Selection of Remote Sensing Images And Evapotranspiration Models On Complex Land Surfaces for a Humid Karst Catchment

strongly regulated by vegetation coverage and available energy in subtropical humid karst catchments.

typically exhibit significant and rapid responses to precipitation events (Hartmann et 107 al., 2014a), due to various porosities (such as micropores, small fissures and fractures, 108 large fractures and conduits) (Bakalowicz, 2005) and low water retention in thin soil 109 (Liu et al., 2016). Moreover, severe rocky desertification has occurred in many fragile 110 karst ecosystems, mainly due to deforestation and agricultural intensification in recent 111 years (Yang et al., 2016). When rainfall is low, water stress and water shortages occur 112 frequently and rapidly, and there is an urgent need to quantify water fluxes to provide 113 reliable evidence for the protection and sustainable management of karst water 114 resources. 115 Therefore, this study aims to: 1) Construct and compare remote sensing models to 116 estimate ET at the regional level for karst landscapes; and 2) Characterize ET processes 117 and impact controls for karst ecosystems.    150 Sap flow for trees and shrubs was monitored using the TDP system to calculate 151 transpiration. Two thermocouple probes of the TDP system were implanted in the 152 sapwood of the tree trunk (Granier, 1985;Granier et al., 2000). Then, the temperature 153 difference (Td) was recorded by these two thermocouples. When transpiration happens 154 in the stem of trees or shrubs, the sap flow in the conduit will make the heated probe (1) 160 where Tdm is the mean value of the temperature difference between 23:00 pm and 3:00 161 am of the next morning (Cramer et al., 1999), and Sa is the area of trees' sapwood. 162 Ventilated chambers are able to measure actual ET for crops, shrub, grass, small 163 trees and plantations (Reicosky and Peters, 1977

Sap flow, ventilated chamber, and micro-lysimeters
where V is the flow rate of the pump (m 3 ·min -1 ) and AH is absolute humidity (g·m -3 ).

179
For ecosystems with installed TDP and ventilated chambers, ET can be estimated 180 as follows: where Tsum (kg· day -1 ) is tree transpiration, ETv is the ET measured by the ventilated 183 chamber in the understory and A (m 2 ) is the cover area of the trees.

184
In this study, Micro-lysimeters were used to measure soil evaporation.  where α is a dimensionless coefficient ranging from 1.08 ± 0.01 and 1.34 ± 0.05, with where n is the number of data sets, r is the linear correlation coefficient between 282 simulations and observations, Q t m is the modeled value, Q t o is observed data m Q is 283 average modelled value, and o Q is the average observed value. As shown in Fig.8, the surface parameters values exhibited distinct seasonality.

331
The average NDVI (Fig.8a-d)  As shown in Fig. 9, ET had a slight positive correlation with DEM, and R 2 was  In our study, daily ET for both HOUZHAI (Fig. 4) and SANCHAHE (Fig. 5)

374
Basins both exhibited seasonality. This finding was similar to previous studies in various climatic zones, including cold climates, temperate climates, tropical systems average daily temperatures were much lower than those regions. Consequently, annual

394
ET was also lower. Daily ET was highest during the growing seasons. In our study, 39 5 annual ET in forest, orchard and cropland was always higher than in fired patch and       index, ground surface conductance, net radiation and ground temperature, respectively.