Estimates of Daily Evapotranspiration in the Source Region of the Yellow River Combining Visible/Near-Infrared and Microwave Remote Sensing

Article Estimates of Daily Evapotranspiration in the Source Region of the Yellow River Combining Visible/Near-Infrared and Microwave Remote Sensing

Rong Liu 1, Jun Wen 2,* , Xin Wang 1, Zuoliang Wang 1, Yu Liu 1 and Ming Zhang 3

1 Key Laboratory of Land Surface Process and Climate Change in Cold and Arid Regions, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China; [email protected] (R.L.); [email protected] (X.W.); [email protected] (Z.W.); [email protected] (Y.L.) 2 College of Atmospheric Sciences, Chengdu University of Information Technology, Chengdu 610225, China 3 Faculty of Geomatics, East China University of Technology, Nanchang 330013, China; [email protected] * Correspondence: [email protected]

Abstract: The spatial variation of surface net radiation, soil heat flux, sensible heat flux, and latent heat flux at different times of the day over the northern Tibetan Plateau were estimated using the Surface Energy Balance System algorithm, data from the FY-2G geostationary meteorological satellite, and microwave data from the FY-3C polar-orbiting meteorological satellite. In addition, the evaporative fraction was analyzed, and the total evapotranspiration (ET) was obtained by the effective evaporative fraction to avoid the error from accumulation. The hourly change of latent heat flux presented a sound unimodal diurnal variation. The results showed the regional ET ranged between 2.0 and 4.0 mm over the Source Region of the Yellow River. The conditional expectations of surface energy components during the experimental period of the study area were statistically analyzed, and the correspondence between different surface temperatures and the effective energy distribution was examined. The effective energy distribution of the surface changed significantly with the increase in temperature; in particular, when the surface temperature exceeded 290 K, the Citation: Liu, R.; Wen, J.; Wang, X.; Wang, Z.; Liu, Y.; Zhang, M. Estimates effective energy was mainly used for surface ET. The aim of this study was to avoid the use of surface of Daily Evapotranspiration in the meteorological observations that are not readily available over large areas, and the findings lay a Source Region of the Yellow River foundation for the commercialization of land surface evapotranspiration. Combining Visible/Near-Infrared and Microwave Remote Sensing. Keywords: visible and near-infrared; microwave; evapotranspiration; FengYun meteorological Remote Sens. 2021, 13, 53. satellite; evaporation fraction; Source Region of the Yellow River https://dx.doi.org/10.3390/ rs13010053

Received: 19 November 2020 1. Introduction Accepted: 24 December 2020 The variation of regional evapotranspiration is of signiﬁcance for the land surface Published: 25 December 2020 process and the energy exchanges in the hydrosphere, atmosphere, and biosphere [1–3].

Publisher’s Note: MDPI stays neu- Due to the development of observation methods and understanding of the turbulence tral with regard to jurisdictional claims process in the atmospheric boundary layer, the method of calculating land surface evap- in published maps and institutional otranspiration has gradually developed from the traditional spot-measurements-based afﬁliations. method to regional evapotranspiration estimation based on the technology of large-scale satellite remote sensing [4]. Since the mid-1970s, numerous scholars have undertaken a large amount of research on the accuracy of evapotranspiration retrieved by remote sensing [5,6]. According to Copyright: © 2020 by the authors. Li- different theories, remote sensing evapotranspiration models can be roughly divided censee MDPI, Basel, Switzerland. This into three categories. First category is based on the Penman-Monteith algorithm with an article is an open access article distributed improvement of its parameters, using remote sensing data and ground observation data to under the terms and conditions of the estimate the actual evapotranspiration [7]. The second type is a statistical empirical model, Creative Commons Attribution (CC BY) which attempts to establish a correlation between evapotranspiration and vegetation license (https://creativecommons.org/ index (VI) or surface-air temperature difference [8,9], or derives the two-dimensional licenses/by/4.0/).

Remote Sens. 2021, 13, 53. https://dx.doi.org/10.3390/rs13010053 https://www.mdpi.com/journal/remotesensing Remote Sens. 2021, 13, 53 2 of 15

scatter diagram of land surface temperature and vegetation index (or albedo) with specific trapezoidal or triangular characteristics [10,11]. The third category comprises the single and dual-source models based on the principle of energy balance, among which the Surface Energy Balance Algorithm for Land (SEBAL) model [12], Simplified Surface Energy Balance Index (S-SEBI) model [13], Surface Energy Balance System (SEBS) model [14], and Mapping Evapotranspiration with Internalized Calibration (METRIC) model [15] are the typical cases of a single-source model while Shuttleworth-Wallace model [16] and N95 model [17] are dual-source models. The Qinghai-Tibet Plateau has an average altitude of more than 4000 m, accounting for a quarter of China’s total land area, and is an important source of water and heat for the East Asian monsoon circulation. Its huge topography and unique geographical location determine that its surface evapotranspiration not only plays an important role in the formation and evolution of the Asian monsoon system but is also closely related to changes in regional surface water resources over the Qinghai-Tibet Plateau. The temperature data of 50 meteorological stations on the Qinghai-Tibet Plateau from the 1950s to the 1990s showed that both the temperature and precipitation of the Qinghai-Tibet Plateau have displayed a significant upward trend [18]. Therefore, evapotranspiration, which is the main process of water loss, is closely related to the changes in water resources over the Qinghai-Tibet Plateau. However, the visible and near-infrared observation of sensors fails in cloud-covered conditions. Hence, for evapotranspiration estimates, most of the models noted above based on satellite remote sensing avoid cloud cover and use observation data provided by polar-orbiting satellites to estimate the instantaneous latent heat flux as the sensor passes. In this process, errors cannot be avoided and the diurnal variation process of the evapotranspiration cannot be obtained. Alternatively, geostationary meteorological satellites can provide high-frequency observation data once per hour, allowing the real- time monitoring of the regional surface state, and the microwave radiometer carried by these satellites can penetrate clouds and fog to retrieve the surface flux under cloud-cover conditions. However, the regional evapotranspiration combined with two kinds of data is not commonly studied. The Source Region of the Yellow River (SRYR) is located in the northeast of the Tibetan Plateau. As an important runoff of the Yellow River, the SRYR area supplies about 45.0% of the water in the upper reaches of the Yellow River, and is thus also called “the Reservoir of the Yellow River” [18]. In this study, the spatial distributions of the hourly surface net radiation, soil heat flux, latent heat flux, and sensible heat flux in the SRYR area were estimated using observations from the FY-2G geostationary meteorological satellite and SEBS model in real-time. During afternoons, when the plateau convective cloud develops vigorously, MicroWave Radiation Imager (MWRI) data from the FY-3C polar-orbiting meteorological satellite was used to obtain the surface energy fluxes. The results were verified with the observation data from the Source Region of the Yellow River Alpine Climate Station. The statistical results and errors were also analyzed, and are discussed in the current paper. This paper intends to provide more reliable results for evapotranspiration in typical summer weather in the SRYR area. The paper is organized as follows: In Section2 the methodology is briefly described. The study area and observation data are presented in Section3. Section4 presents results and verification. Finally, a discussion and conclusions of this study are presented in Sections5 and6, respectively.

