Improvement of Drought Prediction Model Through Sensitivity Analysis of Predictor Variables

The existing drought prediction model for Oklahoma climate division 8 can forecast the timing and intensity of drought, as indicated by the Palmer Drought Severity Index (PDSI), up to 18 months in advance. The statistical drought prediction is achieved by modeling the relationship between the PDSI and climate teleconnection indices. Past, present, and future-forecasted climate indices (Multivariate ENSO Index, Pacific Decadal Oscillation index, and Atlantic Multidecadal Oscillation index) and past and present PDSI are used as predictor variables. Applied interactions among predictors and the model has been optimized through stepwise regression to eliminate non-significant terms.

This study aims to enhance the model by conducting a sensitivity analysis of the predictor variables. The objective is to evaluate the responsiveness of the drought forecasts to changes in input variables and identify the most influential and uncertain predictors affecting PDSI forecasts.

The analysis begins with establishing a base case forecast by generating a drought prediction for a historical event using the current model. In the subsequent phase, sensitivity analysis will be performed by systematically perturbing each predictor while holding others constant. This approach will assess the impact of these perturbations on forecast outcomes and quantify the relationship between input uncertainty and forecast variability.

The study will present preliminary results that illustrate how changes in predictor variables affect the PDSI forecast. By isolating and analyzing the influence of each predictor, the research aims to identify key predictors that significantly impact forecast accuracy and highlight areas of uncertainty within the model.

The insights gained from this sensitivity analysis are expected to improve the existing drought model’s robustness and reliability. This study will provide valuable information for enhancing drought prediction accuracy and for informing decision-making processes in drought management.