Development of high-resolution gridded data for water availability identification through GRACE data downscaling: Development of machine learning models
Article
Article Title | Development of high-resolution gridded data for water availability identification through GRACE data downscaling: Development of machine learning models |
---|---|
ERA Journal ID | 1956 |
Article Category | Article |
Authors | Tao, Hai, Al-Sulttani, Ahmed H., Salih, Sinan Q., Mohammed, Mustafa K.A., Khan, Mohammad Amir, Beyaztas, Beste Hamiye, Ali, Mumtaz, Elsayed, Salah, Shahid, Shamsuddin and Yaseen, Zaher Mundher |
Journal Title | Atmospheric Research |
Journal Citation | 291, pp. 1-13 |
Article Number | 106815 |
Year | 2023 |
Publisher | Elsevier |
Place of Publication | Netherlands |
ISSN | 0169-8095 |
1873-2895 | |
Digital Object Identifier (DOI) | https://doi.org/https://doi.org/10.1016/j.atmosres.2023.106815 |
Web Address (URL) | https://www.sciencedirect.com/science/article/pii/S0169809523002120?dgcid=coauthor |
Abstract | Estimation of total water availability has paramount importance in planning sustainable development of a region, particularly in arid water-scarce areas. Coarse-resolution of existing total water availability or terrestrial water storage anomaly (TWSA) data is the major limitation of their applications in different sectors. An attempt has been made to downscale Gravity Recovery and Climate Experiment (GRACE) TWSA data to develop a high-resolution gridded data product of the total water availability of Iraq. European reanalysis (ERA5) precipitation, evapotranspiration, surface runoff, subsurface runoff and soil water contents data were used to downscale GRACE 1.0° spatial resolution monthly TWSA to 0.1° spatial resolution for the period 2002–2020. A machine learning (ML)-based recursive feature elimination algorithm was used to identify the optimum input combination according to the nonlinear relationship of ERA5 variables with GRACE water equivalence data. The selected subset of inputs was used to develop the downscaling models using three classical ML algorithms for the available GRACE measurement points over Iraq. The models were calibrated at 70% of GRACE grid point locations and validated in the rest of the points. Finally, the model was used to predict TWSA at each ERA5 grid point to generate Iraq's high-resolution water availability dataset. The results showed higher performance of random forest in downscaling TWSA compared to other algorithms. The model estimated the TWSA at validation points with Kling-Gupta Efficiency (KGE) in the range of 0.5–0.91 and Nash-Sutcliff Efficiency (NSE) between 0.54 and 0.88. The modelled high-resolution TWSA data shows higher availability of water resources in the north, particularly northeast of Iraq, and the least in the southeast. The technique developed in this study can be implemented in developing a high-resolution gridded water availability dataset from satellite GRACE data in the region where in-situ estimation is very limited. |
Keywords | Water equivalence data ; Downscaling; GRACE; ERA5; Machine learning |
Contains Sensitive Content | Does not contain sensitive content |
ANZSRC Field of Research 2020 | 461199. Machine learning not elsewhere classified |
Public Notes | Files associated with this item cannot be displayed due to copyright restrictions. |
Byline Affiliations | Qiannan Normal University for Nationalities, China |
Guizhou University, China | |
University of Kufa, Iraq | |
Al-Bayan University, Iraq | |
University of Warith Al-Anbiyaa, Iraq | |
Galgotia college of engineering, India | |
Istanbul Mediniyet University, Turkiye | |
UniSQ College | |
University of Prince Edward Island, Canada | |
University of Sadat City, Egypt | |
Al-Ayen University, Iraq | |
University of Technology Malaysia, Malaysia | |
King Fahd University of Petroleum and Minerals, Saudi Arabia |
https://research.usq.edu.au/item/yy53v/development-of-high-resolution-gridded-data-for-water-availability-identification-through-grace-data-downscaling-development-of-machine-learning-models
64
total views1
total downloads7
views this month0
downloads this month