2. Methodology 2.1. Net Radiation 2.1.1. Net Radiation under Clear Sky Conditions Cloud classification and cloud total amount of products obtained from the FY-2G geostationary meteorological satellite can determine whether the grid point is covered with clouds. When the cloud classification product is defined as the clear sky and the cloud total Remote Sens. 2021, 13, 53 3 of 15

amount is less than 10%, it is considered to be a clear sky condition (no clouds) and the near-surface net radiation Rn is calculated according to:

Rn = (1 − α)Q + L ↓ −L ↑ (1)

where, α is the surface albedo, determined by MODIS observation data; Q is the total solar radiation, including direct radiation, scattered radiation, and reﬂected radiation [19]. L↓ is the downward long-wave radiation, which can be obtained by:

4 L ↓= εa · σ · Ta (2)

where, εa is the air effective emissivity, obtained from actual water vapor pressure and air −8 −2 −4 temperature [20]; σ is the Stefan-Boltzman constant, which is 5.669 × 10 Wm k ; Ta is the measured air temperature. L↑ is the upward long-wave radiation, which can be calculated from the formula:

4 L ↑= ε · σ · Ts (3)

where, ε is the land surface emissivity, obtained by the methods of Chen et al. [19]; TS is the surface skin temperature, which can be obtained using the Black Body Temperature (TBB) data observed by the Visible Infrared Spin Scan Radiometer (VISSR) on FY-2G and the MODerate resolution atmospheric TRANsmission model (MODTRAN). Speciﬁcally, the band response functions of MODTRAN and VISSR are adopted to calculate the atmospheric thermal infrared transmittance and atmospheric thermal infrared radiance. Then, brightness temperature from VISSR thermal infrared band-1 and band-2 are chosen in the surface radiation transfer model. Finally, the results can be derived using an iterative algorithm.

2.1.2. Net Radiation under Cloud Cover The SRYR is located in the northeast of the Qinghai-Tibet Plateau. The development of convective clouds over the plateau in the summer afternoon limits visible and near- infrared remote sensing. Therefore, a simpliﬁed parameterization scheme of surface radiation ﬂux [21] was adopted under cloud-cover conditions. This scheme assumes that the long-wave radiations are approximately equal under cloud cover. (1) Atmospheric attenuation The attenuation effect of the atmosphere on solar radiation can be presented by the transmittance τ: Q τ = E (4) Q0

where, QE is the solar radiation that reaches the surface after atmospheric attenuation; Q0 is the total solar radiation at the top of the atmosphere. (2) The energy received by the satellite sensor The energy received by the satellite sensor can be divided into two parts: one is the energy of solar radiation that has not been attenuated by the atmosphere, and the other is the energy that reaches the surface after atmospheric attenuation and is reﬂected [21].

αTOAQ0 = Q0ταsτ + (1 − τ)Q0 (5)

where, αTOA is the albedo of the top cloud, αTOAQ0 is the energy received by the sensor; αs is the surface albedo; Q0 τ αs τ is the energy passing to the surface after atmospheric attenuation, and reﬂected and attenuated by the surface; (1 − τ)Q0 is the energy reﬂected at the cloud top. Using Equation (5), the formula of atmospheric transmittance can be given by the following: p 1 − 1 − 4α (1 − α ) τ = S TOA (6) 2αS Remote Sens. 2021, 13, 53 4 of 15

where αTOA and αs are taken from the observations of the geostationary and polar-orbiting satellites, respectively. The hourly temporal scale transmittance can be derived from Equation (6), and the net radiation calculation formula under the cloudy condition is:

QE = (1 − α)Q0τ (7)

It should be noted that the atmospheric transmittance obtained from the above formula is not the complete atmospheric transmittance of solar radiation, but the total equivalent extinction effect of the earth-atmosphere system on the radiation transfer path from outside the atmosphere to the ground surface. It represents the interaction between absorption and scattering effects. For satellite sensor observations that are missing or with obviously wrong values, we established the relationship between solar short-wave radiation and solar radiation before the atmospheric effect to calculate the atmospheric transmittance.

2.2. Soil Heat Flux The formula for soil heat ﬂux is given by:

Go = Rn[Γc + (1 − fv)(Γs − Γc)] (8)

where, Γc and Γs represent the ratios between G0 and Rn with 100% vegetation cover NDVI−NDVImin and bare soil area. Vegetation coverage follows fv = , where NDVI is NDVImax−NDVImin min the normalized difference vegetation index of the bare soil surface, and NDVImax is the normalized difference vegetation index when the area is fully covered with vegetation [21].

2.3. Calculation Method of Sensible Heat Flux An iterative method is adopted to calculate the sensible heat ﬂux and the speciﬁc calculation formula is as follows:

u∗ Z − do Z − do Zom u = [ln( ) − ψm( ) + ψm( )] (9) k Zom L L

H Z − do Z − do Zoh θo − θa = [ln( ) − ψh( ) + ψh( )] (10) ku∗ρCp Zoh L L 3 ρCpu∗θv L = − (11) kgH

where Z is the height of the reference surface, u is the wind speed, u∗ is the friction wind speed, do is the zero-plane displacement height, and Zom and Zoh denote momentum and thermal roughness lengths, respectively. ψm and ψh are the stability correction functions for dynamic and thermodynamic transfer, respectively; θo and θa are the potential temperature of the surface air and the potential temperature of the reference height, respectively; H is the sensible heat ﬂux; L is the Obukhov length; k is the von Kármán constant; ρ is the air density, Cp is the gas constant; g is the acceleration of gravity; and θv is the virtual potential temperature near the surface. More parameters can be seen in Su [14]. The Leaf Area Index (LAI) needed for the calculation of Zoh and Zom can be derived from the method of Chen et al. [19]. In this research, the 37 GHz vertically polarized brightness temperature from the MWRI on the FY-3C polar-orbiting meteorological satellite was used to retrieve the surface temperature under cloud-cover conditions at 16:00 [22]. Remote Sens. 2020, 12, x 5 of 15

In this research, the 37 GHz vertically polarized brightness temperature from the MWRI on the FY-3C polar-orbiting meteorological satellite was used to retrieve the sur- Remote Sens. 2021, 13, 53 5 of 15 face temperature under cloud-cover conditions at 16:00 [22].

