Soil moisture simulation by JULES in New Zealand: verification and sensitivity tests

Hourly observations of soil moisture volumetric content (at depths of 0-0.4m) have been conducted for a few years at 55 climate stations with diverse soil types and annual rainfall in New Zealand. Thirty two of these stations also have surface meteorological observations. In this study, these observations were used for the first time to validate a land surface model (LSM) for soil moisture simulation, the Joint UK Land Environment Simulator (JULES). At the 32 stations, the yearly mean absolute differences (MADs) of the simulated soil moisture volumetric ratio were about 5% or smaller, except at four South Island stations, where larger MADs of 8-12% were found. Further analysis of rainfall and soil moisture observations at these four stations showed that the observed soil moisture includes a contribution from lateral soil water flow, which is not considered in the one-dimensional version of JULES used in this study. For the remaining 28 stations, the relatively small yearly MADs (approximate to 5%) were mainly due to biases. To understand the major factors contributing to the errors/biases, three sensitivity tests were conducted. These tests showed that errors in the soil texture data were the most likely contributor to the errors in simulated soil moisture. This study demonstrates that for a better simulation of soil moisture in New Zealand, especially in the lower South Island, a LSM needs to include lateral soil moisture/water flow, and to use more accurate prescriptions of the hydraulic and thermal properties of the soil.