The terrestrial water cycle is of critical importance to a wide array of earth systems. Evapotranspiration (ET) is an important component of the terrestrial water cycle, which controls the distribution of plant communities and the net primary production of ecosystems. Considerable efforts have been made by scientists to use remotely sensed data to estimate the spatial and temporal distribution of evaporation rates. Compared with other methods of estimation based on remotely sensed data, the Penman-Monteith equation has proven effective at both the point scale and the kilometer scale but the effectiveness of the surface conductance model in some specific areas such as arid regions still requires validation. Evaporation processes are complex in arid regions and the processes and patterns of ET under different vegetation types in arid regions are poorly researched. Therefore the Heihe River Basin was chosen as a study area because it is a typical arid area in China. The aim of this study was to evaluate the Penman-Monteith equation for estimating daily evaporation fluxes in arid regions using MODIS LAI data and meteorological data. A simple biophysical model using leaf area index (LAI) from remotely sensed data and the Penman-Monteith equation was used to calculate the daily ET of grassland and farmland at various sites in the middle and upper reaches of the Heihe River. The Penman-Monteith equation is a biophysically based equation for calculating land surface evaporation and has been used extensively. The model for computing the surface conductance contains six parameters requiring local calibration. The shuffled complex evolution (SCE) algorithm was used to optimize the parameters in the surface conductance model for each site. The Nash-Sutcliffe efficiency (NSE) was selected as the optimized objective function. Optimal parameters in the surface conductance model were obtained by minimizing the cost function NSE with the SCE algorithm. Excellent agreement was obtained between the measured mean daily evaporation rates and those calculated using remotely sensed LAI data and the Penman-Monteith equation. The NSE at the A'rou (grassland) and Yingke (farmland) sites were 0.84 and 0.85, respectively and the root mean square error (RMSE) were 1.25 and 1.66 MJ m^-2 d^-1, respectively. With regard to grassland and farmland ecosystems in this study, the transpiration was much stronger than the soil evaporation during the growing seasons, however, the opposite situation was found during the non-growing seasons. From this study it can be concluded that the Penman-Monteith equation combined with LAI data can provide reliable estimates of ET at daily time scales in different ecosystems of arid and cold regions. However, there are still some uncertainties in ET estimations of the ecosystems in this study area, especially the cropland at the Yingke station. These biases are possibly because the model cannot simulate the impact of the extreme soil moisture changes caused by irrigation on ET. The soil evaporation factor is a constant in this study, whereas in reality it is a variable that depends on the moisture status of the soil near the surface. Thus, the effect of soil moisture variation on surface conductance was considered, and it improved the precision of the equation in the simulation of farmland ET.