2.4. Evaporation Fraction 2.4. EvaporationThe evaporation Fraction fraction is the ratio of actual ET to effective energy (the difference betweenThe evaporationnet radiation fractionand soil isheat the flux) ratio and of actual can be ET presented to effective as energy (the difference between net radiation and soil heat flux) and canLE be presented as EF = (12) RGLE− EF = n 0 (12) Rn − G0 where EF is the evaporation fraction, LE is the latent heat flux, Rn is the net radiation flux, whereand G EFis theis the soil evaporation heat flux. fraction, LE is the latent heat flux, Rn is the net radiation flux, and GTheis theSEBS soil model heat flux. defines the limit cases as deficient and sufficient [23]. Since evapo- rationThe fraction SEBS modelis stable defines during the the limit whole cases day, as deficientif we know and the sufficient net radiation [23]. Since and evapora-soil heat tionflux fractionat a certain is stable time, duringthen the the latent whole heat day, flux if at we that know time the can net be radiation calculated and using soil the heat latent flux atheat a certain of vaporization time, then (2.49 the × latent 106 Wm heat−2mm flux−1). at that time can be calculated using the latent heat of vaporization (2.49 × 106 Wm−2mm−1). 2.5. Validation 2.5. Validation Ground observations from the regular meteorological station and eddy covariance systemGround from observationsthe SRYR Alpine from Climate the regular Station meteorological (102°08′E, 33°53 station′N, 3443 and eddym) were covariance used to system from the SRYR Alpine Climate Station (102◦080 E, 33◦530 N, 3443 m) were used to validate the instantaneous heat fluxes that were estimated from remote sensing retrievals validate the instantaneous heat fluxes that were estimated from remote sensing retrievals from 1–30 August 2019. The statistical analysis using the root mean square error is also from 1–30 August 2019. The statistical analysis using the root mean square error is also presented in Section 4.2. presented in Section 4.2. Since the soil heat flux observation was at 10 cm, we used the Averaging Soil Ther- Since the soil heat flux observation was at 10 cm, we used the Averaging Soil Thermo- mocouple Probe (TCAV) method to elicit the surface soil heat flux through soil heat stor- couple Probe (TCAV) method to elicit the surface soil heat flux through soil heat storage, age, which was obtained from the average soil temperature [24]. The daily average soil which was obtained from the average soil temperature [24]. The daily average soil volu- volumetric water content of 0–10 cm at the observation site was 0.39, and the average metric water content of 0–10 cm at the observation site was 0.39, and the average porosity porosity of shallow soil was 0.50. of shallow soil was 0.50.

3.3. Study Area and Observation DataData TheThe sourcesource regionregion of of the the Yellow Yellow River River is is located located in in the the northeastern northeastern part part of theof the Qinghai- Qing- Tibethai-Tibet Plateau Plateau (95.5 (95.5°E–103.5°E,◦ E–103.5◦ E, 31.5 31.5°N–36◦ N–36.5◦.5°N),N), with with an an area area of of about about 1.22 1.22× ×10 1055km km22,, accountingaccounting for 16% of thethe totaltotal areaarea ofof thethe YellowYellow RiverRiver Basin.Basin. ThisThis area,area, mostlymostly withwith anan altitudealtitude ofof aboveabove 40004000 m,m, isis mainlymainly coveredcovered withwith alpinealpine meadowsmeadows (Figure(Figure1 1).). The The source source regionregion ofof thethe YellowYellow RiverRiver isis locatedlocated inin thethe “arid-semi-arid”“arid-semi-arid” transitiontransition zonezone betweenbetween thethe Qinghai-TibetQinghai-Tibet PlateauPlateau and and the the Loess Loess Plateau. Plateau. It hasIt has a typical a typical continental continental alpine alpine climate climate with anwith average an average annual annual temperature temperature of 2.9 ◦ofC. 2.9 Precipitation °C. Precipitation is mainly is mainly concentrated concentrated in the period in the fromperiod May from to May September, to September, accounting accounting for more for than more 80% than of the 80% total, of the which total, is which 560 mm is per560 yearmm (1956–2016).per year (1956–2016). The annual The evaporation annual evaporation is 920 mm, is gradually920 mm, gradually decreasing decreasing from southeast from tosoutheast northwest to northwest [21]. [21].

FigureFigure 1.1. Location and land useuse ofof thethe SourceSource RegionRegion ofof thethe YellowYellow RiverRiver (SRYR).(SRYR).

The regular meteorological and eddy covariance observations from the SRYR Alpine Climate Stations (102◦080 E, 33◦530 N, 3443 m) were used from 1 August 2019. The underlying soil was alpine meadow soil, mainly covered by meadows with a height of 10 cm. The automatic meteorological station was set with three ﬂoors (1, 5, and 10 m), observing Remote Sens. 2021, 13, 53 6 of 15

