Modelling rapeseed-pea and wheat-legumes intercrops under varying water availability in Mediterranean conditions

Modelling intercropping is still in its infancy, despite its potential to pre-screen the multiple scenarios where this practice can be beneficial. This study assessed the performance of the STICS soil-crop model in simulating three intercrops under Mediterranean conditions and used it to evaluate multi-year scenarios differing in water availability. The model was calibrated with sole crop (SC) data from two irrigated and two rainfed experiments conducted in Ebro valley (NE Spain). It was evaluated with intercropping (IC) data from an on-farm irrigated experiment involving rapeseed-pea (IC-RP), durum wheat-pea (IC-WP), and durum wheat-faba bean (IC-WF). Multi-year simulations (2000–2020) were conducted for SC and IC under irrigated and rainfed scenarios, relay-sowing dates were tested for IC-RP. Calibration yielded good results, particularly for grain yield (rRMSE ≤ 29%; MB ≤ 0.36 t ha −1 ). In evaluation, the model accurately captured the inversion of species dominance across years in IC-RP, with good agreement for grain yields (rRMSE = 22% for pea; 32% for rapeseed), though it underestimated grain-yield LER (LER Y ). A 2-month relay sowing (first rapeseed, then pea) maximised yields, improving LER Y in both rainfed (1.11) and irrigated (1.26) conditions, and achieving non-significant overyielding (+6%) under irrigation. For IC-WP and IC-WF, STICS exaggerated the dominance of one species in terms of LAI and crop height, leading to overestimated yields (MB = + 2.3 t ha −1 for faba bean), LER Y , and inaccurate N 2 fixation. Hence, results under rainfed and irrigated scenarios for these intercrops remain uncertain. Modelling intercropping in Mediterranean conditions with STICS remains challenging. Simulating species of similar height is particularly difficult, as competition for light and its impact on crop height still require improvements. Calibration strategies would likely benefit from combining optimisation tools with trial-and-error methods and from incorporating IC data into parameter estimation. • Three intercrops in Mediterranean conditions were modelled with STICS. • Rapeseed-pea intercropping shows higher benefits without water stress. • Rapeseed-pea yields are higher when relay sown (2-month delay for pea). • Poor simulation of light interception worsens overall model performance. • Modelling wheat–legume intercrops remains challenging under semi-arid conditions.