ARIMA, ARIMAX, or Univariate Linear Model? Modeling Sea Surface Temperature Anomalies' Affect on California's Crop Production

Abstract

It is well known that agriculture is a ected by local climate, but what are the impacts of factors such as El Ninos and La Ninas? This study examines the direct influence of ENSO events on Field crop, Nuts and Fruits, and Vegetables and Melons production in California. The purpose is to determine if knowledge of ENSO events is enough to give an accurate forecast of crop production that way policy and decision makers can be effective in how they manage resources and plan. The relationship between these two variables is represented by univariate linear, ARIMA, and ARIMAX models. These models are compared to one another and tested for their ability to fit and predict crop observation using ENSO as an exogenous variable. Ultimately, ARIMAX was the best at fitting and ARIMA produced the best forecast data. The univariate linear model didn't perform well in both regards. One major component that influenced that results was the method of data transformation. By using spline interpolation, unnecessary and random noise was added to the series. Therefore, the final models may not be as accurate as expected. Nonetheless, this process of creating, comparing, and testing models has shed light on modeling procedures and various topics to consider such as cointegration.