Abstract
Understanding the relationship between precipitation and groundwater levels is important for effective water management, particularly amidst climatic uncertainties and urban development. Lee County, Florida, the focus of our initial exploratory study, has undergone rapid growth, amplifying the relevance of this research. Our investigation spans ten years (2004-2014), during which we analyzed the precipitation patterns and groundwater elevations at the Corkscrew Water Plant. Our initial analyses revealed that although there appears to be a marginal downward trend in precipitation over this period, this change was not statistically significant (coefficient = -1.24 X 10 (super -9) , p-value = 0.72). In contrast, we observed a subtle yet significant upward trend in groundwater elevations (coefficient = 4.92 X 10 (super -9) , p-value = 0.002). The divergence between precipitation and groundwater levels underscores a potential disparity, highlighting the need for more comprehensive studies. We deployed a variety of 'train and test' models to forecast trends for 2015, selecting those with minimal errors. When forecasting precipitation, the TBATS model delivered the lowest Mean Absolute Percentage Error (MAPE) score of 612.72. In contrast, the Seasonal Autoregressive Integrated Moving Average (SARIMA) model outperformed others with a MAPE score of 4.73 in predicting groundwater levels. While the MAPE value for the precipitation forecast may not seem compelling, it's noteworthy that a positive correlation was found between actual and forecasted values (y = 0.5738x + 0.7319, R (super 2) = 0.90), showing a slight underestimation of precipitation in the forecasts. As Lee County experiences continuous expansion, acquiring fresh insights into the relationship between development and water resources across the region becomes essential, which will enable us to comprehend the implications of changes in land use and population growth on the unconfined aquifer more effectively.