Drought monitoring through a joint remote-sensing based index

Drought monitoring is one of the significant issues of water resources assessment, in fact appropriate water resources management strategies can be timely implemented thanks to drought early warning systems. Drought characterization, through the determination of onset, offset, severity, intensity and duration, is crucial to manage and reduce the risk. A single variable can be insufficient to fully characterize drought. More than 100 indices have been proposed to monitor drought, each one with its aim and specificity. In addition, drought monitoring can be a hard task in countries with few meteorological stations on the ground. Drought events cause high impacts on various economic sectors, but agriculture is the most affected due to its strong dependency on water availability. The present study aims at implementing a framework to identify drought events that cause impacts on crops, through the following two steps: 1. Creation of an index that combines, by using a bivariate normal distribution function, two of the most renown drought indices, the Standardized Precipitation Index (McKee, 1993), which is based on precipitation only, and the Vegetation Health Index (Kogan, 1997), which is a measure of the effects of drought on the vegetation. 2. Implementation of a framework to identify drought events, by establishing starting and end date of the events on the basis of the percentage of a country’s area under drought according to the new index. Two remote-sensing datasets with a global coverage have been employed. Precipitation were retrieved from CHIRP (Funk, 2015), and the VHI was retrieved from the Global VHP of the NOAA (NOAA, 2011). SPI was computed from CHIRP precipitation and updated every week. Haiti was chosen as case study; the country was affected by intense droughts that caused huge loss of crops and deeply affected the population. Events were identified starting from text-based information. Drought events identified by the proposed framework were compared with the observed events. The Receiver Operating Characteristic curve was employed in the validation process. Results showed that the new index was able to identify all major drought events hitting Haiti. The approach here proposed showed significant advantages: can be easily implemented over the entire globe at country-scale; can be applied in areas with sparse gauge coverage, and the index can be updated in near-real time, having both the datasets a short latency period.