Basin.However,ensuring sustainable land and water resource use requires a considerable amount of high-quality data and easy access to monitoring,analysis,and planning tools.The Nile Basin is located on the African continent,spanning 11 developing countries,3 of which are highly underdeveloped.The scarcity of ground observation and the severe lack of data sharing between countries have dramatically hindered the basin’s rational allocation and sustainable land and water resources use;furthermore,the uneven distribution of water resources in the basin has led to different socioeconomic problems and,in severe cases,conflicts of interest between countries.The development of Earth observation technology can effectively fill the gaps in the measured data.However,there is limited research on improving the quality of current remote sensing products and tools to meet the needs for a rational assessment of land and water resources development,especially in the Nile Basin,which has a huge area and a lack of ground observation.Using a big data approach,this study uses various Earth observations to generate and improve the relevant parameters required for land and water resource assessment in the Nile Basin.Specific objectives include fusing and generating high spatial and temporal resolution precipitation and evapotranspiration datasets;and further quantifying the risk of water depletion from human-managed land covers and soil erosion,as well as the risk of desertification and land degradation in the Nile Basin;all contribute to achieving land and water-related Sustainable Development Goals（SDGs）,eight years to achieve them.Accurate precipitation data at high spatiotemporal resolution are essential for land and water management at the basin scale,representing the first largest water balance pole.For that,the study applied a spatiotemporal statistical downscaling framework on Tropical Rainfall Measuring Mission（TRMM）precipitation products by integrating the Artificial Neural Network（ANN）machine learning algorithm and Earth observation data from Google Earth Engine（GEE）in Google Colaboratory（Colab）.Geospatial predictors in the downscaling model include one vegetation index（either Normalized Difference Vegetation Index,NDVI;Enhanced Vegetation Index,EVI;or Leaf Area Index,LAI）,topography（Elevation）,and geolocations（Longitude and Latitude）.The spatiotemporal statistical relationships between annual precipitations and annually-averaged vegetation indices were used to downscale TRMM precipitation data from 25 km to 1 km;the latter was converted to monthly precipitation maps by disaggregation.The study proposed three high-resolution precipitation products:ANN-EVI,NDVI,and LAI downscaled products.The validation with rain gauge observations revealed that the ANN-EVI downscaled model yielded higher agreement and accuracy with lower error(coefficient of determination（R²）=0.62,Root Mean Square Error（RMSE）=324 mm yr^-1,and Mean Absolute Error（MAE）=225 mm yr^-1)compared to the original TRMM precipitation(R²=0.58,RMSE=330mm yr^-1,and MAE=237 mm yr^-1).The study also proposed an enhanced monthly downscaled composite as follows;NDVI-based downscaled product for March,EVI-based downscaled products for January,February,May,June,September,November,and December,and LAI-based downscaled products for the remaining months（April,July,August,and October）.The study automatically optimized the models’parameters,estimated features’importance,and downscaled the TRMM precipitation product to 1 km from 2003 to 2019.The proposed high-resolution precipitation products can be used in models that use precipitation as a key input factor for better land and water management.Actual Evapo Transpiration（ET）is needed in various hydrology,climatology,ecology,and agriculture applications.Therefore,ET plays an essential role in land and water management,representing the second largest water balance pole.Remote sensing-based estimation is the only viable and economical method for ET estimation over a large scale.Although ET estimations over a large scale and a long time are challenging,several global ET datasets are available with considerable uncertainty associated with various assumptions regarding their algorithms,parameters,and inputs.In solution,the study validated and investigated 12 global ET products across various land surface types and conditions to select the best performing ET products and then produce a global long-term synthesized ET set using a high-quality flux Eddy Covariance（EC）ET covering the entire globe.According to six comprehensive validation criteria,the evaluated ET products were ranked based on the lowest error metrics and highest accuracy and consistency over different classification levels and times to choose the ensemble members.Hence,this study proposes a long-term synthesized ET product at a kilometer spatial resolution and monthly temporal resolution from 1982 to 2019.Among the 12 global ET products,Penman-Monteith-Leuning（PML）,the operational Simplified Surface Energy Balance（SSEBop）,the Moderate Resolution Imaging Spectroradiometer（MODIS,MOD16A2105）,and the Numerical Terradynamic Simulation Group（NTSG）ET products were chosen to create the synthesized ET set.The proposed synthesized ET product agreed well with flux EC ET over most of all comparison levels with a coefficient of determination（R²）>0.70,maximum Relative Mean Error（RME）at 13.94?mm（17.13%）,and maximum Relative Root Mean Square Error（RRMSE）of 38.61?mm（47.45%）.Furthermore,the product performed better than local ET products in China,the United States,and the African continent and presented ET estimates across all land cover classes.Therefore,the proposed ET can be used without looking at other datasets and performing further assessments for a local,basin,or global studies.The proposed global ET dataset represents the total water consumption;however,quantifying the amount of ET from human activities compared to climate is critical for basin-level integrated water