Use of Convection Permitting climate models for maize yield projection over Italy

Agriculture is highly vulnerable to temperature increase and variations in precipitation patterns associated with climate change. The Mediterranean region is considered a hotspot, with Italy being particularly affected by a raise in the frequency and severity of prolonged periods of drought and extreme floods. The IPCC reported that maize and wheat yields have been negatively affected by the observed climatic changes in several lower-latitude regions during recent decades. Cereal production constitutes a key asset for Italy’s agricultural sector, with wheat and maize being the main cultivated crops, reaching together the 79% of the total harvested area. However, in Italy there is still limited information on the effects of climate change and extreme weather events on maize production, particularly at very high-resolution spatial scale. Given the peculiar topography of the Italian landscape and the sudden spatial variations of weather variables due to the country’s orography, the use of very high spatial resolution climate data could significantly contribute in offering better detailed future crop yield projections. The km-scale Convection Permitting Models (CPMs), which provide a more realist representation of hourly precipitation and dry hours compared to coarser resolution models, could constitute an interesting tool to project future yield with a very fine spatial scale.

This study uses CPMs from the CORDEX-FPS on Convective Phenomena over Europe and the Mediterranean (FPS Convection) to drive the Agricultural Production System sIMulator (APSIM) crop model to project maize yield under RCP 8.5 over the 2090-2099 period.

At first, the ability of APSIM in simulating the observed maize yield at province scale in Italy over the period from 2006 to 2023 is assessed. In this phase, the APSIM crop model is initialized with weather data from the reference dataset Era5Land remapped at the same spatial resolution of CPMs. Then, the performance of nine CPMs (run with boundary conditions provided by ERA-Interim) in reproducing the simulated maize yield over the 2000-2009 period is evaluated. The APSIM crop model is subsequently run over the 1996-2005 period with weather data from CPMs with boundary conditions provided by their respective GCM. Finally, APSIM is run over the 2090-2099 period under RCP 8.5 to get projections of future maize yield.

Results have shown that the APSIM crop model is capable of simulating maize yield over Italy at province scale, with an overall correlation between observed and simulated maize yield of 0.92 (initialization) and 0.86 (testing). Moreover, maize yield simulated through the use of CPMs shows a good agreement with maize yield simulated with Era5Land, with correlation from 0.79 to 0.91 (p<0.001) depending on the considered CPM. At province level, CPMs perform better in the high-producing areas, such as the Po Valley, while the correlation decreases over provinces with significant areas located on the Alps or the Apennines. Finally, over the 2090-2099 period, maize yield will decrease up to -30% over the Po Valley provinces, while it will increase at higher altitudes.

Results demonstrated the importance of high-spatial resolution yield projections to evaluate future adaptation strategies.