Biases in Model-Simulated Surface Energy Fluxes During the Indian Monsoon Onset Period

Abstract

We use eddy-covariance measurements over a semi-natural grassland in the central Indo-Gangetic Basin to investigate biases in energy fluxes simulated by the Noah land-surface model for two monsoon onset periods: one with rain (2016) and one completely dry (2017). In the preliminary run with default parameters, the offline Noah LSM overestimates the midday (1000–1400 local time) sensible heat flux (H) by 279% (in 2016) and 108% (in 2017) and underestimates the midday latent heat flux (LE) by 56% (in 2016) and 67% (in 2017). These discrepancies in simulated energy fluxes propagate to and are amplified in coupled Weather Research and Forecasting model simulations, as seen from the High Asia Reanalysis dataset. One-dimensional Noah simulations with modified site-specific vegetation parameters not only improve the partitioning of the energy fluxes (Bowen ratio of 0.9 in modified run versus 3.1 in the default run), but also reduce the overestimation of the model-simulated soil and skin temperature. Thus, use of ambient site parameters in future studies is warranted to reduce uncertainties in short-term and long-term simulations over this region. Finally, we examine how biases in the model simulations can be attributed to lack of closure in the measured surface energy budget. The bias is smallest when the sensible heat flux post-closure method is used (5.2 $\hbox {W } \hbox {m}^{-2}$ for H and 16 $\hbox {W } \hbox {m}^{-2}$ for LE in 2016; 0.17 $\hbox {W } \hbox {m}^{-2}$ for H and 2.8 $\hbox {W } \hbox {m}^{-2}$ for LE in 2017), showing the importance of taking into account the surface energy imbalance at eddy-covariance sites when evaluating land-surface models.