Comment on gmd-2022-74

Abstract

We assess the land surface model JSBACHv4, which was recently developed at the Max Planck Institute for Meteteorology as part of the effort to build the new Earth System model ICON-ESM. We assess JSBACHv4 in simulations coupled with ICON-A, the atmosphere model of ICON-ESM, hosting JSBACHv4 as land component to provide the surface bounddary conditions. The assessment is based on a comparison of simulated albedo, Land Surface Temperature (LST), Leaf Area Index (LAI), Terrestrial Water Storage (TWS), Fraction of Absorbed Photosynthetic Active Radiation (FAPAR), Net Primary Production (NPP), and Water-Use-Efficiency (WUE) with corresponding observational data. JSBACHv4 is the successor of JSBACHv3, therefore, another purpose of this study is to document how this step in model development has changed model biases. This is achieved by also assessing in parallel results of coupled land-atmosphere simulations with the preceding model ECHAM6 hosting JSBACHv3. Large albedo biases appear in both models over ice sheets and in central Asia. The temperate to boreal warm bias observed in simulations with JSBACHv3 largely remained in JSBACHv4, despite the very good agreement with observed LST in the global mean. For the assessment of changes in land water storage, a novel procedure is suggested to compare the gravitational data from the GRACE satellites to simulated TWS. It turns out that the agreement of changes in the seasonal cycle of TWS is sensitive to the representation of precipitation in the atmosphere model. The LAI is generally too high which is partly caused by too high soil moisture but also by the parameterization of the phenology itself. The pattern of WUE is for both models largely as observed. In India WUE is too high probably because JSBACH does not incorporate irrigation in our simulations. WUE differences between the two models can be traced back to differences in precipitation patterns in the two coupled land-atmosphere simulations. For both models, most NPP biases can be associated with biases in water stress, LAI and FAPAR. Particularly the NPP bias of the Eurasian steppes has switched from positive in JSBACHv3 to negative in JSBACHv4. This difference is mainly caused by weaker precipitation and FAPAR of ICON-A+JSBACHv4 in July, which is most probably caused by a feedback loop between too little soil moisture, evaporation and clouds. While the size and patterns of biases in albedo and LST are largely similar between the two model versions, they are less well correlated for precipitation and vegetation related variables like FAPAR. Overall, we see very good perspectives for further improving JSBACHv4, because there are sound possibilities to mitigate most of the biases without interdependencies or technical problems.