Unlocking the Potential of Dynamical Models for Drought Forecasting in Iran: Insights From Multi-Model Ensemble Analysis

Abstract

This paper assesses dynamical models to construct monthly (January through December for lead times of 0.5–2.5 months) and seasonal (January–March [JFM], April–June [AMJ], July–September [JAS], and October–December [OND] for lead times of 1.5–3.5 months) forecasting of drought based on the standardized precipitation evapotranspiration index (SPEI) over Iran. The air temperature (minimum, maximum, and mean) and precipitation data, as the components of SPEI, are forecasted using six North American Multi-Model Ensemble (NMME) and European Centre for Medium-Range Weather Forecasts (ECMWF) SEAS51 as well as their ensemble multi-model mean (MMM) for a common period from 1991 to 2021. These forecast data are interpolated to stations using inverse distance weighting, and then the SPEI is computed for each model. The observed SPEI is calculated for 67 synoptic stations across Iran. The SPEI forecast skill of the MMM surpasses that of individual models. Additionally, MMM demonstrates improved forecast skill during wet and cold months (November–March) compared to dry and warm months (June–September). There is a statistically significant Pearson correlation coefficient between observed and forecast JFM SPEI in most areas of the study area for lead times of 1.5, 2.5, and 3.5 months at a 5% significance level. Moreover, the SPEI forecast is significant in most areas for JFM, AMJ, and OND for the 1.5-month lead time. The canonical correlation analysis is employed to investigate the relationship between observed global sea surface temperature anomalies (SSTA) and seasonal SPEI to achieve insights into the source of drought predictability in Iran, as well as how the skill of the MMM forecasts is affected by SSTA. The spatial pattern root mean square error of the MMM forecasts and SSTA is similar. The canonical correlation coefficient between SSTA and observed SPEI is stronger than in JFM, indicating that MMM exhibits promising potential for SPEI forecasts.