air temperature, air pressure, relative air humidity, surface temperature, wind speed at 10 m, precipitation, and sunshine hours. An eddy covariance system was erected at 3.02 m, with a frequency of 10 Hz, observing short-wave radiation flux and long-wave radiation flux (upward and downward), sensible heat flux, latent heat flux, soil heat flux (two layers: −10 cm and −20 cm). The system was set to output average flux every 10 min, and the results were finally averaged to obtain the hourly temporal scale data. FY-2G and FY-3C are the geostationary meteorological and the polar-orbiting meteorological satellites, respectively, developed and operated by China. The Visible Infrared Spin Scan Radiometer carried by the FY-2G is the main earth observing instrument on the FY series satellites of the China Earth Observing System. It provides five channels of data from 1 July 2015. The resolution is 1.25 km for the visible channel, and 5 km for infrared and water vapor channels. The MicroWave Radiation Imager (MWRI) on FY-3C can be used for monitoring and research of soil moisture, snow cover, and ocean salinity. The sensor contains five frequencies-10.65, 18.7, 23.8, 36.5, and 89 GHz-and the corresponding spatial resolutions of sub-satellite points are 51 km, 85 km, 30 × 50 km, 27 × 45 km, 18 × 30 km, and 9 × 15 km. The data of the FY series satellites used in the study were all downloaded from the Fengyun Satellite Data Center (http://satellite.nsmc.org.cn/portalsite/default.aspx), including the 1-h average thermal infrared radiation brightness temperature and visible wide-band albedo, and MWRI 36.5 GHz microwave radiation brightness temperature. For specific processing steps, refer to the relevant literature [25]. In addition, the remote sensing data used in this study also included surface albedo, NDVI, LAI, and DEM. Surface albedo, NDVI, and LAI were obtained from MODIS (http://modis.gsfc.nasa.gov/). NDVI was adopted by the maximum value composite from an 8-day daily MODIS-NDVI, which eliminated the adverse effects of clouds, atmosphere, and satellite zenith angles. The DEM was obtained from the Shuttle Radar Topography Mission (SRTM) (http://srtm.csi.cgiar.org/), and its spatial resolution was 90 m. After the data or products were resampled to the nominal projection format of FY-2G, the final results were converted to the 0.078◦ longitude and latitude coordinate system to complete the drawing and statistical analysis. Figure2 shows the surface albedo, NDVI, and LAI in the SRYR retrieved by satellite remote sensing during the study period. The distribution figure shows that the surface albedo, NDVI, and LAI in the study area were highly consistent with its land use. Since the study area was dominated by alpine meadow grasslands, the surface albedo of the entire area was generally below 0.25. The minimum albedo (<0.10) occurred mainly in the west of the study area, indicating that there were large lakes, whereas the maximum value (>0.25) was in the northwest, indicating there was snow or permafrost tundra. The NDVI value in the area was between 0.45 and 0.85. Due to the distribution of lakes, certain areas also showed low values between 0.10 and 0.40. The value of LAI was between 0.9 and 2.1, indicating that although the distribution of vegetation presented certain spatial differences, there was no fundamental difference in general, and LAI had a more obvious spatial variation range than NDVI. The altitude in the region was around 3200–5200 m, and the location of the observation site was relatively flat. It should be noted that this research did not examine the surface ET of snow and lakes. Remote Sens. 2021, 13, 53 7 of 15

Remote Sens. 2020, 12, x 7 of 15

a b

c d

Figure 2. Surface parameters of the SRYR (a) Surface albedo (b) NDVI (c) LAI (d) DEM. Figure 2. Surface parameters of the SRYR (a) Surface albedo (b) NDVI (c) LAI (d) DEM. 4. Results and Verification 4.1.4. Results Results ofand Energy Verification Flux 4.1. BecauseResults of the Energy underlying Flux surface of the study area was relatively flat and due to the footprintBecause effect, the the underlying surface observation surface of the was study of spatial area was representativeness relatively flat and from due tens to ofthe metersfootprint to several effect, the kilometers. surface observation That means was the energyof spatial flux representativeness from the surface observation from tens of can me- beters approximately to several kilometers. taken as That the surface means energythe energy flux flux of the from entire the studysurface area. observation To verify can the be satelliteapproximately remote sensingtaken as algorithm the surface and energy the estimated flux of the results, entire thestudy field area. observations To verify the were sat- comparedellite remote with sensing the estimated algorithm results. and the Figure estima3 showsted results, the daily the variation field observations of the near-surface were com- energypared fluxwith inthe the estimated SRYR retrieved results. Figure by satellite 3 shows remote the sensingdaily variation on 18 August of the near-surface 2019. The figureenergy shows flux in the the hourly SRYR net retrieved radiation, by satell soil heatite remote flux, sensible sensing heaton August flux, and 18th, latent 2019. heat The fluxfigure from shows 11:00 the to hourly 16:00 Beijing net radiation, time.The soil solarheat flux, radiation sensible absorbed heat flux by, and the latent ground heat after flux sunrisefrom 11:00 was to greater 16:00 thanBeijing the time. expended The solar radiation. radiation At absorbed 11:00, the by net the radiation ground after was aboutsunrise −2 −2 400.0was greater Wm , than the soilthe expended heat flux wasradiation. about At 80.0 11:00, Wm the ,net the radiation sensible was heat about flux was400.0 about Wm−2, −2 −2 120.0the soil Wm heat, andflux the was latent about heat 80.0 flux Wm value−2, the was sensible about heat 200.0 flux Wm was; about At midday, 120.0 whenWm−2, the and solarthe latent radiation heat strengthened, flux value was the netabout radiation 200.0 Wm continued−2; At midday, to increase when at 12:00 the andsolar 13:00, radiation and thestrengthened, net radiation, the soil net heat radiation flux, andcontinued sensible to heat increase flux allat 12:00 reached and the 13:00, daily and maximum the net radiat −2 −2 −2 12:00ation, and soil 13:00, heat which flux, wereand sensible about 650.0 heat Wm flux all, 100.0 reached Wm the, and daily 150.0 maximum Wm , respectively.at 12:00 and −2 The13:00, latent which heat were flux about was about650.0 Wm 300.0−2,Wm 100.0 Wm. In− the2, and afternoon, 150.0 Wm when−2, respectively. the solar radiation The latent weakened,heat flux was the netabout radiation 300.0 Wm continued−2. In the to afternoon, decrease at when 14:00 the and solar 15:00, radiation to between weakened, 500.0 and the −2 650.0net radiation Wm , although continued the to studydecrease area at showed 14:00 and relatively 15:00, to lower between values 500.0 at and 15:00. 650.0 The Wm soil−2, −2 −2 heatalthough flux and the sensible study area heat showed flux reached relatively 60.0 Wm lower andvalues 100.0 at Wm15:00. Theat 15:00, soil heat respectively. flux and Thesensible latent heat heat flux flux reached reached 60.0 the dailyW.m− maximum2 and 100.0 atW.m 14:00−2 at and 15:00, decreased respectively. at 15:00, The which latent −2 −2 wereheat 320.0flux reached Wm and the 250.0daily Wmmaximum, respectively. at 14:00 and As decreased the solar altitude, at 15:00, radiation which were and 320.0 air saturationW.m−2 and deficit 250.0 weakenedW.m−2, respectively. in the afternoon, As the thesolar energy altitude, fluxes radiation decreased andat air 16:00. saturation The valuesdeficit of weakened the energy in fluxes the afternoon, at 16:00 were the energy approximately fluxes decreased similar to at those 16:00. at The 12:00. values of the energy fluxes at 16:00 were approximately similar to those at 12:00.

Remote Sens. 2021, 13, 53 8 of 15

Remote Sens. 2020, 12, x 8 of 15

11:00

12:00

13:00

Figure 3. Cont.

