Drought prediction boosted by AI

These AI models were shown to forecast droughts with high levels of accuracy

Spotted: An estimated 55 million people are negatively affected by droughts globally every year. To mitigate these effects, meteorologists monitor changes in precipitation, temperature, and other factors to determine when droughts will occur and combat them. However, current monitoring systems can be unreliable. With the help of AI, some researchers are working to change that.

To start, the researchers – from the University of Sharjah, UAE, and La Trobe University, Australia – took drought indicator data from seven meteorological stations in Alice Springs, Australia, between 1985 and 2020. They then applied nine conventional drought indices, including the Standard Precipitation Index, Palmer Drought Severity Index, and Percent of Normal Index, comparing the outcome of the indices with reality.  

To test the AI, they used six soft computing models, including decision trees, deep learning, and generalised linear models with their data, also comparing the results against what actually happened. After testing the conventional indices and AI models against the drought indicators, they found that all models were accurate and correlated with traditional indices. However, the generalised linear model correlated best with indicators, suggesting its potential for more precise drought assessment.

If adopted to monitor drought conditions globally, these AI models could standardise monitoring and provide more precise drought forecasting. With the accurate and quick results these AI models provide, preparations could begin well in advance to reduce the effect of droughts. Plus, it could enable better water optimisation strategies to ensure healthy crop growth and food security in arid climates.

Written By: Joshua Solomon