Seasonal forecasting of Agroclimatic indices in Mediterranean areas: testing tools for improving predictability

Abstract:

Land managers might greatly benefit of accurate climate predictions with several months in advance (i.e. seasonal forecasts) for best decision making in planning activities and allocate resources, also under the context of climate change mitigation and adaptation. However, previous studies showed that the accuracy of seasonal predictions was limited over extra tropic regions such as Europe. Currently, the MEDSCOPE project is developing tools to improve the accuracy of seasonal predictions and thus the production of climate services over the Mediterranean area. This work aims to assess the performance of some post-processing tools produced in MEDSCOPE (R package CSTools) using agroclimatic indicators related to the agriculture and forestry sectors (e.g. fire risk, vegetation thermal needs and water availability). Seasonal forecasts from the CMCC SPSv3 model were corrected using different techniques implemented on CSTools, named simple bias correction (SBC), calibration (CAL) and quantile mapping (QM), that differently correct the statistical properties of climate datasets. Correlations with up-scaled ERA5 reanalysis climate data showed that QM provides similar spatial patterns as raw data (no post-processing), whereas with CAL and SBC the extent of significant correlation decreased. The study suggests that results are highly influenced by the selected method and thus a cautious use is recommended.