Modeling root water uptake patterns of oil crops grown on semiarid loess

Oil crops are often used as cover crops by smallholders managing orchards on drylands, because they have an economic return and help regulate ecosystem services. Improving the ability of models in root water uptake (RWU) simulation is important to find the right crop species and predict the dynamics of these oil crops in water -limited regions. In this study, two different process-based models were used to simulate the RWU of three oil crops, i.e. soybean (Glycine max L., SYB) and two canola cultivars (turnip type rape (Brassica campestris L.), TTR, and cabbage type rape (Brassica napus L.), CTR), grown on the semiarid Loess Plateau of China, based on two-year in-situ measurements. The first model, IM-SVD, was developed by inverse modeling (IM) of the Richards equation and incorporating singular value decomposition (SVD) parameter optimization algorithms, which simulated the RWU pattern (including soil evaporation and plant transpiration). The second model was the Feddes model implemented in one-dimensional HYDRUS (HYDRUS 1-D). Both models reliably predicted the soil water content (SWC) in different layers. However, the IM-SVD model showed better modeling accuracy of SWC, and produced higher simulated RWU values than the HYDRUS 1-D model irrespective of crop. The results from both models implied that SYB had a higher daily and seasonal RWU than TTR and CTR, and its water source was derived mainly from the top 30 cm soil layer. Additionally, the simulated RWU was positively and significantly (p<0.05) correlated with fine root length density (FRLD). The RWU showed lagged correlations with SWC, which varied between crop species and growth stages, but overall were more positively correlated with the SWC on preceding days and more negatively correlated with the SWC on subsequent days. Our study provides insights into the improving RWU simulation for oil crops and similar crops in drylands.