resources management to achieve sustainable water use and peaceful development.Herein,we applied a data-driven framework that integrates Random Forest Regressor（RFR）machine learning and Earth observation data from Google Earth Engine（GEE）in Google Colaboratory（Colab）to distinguish between human activities and natural forces to the total water consumption of managed land covers.The framework effectively distinguished natural forces and human activities to ET from rainfed,mixed,irrigated croplands and urban areas in the Nile Basin with a coefficient of determination（R²）and Nash-Sutcliffe Efficiency（NSE）>0.85,Relative Bias（RB）within 1%,Mean Absolute Error（MAE）<24 mm yr^-1 and Root Mean Square Error（RMSE）<37 mm yr^-1.The study results also revealed that precipitation in sub-humid and humid regions and latitude and specific humidity in arid and semi-arid parts are the top essential predictors in building the ET separation prediction model.The results showed that human activities that occur in drier basins tend to increase water depletion more significantly.The results also showed that the contribution of human activities to the total water consumption reached69%in the arid area,reduced to 23%in semi-arid areas,and less than 15%in sub-humid and humid areas of the Nile Basin.In light of human activity trends from 2003 to 2019,the study revealed an increasing effect in arid and semi-arid areas（ranges from 2%to 5%）and a decreasing effect in sub-humid and humid regions（ranges from 0.4%to 2%）.The key findings of this study and their implications could be helpful for land and water manager in the Nile Basin to achieve water-related SDGs.The Nile Basin faces an uneven distribution of water resources resulting in different socioeconomic problems.Accordingly,upper and lower riparian countries’water availability consequences are sensitive issues and might be a source of conflict due to their vast effect on societies’stability.Herein,the study aims to understand the spatial distribution of the soil loss by water and vulnerability to desertification in light of the Nile Basin water availability.For that,two well-known frameworks,including the Revised Universal Soil Loss Equation（RUSLE）and the MEditerranean Desertification And Land Use（MEDALUS）,were fully implemented in Google Earth Engine（GEE）environment to investigate the environmentally sensitive areas（ESA）to desertification and land degradation,giving more consideration to soil erosion as a key input factor in land degradation assessment.The results of the RUSLE model showed that the highly prone areas to soil loss by water are the Ethiopian highlands,followed by the Equatorial Lakes region.The main driver of soil loss in these areas is the erosive rainfall that falls in a short period on steep slopes and fragile soils.Conservation practices that can reduce soil loss by water are highly needed in these areas.The results of the MEDALUS framework indicated that the tailwater areas are highly affected by desertification and land degradation phenomena due to the lack of water availability.Generally,these results indicate that soil erosion by water and sensitive areas to desertification are spatially distributed inversely according to water availability,whereas less sensitive areas to desertification are more likely to occur in areas with higher water availability（P>ET）which are in the same time more vulnerable to soil loss by water;vice versa（P<ET）.The study implications can help design appropriate management scenarios for soil and water conservation,reservoir management,and agricultural development of the Nile Basin to achieve land-related SDGs.Generally,the study proves the effectiveness of comprehensive utilization of machine learning and cloud computing in land and water management to help achieve SDGs at a basin scale.Furthermore,the study proved that downscaling precipitation plays a significant role in mapping soil loss by water（RUSLE estimates）;in contrast,a nonsignificant effect of the downscaling precipitation is observed in mapping human-induced total water consumption（ET separation estimates）and desertification and land degradation（MEDALUS estimates）.In addition,the study concludes that the arid and semi-arid parts of the Nile Basin are experiencing high human interventions and low water availability resulting in high vulnerability to desertification and land degradation.In contrast,sub-humid and humid regions are experiencing low human interventions,and high water availability results in low vulnerability to desertification and land degradation,except for the Ethiopian highlands,where considerable soil loss by water exists.The innovation of this study is mainly reflected in the following three aspects:1)a global monthly remote sensing ET dataset from 1982 to 2019 at 1 km spatial resolution is generated,which fills the gap of long-time series of ET data;2)a data-driven approach is proposed to quantify the contribution of human activities and natural effects in ET,which provides accurate management of water resources in managed land covers as a new solution for the cognition of water consumption by human activity;and 3)Earth observations,machine learning,and cloud computing are fully utilized as online mapping and assessment framework for spatiotemporal statistical downscaling of precipitation,soil water erosion,and desertification and land degradation,which provide good support for land and water resource management in data-scarce basins for achieving land and water-related SDGs.Although this study has achieved its primary objectives,the study implications and future directions need real and urgent collaboration and communication between the Nile countries’scientists and communities,particularly regarding data and information sharing for better basin-wide land and water management.In addition,an integrated nexus strategy and plan for the inter-and intra-country is desired to address the challenges related to water and food security in the Nile Basin.