Assessment of December to May 1-month lead statistical drought hindcasts for the Philippines

Abstract

We present the quantitative assessment of 1-month lead statistical hindcasts for detecting drought during the dry season in the Philippines using the Standardized Vegetation−Temperature Ratio (SVTR) drought index. The Oceanic Niño Index (ONI) and the satellite-derived measurements of land surface temperature (LST) and the Normalized Difference Vegetation Index (NDVI) from the Moderate Resolution Imaging Spectroradiometer are used as predictors for the hindcasts. We employ the Autoregressive Integrated Moving Average (ARIMA) model to generate 1-month lead SVTR hindcasts using the aforementioned predictors for 2011 to 2022. Nationwide hindcasts are accurate by (70±10)% across December to May. Areas with 100% hit rate tend to follow the monsoonal rains from December to February; however, the chance of false alarms increased as well. March to May had no chance of false drought warnings for nearly the entire country. Consolidating the different verification metrics, forecast reliability maps indicated ARIMA had high skill in predicting non-drought areas, particularly from February to May. Hindcasts were reliable in discriminating drought and non-drought areas in Maguindanao and Davao del Sur for January and select regions in Mindanao for May. Low reliability for forecasting drought elsewhere may partly be due to the infrequent drought occurrences, wherein we recommend using other forecasts and drought indices for these cases. The developments of this research will guide stakeholders and water managers in their drought mitigation and early warning response, more so in light of the anticipated El Niño.