Evapotranspiration In The Huangfuchuan Watershed Of Ordos Plateau, China, Estimated By A Scaling Model

Hydrologic cycle is the key component of the global climate system. Evapotranspiration is one the major processes driving this cycle. Two opposing fluxes, energy release and energy absorption, take place during evapotranspiration. In arid and semiarid regions evaporation is generally high and soil moisture content is low, especially in soils developed on loess substrates. While water is often a limiting factor of vegetation growth in such areas, vegetation cover substantially regulates the hydrologic cycle here. Estimating evapotranspiration at different scales and assessing the uncertainty of such estimations are ecological research problems of high importance. This paper presents a method for estimating and scaling the evapotranspiration.Our study area is the Huangfuchuan watershed of Ordos Plateau which has been intensively studied in the past. We developed the evapotranspiration estimation and scaling model of leaf-individual-community-landscape which is described below:1) Individual level. Transpiration models* for herbs and shrubs were developed taking Penman and Priestley-Taylor models as reference and considering differences in plant life forms. These models allowed us to scale up from the leaf to the individual plant level. The instantaneous transpiration model was developed using data on environmental factors, such as incident solar radiation, air temperature, wind speed at 2m (for herbs), and relative air humidity (for shrubs), and plant characteristics, such as leaf Stomatal conductance and leaf dry weight of individual plants. Daily transpiration was calculated by the integration procedure. The verification of the model resulted in the average relative error of 16.87% for herbs and 14.48% for shrubs, which is an acceptable level of uncertainty.2) Community level models* were developed in three different ways - the model based on individual transpiration models, the model based on direct relationship between leaf Stomatal conductance and community level evapotranspiration, and the model that used the original Penman-Monteith model with additional adjustements of some factors by soil water content. These model modifications resulted in improved accuracy.The model developed from individual transpiration models used the relationships between evapotranspiration and environmental factors such as incident solar radiation, air temperature, and relative air humidity, and incorporated plant leaf area index, individual transpiration of different plant life forms and leaf dry weight. The correlation of herb community model and shrub community models were respectively 0.8845 and 0.9238.The second model used relationships with meteorological factors such as incident solar radiation, air temperature, and relative air humidity, and plant characteristics such as leaf Stomatal conductance, percent cover, and leaf area index. The correlations achieved for herb community and shrub community were 0.9145 and 0.9449 respectively.The third model used the Penman-Monteith model that was corrected by soil water content. The correlations were 0.9254 for the herb community and 0.8563 for the shrub community. Sensitivity analysis produced±13% change in evapotranspiration when model parameters were modified within±20%. Therefore, the model is regarded stable and functional.The first evapotranspiration model is useful when individual evapotranspiration is calculated but leaf Stomatal conductance is not measured. The second model is more appropriate when leaf Stomatal conductance is measured. These two models are more applicable because they use less parameters compared to the Penman-Monteith model.3) Landscape level evapotranspiration was estimated using community scale models. Additionally, we used remote sensing to estimate evapotranspiration at this scale and contrast it with model upscaling.Community model upscaling considered changes in parameters' space and used Geographical Information Systems (GIS). GIS layers, such as topographic slope and aspect, maps of plant communities and vegetation cover were used to characterize landscape heterogeneity. We extrapolated to the entire Huangfuchuan watershed and estimated evapotranspiration in 2003 with the average relative error of 14.95%.Remote sensing approach was used to retrieve instantaneous evapotranspiration based on the estimation of land surface characteristics and fluxes from Landsat-5 TM images collected in 1996, 2003, and 2007, and using auxiliary environmental data from the same time periods. Daily evapotranspiration was estimated by scaling. Calculated daily evapotranspiration had an average relative error of 12.87% when results for 1996 and 2007 were verified using the FAO method. The average relative error for 2003 was 17.47%. This level of uncertainty is acceptable so we conclude the method is applicable. Because water was treated as bare land when using this method calculated evaporation from the water surface using the Penman model and merged this estimates into the daily evapotranspiration which resulted in improved accuracy. Finally, we compared these three images and found that vegetation cover decreased during the 11 year between 1996 and 2007. The average vegetation cover shrank from 17.33% in 1996 to 11.25% in 2007. Evapotranspiration also decrease uniformly during these time periods. No only the average vegetation cover decreased but farmland area has reduced because of the implementation of the policy promoting the return of cultivated lands to forests and grasslands. The comparison of vegetation maps from 1996 and 2007 showed that bare land area decreased, the area of newly planted shrubs greatly increased, farmland area decreased, and woodlands and water area increased insignificantly.The two methods of estimating landscape level evapotranspiration were compared. Both estimations were found plausible with the model upscaling approach being more biologically meaningful and slightly more accurate because of the use of manually interpreted vegetation maps of higher accuracy. Remote sensing approach, on the other hand, is useful in the absence of good quality biological information.Such important biological information is used to parameterize the aforementioned models and include leaf Stomatal conductance, leaf dry weight of individual plants, leaf area index and coverage. All models were coded in Visual FoxPro 6.0.