Remote Sens. 2021, 13, 53 9 of 15

Remote Sens. 2020, 12, x 9 of 15

14:00

15:00

16:00

Figure 3. HourlyFigure surface 3. energy Hourly fluxes surface over theenergy study fluxes area onover 18 Augustthe study 2019 area (11:00 on 18 to 16:00August Beijing 2019 time). (11:00 to 16:00 Bei- jing time). From Figure3, it can be seen that as the plateau convective cloud started to develop at 15:00, the near-infraredFrom Figure band of 3, theit can geostationary be seen that satellite as the plateau was affected convective by cloud cloud cover, started to develop thus, the flux couldat 15:00, not the be near-infrared retrieved in band most of the the geosta area.tionary However, satellite the MWRI was affected on the by cloud cover, polar-orbitingthus, satellite the passedflux could through not be the retrieved SRYR in at most around of the 15:40. area. Figure However,4 shows the MWRI the on the polar- latent heat fluxorbiting retrieved satellite from the passed FY-3C through MWRI andthe SRYR FY-2G at visible/near-infrared around 15:40. Figure data. 4 shows The the latent heat missing data overflux theretrieved cloud coverfrom the area FY-3C from FY-2GMWRI wasand replacedFY-2G visible/near-infrared with data from FY-3C. data. The missing As shown in thedata figure, over in the areas cloud with cover no area cloud from cover, FY-2G the spatialwas replaced distribution with data of high from and FY-3C. As shown low values retrievedin the figure, by the in two areas methods with no were cloud basically cover, th consistent.e spatial distribution Cloudy areas of high were and low values mainly concentratedretrieved in the by norththe two and methods south of were the study basically area, consistent. and the latent Cloudy heat areas flux was were mainly concentrated in the north and south of the study area, and the latent heat flux was about 200.0

Remote Sens. 2021, 13, 53 10 of 15

Remote Sens. 2020, 12, x 10 of 15

−2 aboutWm−2 200.0(Figure Wm 4b). However,(Figure4b). the However, latent heat the flux latent obtained heat ﬂuxfrom obtained the visible from and the near-infra- visible andred near-infraredband displayed band sharp displayed jumps in sharp the cloudy jumps area in the and cloudy its edge(Figure area and its 4a), edge indicating(Figure4 thata) , indicatingthe results thatof the the geostationary results of the satellite geostationary obviously satellite affected obviously the retrieval affected results. the The retrieval latent results.heat flux The retrieved latent from heat ﬂuxthe microwave retrieved from was themore microwave continuous was and more smoother continuous in the same and smootherarea, and inthere the were same no area, jumps and in there edge were regions no jumps (Figure in edge4b). regions (Figure4b).

a b

FigureFigure 4.4. ((aa)) SensibleSensible heat flux ﬂux from FY-2G FY-2G and and FY-3C FY-3C (at (at Beijing Beijing time time 15:40). 15:40). (b (b) )Latent Latent heat heat flux ﬂux fromfrom FY-2G FY-2G and and FY-3C FY-3C (at (at Beijing Beijing time time 15:40). 15:40).

ThereThere waswas nono significantsignificant differencedifference inin the the spatial spatial distribution distribution ofof the the retrieved retrieved latent latent heatheat fluxesfluxes between the two two data data sources. sources. The The two two methods methods had had specific specific advantages advantages and anddisadvantages. disadvantages. Although Although the surface the surface temper temperatureature retrieved retrieved by the by FY-3C the FY-3C microwave microwave was wasnot affected not affected by clouds, by clouds, its resolution its resolution was ab wasout about25 km. 25 The km. surface The surfacetemperature temperature retrieval retrievalneeded to needed be spatially to be interpolated spatially interpolated before being before used, being and the used, accuracy and thewas accuracy slightly lower was slightlythan that lower of the than thermal that of infrared the thermal remote infrared sens remoteing. The sensing. polar-orbitingThe polar-orbiting satellites observed satellites observedvariables variablesinstantaneously, instantaneously, whereas whereasthe geosta thetionary geostationary satellites satellites obtained obtained one-hour one-hour average averagevalues. values. 4.2. Verification 4.2. Verification Due to the footprint effect in the turbulence observations, the value observed by the Due to the footprint effect in the turbulence observations, the value observed by the eddy covariance system can not only be regarded as a value at a certain point but is of eddy covariance system can not only be regarded as a value at a certain point but is of spatial representativeness from tens of meters to several kilometers. The time scale of the spatial representativeness from tens of meters to several kilometers. The time scale of the remote sensing was one hour, so the hourly observation data from the eddy covariance remote sensing was one hour, so the hourly observation data from the eddy covariance system were collected during the experiment to verify the estimated results. Figure5 shows system were collected during the experiment to verify the estimated results. Figure 5 the comparisons between hourly surface energy flux measurements and the estimates on shows the comparisons between hourly surface energy flux measurements and the esti- 18 August 2019, including net radiation flux, soil heat flux, sensible heat flux, and latent mates on August 18th, 2019, including net radiation flux, soil heat flux, sensible heat flux, heat flux. The net radiation flux and sensible heat flux estimated by satellite remote sensing and latent heat flux. The net radiation flux and sensible heat flux estimated by satellite were compared with the ground measurements, and the root mean square error were remote sensing were compared with the ground measurements, and the root mean square 108.31 Wm−2 and 68 Wm−2 respectively; the root mean square error of the estimated soil −2 −2 heaterror flux were and 108.31 that obtained Wm and by 68 the Wm TCAV respectively; method was the 14.28 root Wm mean−2 and square the root error mean of the square esti- −2 errormated of soil the heat estimated flux and latent that heat obtained flux and by thethe groundTCAV method measurement was 14.28 was Wm 45.64 Wmand −the2. Theroot errorsmean weresquare all error within of a the reasonable estimated range. latent heat flux and the ground measurement was 45.64 Wm−2. The errors were all within a reasonable range.

a b

Remote Sens. 2020, 12, x 10 of 15

Wm−2 (Figure 4b). However, the latent heat flux obtained from the visible and near-infrared band displayed sharp jumps in the cloudy area and its edge(Figure 4a), indicating that the results of the geostationary satellite obviously affected the retrieval results. The latent heat flux retrieved from the microwave was more continuous and smoother in the same area, and there were no jumps in edge regions (Figure 4b).

a b

Figure 4. (a) Sensible heat flux from FY-2G and FY-3C (at Beijing time 15:40). (b) Latent heat flux from FY-2G and FY-3C (at Beijing time 15:40).

There was no significant difference in the spatial distribution of the retrieved latent heat fluxes between the two data sources. The two methods had specific advantages and disadvantages. Although the surface temperature retrieved by the FY-3C microwave was not affected by clouds, its resolution was about 25 km. The surface temperature retrieval needed to be spatially interpolated before being used, and the accuracy was slightly lower than that of the thermal infrared remote sensing. The polar-orbiting satellites observed variables instantaneously, whereas the geostationary satellites obtained one-hour average values.

