Performance and sensitivity to climate change of two DSSAT-CSM cassava models for the Brazilian cultivar BRS Formosa

Context or problem

Cassava is well-adapted to diverse climates, but its response to future climate conditions remains uncertain. Crop growth models are crucial for predicting climate effects on yield and evaluating strategies to mitigate or adapt to these changes. However, these models require evaluation to simulate yield accurately.

Objective or research question

The goal of this study was to evaluate the performance of the of the CSM-CROPSIM-Cassava and CSM-MANIHOT-Cassava models in the DSSAT platform for simulating the yield of the drought- and bacterial blight-tolerant cassava variety BRS Formosa in Brazil’s northeast climate. The objectives were: i) to calibrate and assess model performance for estimating BRS Formosa yield, and ii) to analyze the models' sensitivity to changes in temperature, rainfall, and CO₂, exploring their potential for climate impact studies.

Methods

Calibration was based on the irrigated treatments from Embrapa experiments conducted in Cruz das Almas (2017–2020) and Petrolina (2015/16), while for evaluation the data from rainfed treatments and additional experiments in Cruz das Almas (2012–2018), Guanambi (2013–2015), Laje (2012/13, 2015/16), and Petrolina (2013/14), were used. Calibration involved adjusting 23 genetic coefficients for CSM-CROPSIM-Cassava and 13, for CSM-MANIHOT-Cassava.

Results

Model performance was evaluated using statistical indices, showing good accuracy for cassava storage root yield, with MAEs of 2195 and 2425 kg ha⁻¹ and d values of 0.92 and 0.83 for CSM-CROPSIM-Cassava and CSM-MANIHOT-Cassava, respectively. The ensemble of both models, applied to final storage root yield only, further improved performance (MAE = 1973 kg ha⁻¹; d = 0.91). Both models also simulated phenology and canopy traits (leaf number, stem and aboveground biomass) with good accuracy, and were sensitive to air temperature, rainfall, and [CO₂].

Conclusions

The ensemble provided the most reliable estimates of final storage root yield with the current DSSAT versions, whereas the individual models remain valuable for simulating phenology and canopy traits.

Implications or significance

The model analyses from this study are essential for estimating future cassava yield in Northeast Brazil, assessing climate risks, and guiding adaptive strategies for sustainable production.