Abstract Archive Select a year below to view:



Scaling down GRACE data for smaller regions: utilizing artificial neural networks and climate data for enhanced hydrological predictions in the state of Mississippi
Proceedings of the 2023 Mississippi Water Resources Conference

Year: 2023 Authors: Awawdeh A.R., Yasarer H., Pulla S., Kumar M.


The importance of having accurate and high-resolution hydrological data has risen with recent climate change and ongoing dependence on underground water. The launch of the Gravity Recovery and Climate Experiment (GRACE) in 2002 made it possible to acquire such data, enabling researchers to extract information on terrestrial water storage, ice loss, and sea-level change at a temporal resolution of one month. However, this data is not adequate for smaller regions due to the coarseness of the GRACE data grids, which have a resolution of 56 km by 56 km. This study aims to investigate the effectiveness of using Feedforward Artificial Neural Networks (ANNs) with the backpropagation error algorithm to scale down GRACE data for the State of Mississippi to smaller grids of 4 km by 4 km that can be used in smaller regions. This process utilized Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) and TerraClimate databases. CHIRPS provides high-resolution rainfall data, while TerraClimate provides a dataset of monthly climate and climatic water balance for global terrestrial surfaces. A script in Python programming language executed via Jupyter Notebook was developed to download all the necessary data and develop the ANN model. Initial results indicated that the ANN approach performed well with an R2 of 0.93, and the developed models can be utilized to predict equivalent water thickness with high accuracy in the Mississippi region. The current effort is focusing on validating the model by comparing the downscaled GRACE data against well-water data in the Mississippi Delta area and assessing the model's ability to identify drought activities that occurred within the study period from 2002 to 2020.

Tweets by @MS_WRRI