4.2. Verification Due to the footprint effect in the turbulence observations, the value observed by the eddy covariance system can not only be regarded as a value at a certain point but is of spatial representativeness from tens of meters to several kilometers. The time scale of the remote sensing was one hour, so the hourly observation data from the eddy covariance system were collected during the experiment to verify the estimated results. Figure 5 shows the comparisons between hourly surface energy flux measurements and the estimates on August 18th, 2019, including net radiation flux, soil heat flux, sensible heat flux, and latent heat flux. The net radiation flux and sensible heat flux estimated by satellite remote sensing were compared with the ground measurements, and the root mean square error were 108.31 Wm−2 and 68 Wm−2 respectively; the root mean square error of the esti- Remote Sens. 2021, 13, 53 mated soil heat flux and that obtained by the TCAV method was 14.28 Wm−2 and the11 ofroot 15 mean square error of the estimated latent heat flux and the ground measurement was 45.64 Wm−2. The errors were all within a reasonable range.

a b

Remote Sens. 2020, 12, x 11 of 15

c d

FigureFigure 5.5. ComparisonsComparisons betweenbetween hourly hourly surface surface energy energy flux flux measurements measurements and and the the estimated: estimated: (a )(a net) radiation,net radiation, (b) soil (b heat) soil flux, heat (c )flux, sensible (c) sensible heat flux, heat (d) latent flux, heat (d) latent flux. heat flux.

5.5. DiscussionsDiscussions 5.1. The Relationship between the Conditional Expectation and the Temperature 5.1. The Relationship between the Conditional Expectation and the Temperature Surface temperature is the most direct feedback of the surface to meteorological condi- Surface temperature is the most direct feedback of the surface to meteorological con- tions and is one of the key factors affecting the net radiation distribution on the surface. ditions and is one of the key factors affecting the net radiation distribution on the surface. The surface temperature and energy distribution showed certain relationships at the spatial The surface temperature and energy distribution showed certain relationships at the spa- scale due to the climate and land cover of the study area. The spatially conditional expecta- tial scale due to the climate and land cover of the study area. The spatially conditional tion of the surface energy components and surface temperature retrieved by remote sensing wasexpectation derived andof the analyzed surface afterenergy smoothing components (Figure and6a). surface The trend temperature line was smoothed.retrieved by The re- resultsmote sensing show that was the derived distribution and analyzed of net radiation after smoothing generally (Figure displayed 6a). The a similar trend trendline was to thatsmoothed. of sensible The heat.results From show a starting that the point distribu of 275tion K, of with net theradiation increase generally in temperature, displayed the a conditionalsimilar trend expectations to that of sensible of net radiation heat. From and a latent starting heat point flux of increased, 275 K, with until the stabilizing increase atin 298–301temperature, K, and the decreasing conditional after expectations 302 K. Sensible of ne heatt radiation flux showed and anlatent upward heat trendflux increased, from 275 tountil 288 stabilizing K, but decreased at 298–301 slowly K, afterand 288decreasing K. In general, after 302 the K. expected Sensible values heat offlux sensible showed heat, an netupward radiation trend and from latent 275 to heat 288 flux K, but from decreased 275 to 280 slowly K were after consistent, 288 K. In showinggeneral, the an expected upward trend.values However, of sensible starting heat, net from radiation 290 K, theand expected latent heat value flux of from sensible 275 to heat 280 decreasedK were consistent, slightly, whileshowing the an expected upward latent trend. heat However, increased. starting The expectedfrom 290 valueK, the ofexpected soil heat value flux of increased sensible slowly.heat decreased Therefore, slightly, in mid-August, while the the expected daily surface latent sensibleheat increased. heat flux The in theexpected SRYR overvalue the of Qinghai-Tibetsoil heat flux Plateauincreased rose slowly. with theTherefore, increase in in mid-August, temperature belowthe daily 288 surface K. The sensiblesensible heatheat fluxflux changedin the SRYR slightly over within the Qinghai-Tibet the temperature Plateau range rose of 280–285 with the K, increase and the latentin temperature heat flux increasedbelow 288 significantly K. The sensible between heat flux 280 changed and 298 K.slightly within the temperature range of 280– 285 K, and the latent heat flux increased significantly between 280 and 298 K.

a b

Figure 6. (a) Trend of energy fluxes conditional expectation with the land surface temperature. (b) Relationship between evaporation fraction (EF) and latent heat flux.

The above analysis shows that under the abundant surface soil moisture conditions, with the increase in temperature and net radiation, the effective energy distribution of the surface changed significantly with the increase in temperature; in particular, when the surface temperature exceeded 290 K, the effective energy was mainly used for surface ET.

Remote Sens. 2020, 12, x 11 of 15

c d

Figure 5. Comparisons between hourly surface energy flux measurements and the estimated: (a) net radiation, (b) soil heat flux, (c) sensible heat flux, (d) latent heat flux.

5. Discussions 5.1. The Relationship between the Conditional Expectation and the Temperature Surface temperature is the most direct feedback of the surface to meteorological conditions and is one of the key factors affecting the net radiation distribution on the surface. The surface temperature and energy distribution showed certain relationships at the spatial scale due to the climate and land cover of the study area. The spatially conditional expectation of the surface energy components and surface temperature retrieved by remote sensing was derived and analyzed after smoothing (Figure 6a). The trend line was smoothed. The results show that the distribution of net radiation generally displayed a similar trend to that of sensible heat. From a starting point of 275 K, with the increase in temperature, the conditional expectations of net radiation and latent heat flux increased, until stabilizing at 298–301 K, and decreasing after 302 K. Sensible heat flux showed an upward trend from 275 to 288 K, but decreased slowly after 288 K. In general, the expected values of sensible heat, net radiation and latent heat flux from 275 to 280 K were consistent, showing an upward trend. However, starting from 290 K, the expected value of sensible heat decreased slightly, while the expected latent heat increased. The expected value of soil heat flux increased slowly. Therefore, in mid-August, the daily surface sensible heat Remote Sens. 2021, 13, 53 flux in the SRYR over the Qinghai-Tibet Plateau rose with the increase in temperature12 of 15 below 288 K. The sensible heat flux changed slightly within the temperature range of 280– 285 K, and the latent heat flux increased significantly between 280 and 298 K.

a b

FigureFigure 6. 6. (a()a Trend) Trend of of energy energy fluxes ﬂuxes conditional conditional expectat expectationion with with the the land land surface surface temperature. temperature. (b ()b ) RelationshipRelationship between between evaporation evaporation fraction fraction (EF) (EF) and and latent latent heat heat flux. ﬂux.

