Authors: Ujjwal Marasini – New Mexico State University
Title: Snow Water Equivalent Prediction for Northern New Mexico Using the Convolutional LSTM Machine Learning Method
Presentation Type: Poster
Abstract: In New Mexico, earlier snow melt is anticipated because climate models project an average temperature between 5°F – 7°F over the next 50 years. This will lead to a lack of water during the Summer and may even cause snow melt flooding. However, based on an extensive literature review, there are no studies focusing on the long-term Snow Water Equivalent (SWE) in the study area. We aim to fill this gap by making a long-term forecast of SWE for the northern part of the state by creating a ConvLSTM machine learning model for SWE and forecasting changes in SWE for the next 30 years under different climate change scenarios. The required hydrological, meteorological, topographical, and snow cover data are obtained from the National Hydrography Dataset, the National Centers for Environmental Information, the United States Geological Survey, and National Snow and Data Center. The ConvLSTM model is selected for the study because it can learn complex associations between inputs and outputs instantly, without prior knowledge of the underlying process and works with high accuracy. The output is evaluated using Nash-Sutcliffe efficiency, and root mean square error. This study will provide clear insights into future SWE in northern New Mexico using machine learning techniques. Overall, it will assist in policy making and help decision makers with the water allocation, flood risk management and overall planning of water resources.