Predicting a drought six months in advance was a utopia. The UPV has achieved this with a system that uses AI

In recent years drought episodes have intensified in some regions and fear of a global drought flies over the environment. In this scenario, a team of researchers from the Polytechnic University of Valencia have created a system that can predict whether there will be a drought six months in advance.

The system. The work has been carried out by the team from the Institute of Water and Environmental Engineering (IIAMA) of the UPV and has been published in the journal Earth Systems and Environment. The method integrates predictions from four reference climate systems (ECMWF-SEAS5, Météo-France System8, DWD-GCF2.1 and CMCC-SPSv3.5) and are processed using artificial intelligence techniques.

From this data, the team calculated two of the most important international drought indices (the Standardized Precipitation Index and the Standardized Precipitation-Evapotranspiration Index), using data windows of 6, 12, 18 and 24 months. The method has been applied in the Júcar River basin, which usually goes through stages of recurrent and quite intense droughts.

Why is it important. The novelty of this system is that it is not limited to using a single climate model or index, but rather it merges three pieces that are usually used separately and adds AI processing to correct biases and adapt the models to a regional scale. This allows the prediction to be more reliable since it does not depend on a single model. Furthermore, all of this has been integrated into an operational web toolintended to be used in water management and not only as an academic exercise.

Results. The system is correct with a reliability of 90% when the prediction is made for that same month. If they want to obtain predictions three months in advance, the reliability is 60%, while for longer periods (12, 18 and 24 months) they do not give a percentage, but they affirm that the model is still useful for predicting what will happen up to six months in advance. Héctor Macián, co-author of the study, states that “The results confirm that the system is especially effective in reinforcing early warning of droughts, a fundamental aspect to anticipate management measures, reduce socioeconomic impacts and increase resilience to climate change.”

Action window. As we said, the methodology has been developed in the Júcar river basin, which is a semi-arid area with long, dry and very hot summers, although researchers highlight that it is transferable to other drought-prone areas. Being able to foresee these episodes with up to six months of margin opens a window to implement the drought management plans much more in advance and thus be able to mitigate the effects.

Image | UPV

In Xataka | The remains of an ancient Mayan city leave us lessons for the future: an amazing system against drought