TheThe above above analysis analysis shows shows that that under under the the ab abundantundant surface surface soil soil moisture moisture conditions, conditions, withwith the the increase increase in in temperature temperature and and net net radiat radiation,ion, the the effective effective energy energy distribution distribution of of the the surfacesurface changed changed significantly signiﬁcantly with with the the increase increase in in temperature; temperature; in in particular, particular, when when the the surfacesurface temperature temperature exceeded exceeded 290 290 K, K, the the effective effective energy energy was was mainly mainly used used for for surface surface ET. ET. This also indicates that, when the surface temperature exceeded 290 K, its retrieval accuracy had a signiﬁcant impact on the ET retrieval, i.e., the higher the surface temperature, the greater the ET retrieval error.

5.2. The Relationship between Evaporation Fraction, Latent Heat Flux, and Daily ET The evaporation fraction represented the energy exchange between the lower atmosphere and the near-surface at a regional scale, that is, the distribution of effective energy between surface sensible heat and latent heat flux. Under ideal conditions, when the soil was at its dry limit and the soil moisture was not available for evaporation, the latent heat flux was approximately zero and the sensible heat flux reached its maximum. In the wet limit environment, in which the soil moisture was sufficient, the latent heat flux reached its maximum. The First ISLSCP Field Experiment (FIFE) indicated that the evaporation fraction remained relatively stable under typical clear sky conditions, with the exception of mornings and evenings, so the ET throughout the day could be derived from the evaporation fraction at a certain moment and the surface effective energy of the whole day [26,27]. The observation from geostationary satellites made it possible to obtain the hourly scale evaporation fraction by remote sensing and provided a data source for verification of the hourly scale evaporation fraction using the eddy covariance observation system. Figure6b presents the retrieved and observed latent heat flux and evaporation fraction at an hourly scale. Although the evaporation fractions of the two were similar, with values between 0.51 and 0.65, the fitting straight line showed different trends. The observed evaporation fraction showed a more obvious trend of decreasing with the increase in latent heat flux, whereas the retrieved evaporation fraction displayed a slight increase. The two lines intersected when the latent heat flux was 210.19 Wm−2. However, due to the lack of accurate model-driven and surface verification data-sets, which leads to errors in the model forcing field and uncertainty of the model parameters, model calibration and error analysis based on land surface observations was complicated. More importantly, SEBS is a closure energy model, which retrieves surface ET based on the energy residual method, whereas the eddy covariance system used by the station for surface flux observation is not closed, however, and is still widely used for surface ET observation. The evaporation fraction was evaluated using the observations from the eddy covariance system and the retrievals from the remote sensing data. At 12:00, the observed evaporation fraction was the closest to the estimated evaporation fraction. Hence, 12:00 was defined as the analysis time. The retrieved evaporation fraction at this time was 0.57, which was considered to be the effective evaporation fraction of the day. The daily ET calculated based on this value was 4.85mm, and the value of the total ET derived from the Remote Sens. 2021, 13, 53 13 of 15

evaporation fraction at other times was within the range of 4.04–5.25 mm. Compared with the cumulative value, the daily ET derived from the effective evaporation fraction was closer to the observed value with the same input forcing and the model physics schemes. Therefore, the evaporation fraction at 12:00 was used to obtain the daily ET and this method was able to eliminate the retrieval error at a certain moment without destroying the physical mechanism of the retrieval model. The method has the potential to become an effective means of retrieving ET using geostationary satellite observations.

5.3. Daily ET The spatial distribution of daily ET on 18 August 2019 in the SRYR was obtained by collecting the evaporation fraction and accumulating the hourly latent heat ﬂux (Figure7). The regional ET was between 2.0 and 4.0 mm. The low-value area appeared in the north of the study area with ET below 2.5 mm. The maximum daily ET was about 4.0 mm in the southeast of the SRYR, indicating that the surface soil water, vegetation, water available for ET, and meteorological conditions were all conducive to ET. In contrast, the northern part of the study area is located at the northern foot of Tanggula Mountain, which is not favorable for energy transfer from the surface to the atmosphere due to the high altitude Remote Sens. 2020, 12, x and harsh weather conditions. 13 of 15

FigureFigure 7. 7.SpatialSpatial distribution distribution of of daily daily evapotranspirati evapotranspirationon (ET) (ET) over over Northern Northern Tibetan Tibetan Plateau Plateau on on 18 AugustAugust 18th, 2019. 2019.

6.6. Conclusions Conclusions In this research, the spatial distribution of energy fluxes and daily ET at different times In this research, the spatial distribution of energy fluxes and daily ET at different of the day in the SRYR were estimated using the SEBS algorithm and ground meteorological times of the day in the SRYR were estimated using the SEBS algorithm and ground mete- observation data, combined with the data of visible/near-infrared and microwave obser- orological observation data, combined with the data of visible/near-infrared and micro- vations from the Fengyun series of meteorological satellites. Considering the influence of wave observations from the Fengyun series of meteorological satellites. Considering the convective clouds during the afternoon above the plateau, which prevents visible and near- influence of convective clouds during the afternoon above the plateau, which prevents infrared observation in cloud-cover conditions, we used the MWRI brightness temperature visible and near-infrared observation in cloud-cover conditions, we used the MWRI to estimate the surface temperature. The hourly change of latent heat flux presented a brightness temperature to estimate the surface temperature. The hourly change of latent sound unimodal diurnal variation. As the solar radiation increased after sunrise, the latent heat flux presented a sound unimodal diurnal variation. As the solar radiation increased heat flux gradually increased from 180.0 Wm−2 at 11:00 to 200.0 Wm−2 at 12:00. The latent after sunrise, the latent heat flux gradually increased from 180.0 Wm−2 at 11:00 to 200.0 heat flux was about 280.0 Wm−2 at 13:00. At 14:00, the value reached its daily maximum Wm−2 at 12:00. The latent heat flux was about 280.0 Wm−2 at 13:00. At 14:00, the value of around 350.0 Wm−2, and in some areas reached 400.0 Wm−2. As the solar altitude, −2 −2 reachedradiation, its daily and airmaximum saturation of around deficit weakened 350.0 Wm in, and the afternoon,in some areas at 16:00, reached we 400.0 adopted Wm the. Assurface the solar temperature altitude, radiation, retrieved and from air the satu microwaveration deficit data weakened and found in the the latent afternoon, heat fluxat 16:00,in the we cloud-free adopted the area surface was about temperature 280.0 Wm retrieved−2, and from the cloud-covered the microwave area data 150.0 and found Wm−2 . −2 theConverting latent heat the flux latent in the heat cloud-free flux to the area actual was landabout surface 280.0 Wm ET, it,was andfound the cloud-covered that the daily −2 areaET in150.0 most Wm of the. Converting study area the was latent between heat flux 2.0 mmto the and actual 4.0 mm. land Thesurface results ET, wereit was within found a thatreasonable the daily error ET in range most andof the lay study a foundation area was for between realizing 2.0 themm commercialization and 4.0 mm. The results of land weresurface within ET ina reasonable the future. error range and lay a foundation for realizing the commerciali- zationThe of land focus surface of this ET research in the future. was to investigate the hourly retrieval of ET using satellite remoteThe sensingfocus of data.this research Due to thewas limitations to investigate of ground the hourly observation retrieval data of ET and using the overpasssatellite remotetime of sensing polar-orbiting data. Due satellites, to the limitations the accuracy of ground of the algorithm observation under data cloud and coverthe overpass was not time of polar-orbiting satellites, the accuracy of the algorithm under cloud cover was not further analyzed. In addition, cloud-free and cloud-covered represent theoretical conditions, and cloud total amount (CTA) was obtained using the radiation transfer equation and the data from the visible and near-infrared channels. This could be used to determine whether the study area at the regional scale was under cloud cover, but the accuracy of determining whether a single spot was affected by cloud cover was yet to be considered. In the next step, it will be necessary to take manual field observation experiments, that is, to record the cloud amount at the observation spot for each moment, and analyze the difference between the measured ET and the estimated results under cloud cover. The focus will be to improve the estimation accuracy of the model under cloud cover for a more reliable estimate of regional ET under all weather conditions.

