Development of a Statistical Predictive Model for Daily Water Table Depth and Important Variables Selection for Inference

TL;DR

This paper develops a statistical model to accurately predict daily water table depth using hydroclimatic data, aiding groundwater management and ecological conservation.

Contribution

It introduces a flexible BigVAR-based approach that incorporates sparsity, autoregression, and lag effects for water table prediction using multi-site hydroclimatic data.

Findings

RMSE of 14.94 cm during testing

R^2 of 0.96 in a wet year

Key variables include solar radiation, rainfall, wind direction

Abstract

Accurately predicting water table dynamics is vital for sustaining groundwater resources that support ecological functions and anthropogenic activities. This study evaluates a statistical model (BigVAR) that handles three major flexibilities: (a) prediction under a sparsity assumption in coefficients, (b) consideration of a time series autoregression framework, and (c) allowance for lags in both dependent and independent variables for estimating water table depth using daily hydroclimatic data from the USDA Forest Service Santee Experimental Forest (SC) and a site in NC. Data from 2006--2019 (SC) and 1988--2008 (NC) were used, with key predictors including soil and air temperature, precipitation, wind, and radiation. For WS80, RMSE during the dormant season was 10.09 cm, with a daily testing phase RMSE of 14.94 cm. The model achieved an R^2 of 0.93 for 2019 (a dry year) and 0.96 for 2016 (a wet year). Solar radiation, rainfall, and wind direction were among the most influential variables. This predictive model aids in managing wetland hydrology and supports decision-making for forest managers and hydrologists.