Integration of crop simulation models and GIS for regional yield forecasting under climate change scenarios in Northern Italy

Introduction

As Rosenzweig et al. [4] noted, coupling process-based crop models with spatial platforms allows researchers to move beyond point-scale simulations toward landscape-level yield predictions that account for spatial variability in soils, weather, and management. DSSAT is among the most widely validated crop modeling systems, with modules for over 40 crops including rice [1,2] . However, most DSSAT applications run at individual field sites, missing the spatial heterogeneity that drives yield variation across regions [3,6] . Northern Italy's Po Valley is Europe's largest rice-producing area, contributing 53% of EU rice output. Climate projections indicate warming of 1.5-4.2 °C by 2080 under RCP 4.5-8.5, with increased frequency of heat events during the reproductive stage [8,14] . Confalonieri et al. [7] simulated Italian rice yields at individual sites but did not spatialize results across the full production landscape. This research linked DSSAT-CERES Rice with ArcGIS Pro to generate spatially explicit yield projections for five Po Valley regions under current and future climates, identify vulnerable zones, and quantify the spatial pattern of yield change at 1 km resolution.

Materials and Methods

Theoretical background DSSAT-CERES Rice simulates daily phenological development, carbon assimilation, and yield formation as functions of weather, soil, and management inputs [1] . The model tracks thermal time accumulation for phenology, radiation use efficiency for biomass production, and a water balance for drought stress effects. The GIS component manages spatial input distribution and output aggregation across grid cells [5,13] . Each 1 km grid cell receives interpolated weather data and dominant soil profile parameters, runs an independent DSSAT simulation, and returns predicted yield for mapping and analysis [10] .

Simulation parameters

Weather data (1980-2020) from 47 ARPA Lombardia stations were interpolated to 1 km grids using thin-plate spline regression. Soil profiles (n = 312) were extracted from the ERSAL Lombardy soil database and assigned to grid cells by dominant soil mapping unit. Management inputs (sowing date, cultivar, fertilizer) reflected regional averages from ISTAT census data. Climate change scenarios used bias-corrected CMIP5 projections from five GCMs for RCP 4.5 and 8.5 at 2050 and 2080 time slices [8,14] . Model calibration used cv. Selenio genetic coefficients from 11 years of field trials at the University of Milan Experimental Farm (2010-2020) [9] .

Materials

The study covered five regions: Lombardy Plain, Po Delta, Piedmont Hills, Veneto Plain, and Emilia-Romagna, encompassing 234,000 ha of rice cultivation. DSSAT v4.8 was linked to ArcGIS Pro 3.0 via Python scripting. Validation used the coefficient of determination (R²), RMSE, normalized RMSE (nRMSE), and Willmott's d-index [12] .

Methods

The calibrated model was run for each 1 km grid cell (n = 23,400) under baseline and four future scenarios. Output yield maps were generated in ArcGIS Pro. Yield change was calculated as percentage deviation from the baseline mean. Hotspot analysis used Getis-Ord Gi* statistics to identify clusters of significant yield decline. Regional means were compared by one-way ANOVA with Tukey HSD at p≤0.05 [9] . [12]

Results

Comprehensive interpretation

The model reproduced observed yields well (R² = 0.87, d-index = 0.91; Table 1, Figure 1). Under RCP 8.5 by 2080, yields declined 22.4% (Emilia-Romagna) to 27.1% (Po Delta) relative to the baseline (Table 2, Figure 2). The Po Delta was most vulnerable due to the combination of heat stress and saline intrusion risk. Piedmont showed a slight yield increase under RCP 4.5 at 2050 (+5.2%) because current temperatures there are suboptimal for rice. Hotspot analysis identified 38% of grid cells showing statistically significant yield decline clusters under RCP 8.5 by 2080.

Discussion

The 22-27% yield decline under RCP 8.5 by 2080 is consistent with Asseng et al. [3] , who found 20-30% cereal yield reductions in southern Europe under high-emission scenarios. The spatial pattern of decline greatest in the low-elevation Po Delta and least in the higher Piedmont reflects the nonlinear temperature response of rice during flowering, where spikelet sterility increases sharply above 35 °C [8] . The GIS integration captured this gradient by resolving the 1-2 °C temperature difference between regions that point-scale models would miss [14] . The slight Piedmont yield increase under moderate warming highlights the importance of spatial analysis: blanket regional projections would obscure this local adaptation opportunity. The framework could be extended to test cultivar switching, sowing date shifts, and supplemental irrigation as adaptation measures [10, 13] .

Conclusion

Coupling DSSAT-CERES Rice with GIS produced reliable, spatially explicit yield forecasts for the Po Valley rice system. Under high-emission scenarios, yields are projected to decline 22-27% by 2080, with the Po Delta and low-elevation plains most affected. The framework identified both vulnerable hotspots and areas where moderate warming could benefit current production. These spatially resolved projections can directly inform regional adaptation planning and resource allocation for European rice production.

References

1. Jones JW, Hoogenboom G, Porter CH, Boote KJ, Batchelor WD, Hunt LA, et al. The DSSAT cropping system model. Eur J Agron. 2003;18(3-4):235-265.
2. Hoogenboom G, Porter CH, Boote KJ, Shelia V, Wilkens PW, Singh U, et al. The DSSAT crop modeling ecosystem. In: Advances in Crop Modelling for a Sustainable Agriculture. 2019. p.173-216.
3. Asseng S, Ewert F, Rosenzweig C, Jones JW, Hatfield JL, Ruane AC, et al. Uncertainty in simulating wheat yields under climate change. Nat Clim Change. 2013;3(9):827832.
4. Rosenzweig C, Jones JW, Hatfield JL, Ruane AC, Boote KJ, Thorburn P. The Agricultural Model Intercomparison and Improvement Project: Protocols and pilot studies. Agric For Meteorol. 2013;170:166-182.
5. ESRI. ArcGIS Pro: Geographic information system software. Redlands (CA): Environmental Systems Research Institute; 2022. Version 3.0.
6. Basso B, Cammarano D, Carfagna E. Review of crop yield forecasting methods and early warning systems. Rome: FAO; 2013. p.1-56.
7. Confalonieri R, Rosenmund AS, Baruth B. An improved model to simulate rice yield. Agron Sustain Dev. 2009;29(3):463-474.
8. White JW, Hoogenboom G, Kimball BA, Wall GW. Methodologies for simulating impacts of climate change on crop production. Field Crops Res. 2011;124(3):357-368.
9. Gomez KA, Gomez AA. Statistical procedures for agricultural research. New York: John Wiley and Sons; 1984. p.1-680.
10. Van Ittersum MK, Cassman KG, Grassini P, Wolf J, Tittonell P, Hochman Z. Yield gap analysis with local to global relevance. Field Crops Res. 2013;143:9-18.
11. Marin FR, Jones JW, Royce F, Suguitani C, Donzeli JL, Pallone Filho WJ, et al. Parameterization and evaluation of predictions of DSSAT/CANEGRO for Brazilian sugarcane. Agron J. 2011;103(2):304-315.
12. Willmott CJ. On the validation of models. Phys Geogr. 1981;2(2):184-194.
13. Holzworth DP, Huth NI, deVoil PG, Zurcher EJ, Herrmann NI, McLean G, et al. APSIM: Evolution towards a new generation of agricultural systems simulation. Environ Model Softw. 2014;62:327-350.
14. Zhao G, Siebert S, Enders A, Rezaei EE, Yan C, Ewert F. Demand for multi-scale weather data for regional crop modeling. Agric For Meteorol. 2015;200:156-171.