Scaling Effects In Land Surface Water And Heat Flux Modeling

The water and heat flux exchanges between soil and vegetation and atmosphere are the main drivers of aerodynamic and thermodynamic processes in land-atmosphere systems,determine the global energy balance and water cycle,and play a significant role in climate change adaptation,ecosystem protection and agricultural water management.At present,modeling of water and heat flux exchanges at larger scales between surface and atmosphere using remote sensing models and Land Surface Models(LSM-s)is suffering some uncertainties.Firstly,the parameters and parameterization methods applying to homogeneous land surfaces,may not be applicable anymore at larger scales and heterogeneous landscapes.Secondly,modeling by remote sensing models with low spatial resolutions and by "tile" based LSM-s with coarser grids lead to uncertainties in both depicting the real spatial distribution of water and heat flux exchanges and modeling the mean values of water and heat flux exchanges at pixel,subgrid and grid scales.Finally,validation of both remote sensing algorithms and of LSM-s against surface flux observations is far from straightforward due to the difference between the model grid/pixel size and the source area of surface flux observations.These uncertainties form the core scaling issue in modelling the water and heat flux exchanges between land surface and atmosphere.In this study,the scaling effect in water and heat flux modeling from the remote sensing high resolution(<100 m)scale to the LSM's coarser grid(50 km)scale was analyzed.This led to a better understanding of the determinants of the scale-dependence of water and heat fluxes in heterogeneous land-atmosphere systems.The findings described in this thesis provide some physical basis for further improving the parameterization and validation of LSMs.The key researches and conclusions in this thesis are as follows:Firstly,the influential variables on evapotranspiration under different soil and vegetation states had been identified though analyzing the sensitivity of the adopted remote sensing evapotranspiration model-ETMonitor.A method for a global sensitivity analysis of the evapotranspiration(ET)model considering both multiple parameters in a multi-dimensional phase-space and different distributions of soil water and vegetation properties was designed.The sensitivities of ET to LAI and SSM are higher under low vegetation coverage and low soil moisture conditions.Air temperature and relative humidity are influential under all LAI and SSM distributions.The sensitivity of ET to wind speed is higher with high vegetation coverage and high SSM.The sensitivity of ET to downward shortwave radiation is slightly reduced when SSM and LAI are both high.Secondly,for modeling by remote sensing models,the combined impact on evapotranspiration modelling of the low spatial resolution of remote sensing data in capturing surface heterogeneity in combination with models' nonlinearity was evaluated across spatial scales.To this end the spatial heterogeneities of vegetation,soil moisture and flux properties were analyzed by the method of imaginary multi-scale analysis based on two-dimensional discrete wavelet analysis,and the"aggregation error" on remote sensing ET estimates when using intermediate spatial resolution were analyzed by comparing the results obtained by step by step"parameter aggregation"(90 m,180 m.360 m,720 m to 1440 m)with "flux aggregation".The results of the experiments show that heterogeneity of vegetation,soil moisture and the evapotranspiration can occur at different scales,and that their heterogeneity distribution features can be different.This suggests that the most influential surface data at sufficiently high spatial resolution are required for evapotranspiration estimates of high accuracy.The experiments documented that the heterogeneity of ET and land surface conditions is smoother at smaller spatial scales.So that the impact of aggregating high spatial resolution(90 m)data to the kilometer pixel size for most of the pixels is not obvious in the study area.Larger aggregation errors only occur in areas along the Heihe river,near the border of cropland and desert,near the border of the residential areas and the cropland and in surroundings of the drip irrigation area in the southeast of the study area,where the soil and vegetation properties are more heterogeneous.Since the nonlinear degree of the relation between ET to the model inputs can be different in different conditions,the distribution of "aggregation error"is not always consistent with the distribution of surface heterogeneity.Thirdly,the influence of filtering out surface heterogeneity within the same subgrid and boundary layer meteorological heterogeneity within the same grid,by using homogeneous subgrid/grid assumption in LSMs has been analyzed.A numerical scenario analysis and a case study were carried out.In the numerical scenario analysis,the SSM and LAI scenarios and meteorological conditions were designed to evaluate the impact of each factor and of the interactions of SSM and LAI.The influence of meteorological conditions on the nonlinear relation of ET to SSM/LAI and the final aggregation error was evaluated.The results of the scenario analysis show that:the surface heterogeneity distribution type has an important influence on the aggregation error of estimated ET.Larger errors occur when SSM and LAI have positive skewed distribution(arid and semi-arid areas with low vegetation fractional cover).Results obtained with different meteorological conditions show that neglecting the heterogeneity of SSM leads to larger errors than when neglecting the heterogeneity of LAI.Neglecting the heterogeneity of SSM and LAI leads to underestimate ET under conditions with higher wind speed,higher air temperature and lower air relative humidity,and leads to overestimate ET under conditions with lower wind speed,lower air temperature and higher air relative humidity.The case study in the Heihe river basin in northwest of China,was done by using three frequently adopted grid sizes(10 km,25 km and 50 km)to analyze the influence of the surface and boundary layer meteorological heterogeneity on ET modelling by the simplization method adopted by LSMs.The results of the case study show that:for tile and grid scale simulations,neglecting the surface heterogeneity in each tile and the meteorological heterogeneity in each grid is not significant for evapotranspiration estimates at sub-regional scale,while it can induce large errors at subgrid and grid scales.Larger errors occur in areas with low soil moisture,i.e.the accuracy of LSMs based on land cover related parameterizations depends on different application objectives and study areas.The result show that in most of the grids,the impact of neglecting the surface(?)heterogeneity in each tile and the meteorological heterogeneity in each grid usually has same sign and their interactions have more obvious impacts when applying larger grid-sizes.