Roots are an important water transport pathway between soil and plant. Root water uptake (RWU) plays a key role in water and heat exchange between plants and the atmosphere. Inaccurate RWU schemes in land surface models are one crucial reason for decreased model performance. Despite some types of RWU functions being adopted in land surface models, none have been certified as suitable for maize farmland ecosystems. Based on 2007–2009 data observed at the maize agroecosystem field station in Jinzhou, China, the RWU function and root distribution (RD) in the Common Land Model (CoLM) were optimized and the effects of the optimizations on model performance were compared. Results showed that RD parameters calculated with root length density were more practical relative to root biomass in reflecting soil water availability, and they improved the simulation accuracy for water and heat fluxes. The modified RWU function also played a significant role in optimizing the simulation of water and heat fluxes. Similarly, the respective and integrated roles of two optimization schemes in improving CoLM performance were significant during continuous non-precipitation days, especially during the key water requirement period of maize. Notably, the improvements were restrained within a threshold of soil water content, and the optimizations were inoperative outside this threshold. Thus, the optimized RWU function and the revised RD introduced into the CoLM model are applicable for simulation of water and heat fluxes for maize farmland ecosystems in arid areas.

The root water uptake (RWU) process plays an important role in the terrestrial water cycle by connecting the soil environment to the atmosphere. Inaccurate representation of RWU functions (RWUFs) in land surface models (LSMs) has been one of the principal causes of error. Various RWUFs have been proposed for application in LSMs to simulate surface water–heat fluxes within diverse ecosystems. However, none have proven suitable for maize farmland ecosystems where vegetation differs from perennial plants, maize being a typical annual crop with significant human intervention. This study evaluated the performance of the common land model (CoLM) to reproduce energy and water-vapor fluxes during the growing season measured with an eddy covariance system in a maize farmland ecosystem of northeastern China. The simulation performance of the default CoLM differed with the interannual diversity of precipitation: it was relatively poor in the low rainfall year with underestimation of latent heat flux and overestimation of sensible heat flux, but was excellent in the high rainfall year. The effects of two different RWUFs were compared: RWUFZ, an empirical approach to represent a dynamic RWU efficiency by introducing two thresholds; and RWUFL, modifying the original RWUF with a linear function of soil water potential to one with an exponential function. The CoLM performance for simulating water and heat exchanges was significantly improved when the default RWUF was replaced with RWUFL and RWUFZ, respectively, using reasonable parameter assignments. The improvements of modified model performance demonstrated obvious interannual and intermonthly variations, which were more significant for RWUFZ than RWUFL. Thus, RWUFZ is recommended for use in the CoLM in northeast China’s maize farmland ecosystems and possibly for other similar agriculture ecosystems.