Performances of Land Surface Models in Sharp Vegetation Contrast Areas between Desert and Irrigation Oasis

The prediction of convection initiations (CIs) has always been a challenging problem in the prediction of severe convective weather and for the prevention of the related disasters. Boundary layer convergence lines (boundaries), which represent a narrow zone of low-level convergence, are a very important mesoscale CI lifting mechanism, and can sometimes trigger deep moist convections. The uneven heating due to sharp contrast in underlying surface characteristics is an important mode of boundary formation, such as vegetation heterogeneities. Accurate modeling of such kind of non-uniform distribution of the turbulent heat flux (THF, including the sensible heat flux (SH) and the latent heat flux (LH)) and the surface skin temperature (TSK) is important for predicting the formation and evolution of boundary and associated CI. One type of sharp vegetation contrast is across desert oasis border. The Hetao area in Inner Mongolia, China (Hetao) is a typical desert-oasis area. The oasis is the largest one-head irrigation oasis in Asia. Previous study shows that about 60 boundaries form along the desert-oasis border in June, July and August each year with 44% initiating convection, which may sometimes propagate downstream and cause disasters. This study presents a problem in using land surface models (LSMs) in WRF in modeling the uneven heating across desert-oasis border in Hetao area.

Normally, high LH in oasis and high SH and TSK in desert are a typical boundary layer characteristic (BLC) distribution in desert-oasis areas. We found that LSMs of WRF are the decisive factors of BLC simulations. Among the available various LSMs, Noah LSM is the most commonly used one with moderate complexity and computational efficiency, and has been used in most of the current real-world simulations of heterogeneous vegetation with decent performance. The simulations used a triple nested grid with the resolution of 3, 9, and 27 km from the inside to the outside respectively. Four cases were tested with half typical and half atypical in terms of SH and LH and TSK contrast between desert and oasis. Results show that in Hetao area, whatever the real situations are, the BLCs simulated by Noah LSM are always opposite to the typical distributions, namely low LH in oasis and low SH and TSK in desert, while those simulated by another commonly used LSM, PX-LSM, are always the same as the typical distribution. Why?

One possible reason is the oasis in Hetao is an artificial irrigation oasis, for which the two LSMs does not work well. Therefore, we examined the performances of LSMs of WRF for two natural oasis areas neighboring desert, including the Gurbantunggut Desert in Xinjiang, China (Xinjiang, the natural desert-oasis area in the arid and semi-arid region) and the Sahara Desert in Africa (Sahara, the natural desert-oasis area in the tropics). Two cases were tested for both Xinjiang used a double nested grid with the resolution of 3 and 15 km, and for Sahara with the resolution of 6 and 30 km, with one typical and one atypical BLC features. Our simulations show that the simulations of Noah LSM in Xinjiang and Sahara are obviously better than those in Hetao, where the simulated distributions and specific values of BLCs are both closer to the real situations and are obviously better than those of P-X LSM, but those of P-X LSM are still always the same as the typical distribution, no matter what the truth is. Consequently, the current Noah LSM does not fit Hetao area while P-X LSM only uses defaulted BLC regardless what the truth is.

Based on previous studies, the poor performance of Noah LSM in Hetao is most likely due to the high simulation error of soil moisture. However, compared with the observations, Noah LSM's simulations of soil moisture in Hetao oasis are quite accurate. Consequently, these experiments suggest Noah LSM's description of surface-atmosphere interaction may be closer to that in natural oasis, but significantly different from that in irrigation oasis. The overestimated SH in oasis in Hetao area indicates that parameters of Noah LSM may need to be tuned.

For Noah LSM, several previous studies have been done on its parameter sensitivity, which concluded that the parameter Czil describing the aerodynamics impedance, the parameter Sbeta describing the canopy effect and the parameter Rcmin describing the stomatal resistance, etc., are all likely to have a significant impact on the simulations of SH. Our ongoing work uses a parameter optimization software, PEST, to upgrade 16 of the parameters in Noah LSM, by repeatedly adjusting the parameters to gradually match the simulated THF with the observations. Therefrom, we can find parameters that are most sensitive to the simulations of THF and most different from the original values after optimization. As a result, we can reveal the causes for the huge simulation error of THF in artificial irrigation oasis and improve the performance of Noah LSM in areas with neighboring desert and artificial irrigation oasis.