Author Contributions: conceptualization and writing—original draft preparation R.L. supervision, J.W.; investigation, X.W.; formal analysis, Z.W. and Y.L.; software, M.Z. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by supported by The Second Tibetan Plateau Scientific Expedi- tion and Research (STEP) program (grant no. 2019QZKK0105); National Natural Science Foundation of China, grant number 42075065, 91737103, 41530529. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable.

Remote Sens. 2021, 13, 53 14 of 15

further analyzed. In addition, cloud-free and cloud-covered represent theoretical conditions, and cloud total amount (CTA) was obtained using the radiation transfer equation and the data from the visible and near-infrared channels. This could be used to determine whether the study area at the regional scale was under cloud cover, but the accuracy of determining whether a single spot was affected by cloud cover was yet to be considered. In the next step, it will be necessary to take manual ﬁeld observation experiments, that is, to record the cloud amount at the observation spot for each moment, and analyze the difference between the measured ET and the estimated results under cloud cover. The focus will be to improve the estimation accuracy of the model under cloud cover for a more reliable estimate of regional ET under all weather conditions.

Author Contributions: Conceptualization and writing—original draft preparation R.L. supervision, J.W.; investigation, X.W.; formal analysis, Z.W. and Y.L.; software, M.Z. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by supported by The Second Tibetan Plateau Scientific Expedition and Research (STEP) program (grant no. 2019QZKK0105); National Natural Science Foundation of China, grant number 42075065, 91737103, 41530529. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: Not applicable. Acknowledgments: Special thanks are given to the anonymous reviewers who greatly helped to improve the quality of the manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Priestley, C.H.B.; Taylor, R.J. On the assessment of surface heat flux and evaporation using large-scale parameters. Mon. Weather Rev. 1972, 100, 81–92. [CrossRef] 2. Wang, K.C.; Dickinson, R.E. A review of global terrestrial evapotranspiration: Observation, modeling, climatology, and climatic variability. Rev. Geophys. 2012, 50.[CrossRef] 3. Kim, N.; Kim, K.; Lee, S.; Cho, J.; Lee, Y. Retrieval of daily reference evapotranspiration for croplands in South Korea using machine learning with satellite images and numerical weather prediction data. Remote Sens. 2020, 12, 3642. [CrossRef] 4. Su, Z.; Schmugge, T.; Kustas, W.P.; Massman, W.J. An evaluation of two models for estimation of the roughness height for heat transfer between the land surface and the atmosphere. J. Appl. Meteorol. 2001, 40, 1933–1951. [CrossRef] 5. Nedbal, V.; Láska, K.; Brom, J. Mitigation of arctic tundra surface warming by plant evapotranspiration: Complete energy balance component estimation using LANDSAT satellite data. Remote Sens. 2020, 12, 3395. [CrossRef] 6. Cleugh, H.A.; Lenning, R.; Mu, Q. Regional evaporation estimates from flux tower and MODIS satellite data. Remote Sens. Environ. 2007, 106, 285–304. [CrossRef] 7. Mu, Q.; Heinsch, F.; Zhao, M.; Running, S. Development of a global evapotranspiration algorithm based on MODIS and global meteorology data. Remote Sens. Environ. 2007, 111, 519–536. [CrossRef] 8. Jackson, R.D.; Reginato, R.J.; Idso, S.B. Wheat canopy temperature: A practical tool for evaluating water requirements. Water Resour. Res. 1977, 13, 651–656. [CrossRef] 9. Allen, R.G.; Tasumi, M.; Trezza, R. Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC)-model. J. Irrig. Drain. Eng. 2007, 133, 380–394. [CrossRef] 10. Carpintero, E.; Mateos, L.; Andreu, A.; González-Dugo, M.P. Effect of the differences in spectral response of Mediterranean tree canopies on the estimation of evapotranspiration using vegetation index-based crop coefficients. Agric. Water Manag. 2020, 238, 106201. [CrossRef] 11. Jiang, L.; Islam, S. A methodology for estimation of evapotranspiration over large areas using remote sensing information. Geophys. Res. Lett. 1999, 26, 2773–2776. [CrossRef] 12. Bastiaanssen, W.G.M.; Menenti, M.; Feddes, R.A.; Holtslag, A.A.M. A remote sensing surface energy balance algorithm for land (SEBAL): 1. Formulation. J. Hydrol. 1998, 212–213, 198–212. [CrossRef] 13. Roerink, G.J.; Su, Z.; Menenti, M. A simple remote sensing algorithm to estimate the surface energy balance. Phys. Chem. Earth 2000, 25, 147–157. [CrossRef] 14. Su, Z. The surface energy balance system (SEBS) for estimation of turbulent heat fluxes. Hydrol. Earth Syst. Sci. 2002, 6, 85–99. [CrossRef] 15. Allen, R.G.; Tasumi, M.; Morse, A. Satellite-Based Evapotranspiration by METRIC and Landsat For Western States Water Management; US Bureau of Reclamation Evapotranspiration Workshop: Fort Collins, CO, USA, 2005. Remote Sens. 2021, 13, 53 15 of 15

16. Shuttleworth, W.J.; Wallace, J.S. Evaporation from sparse crops-an energy combination theory. Q. J. R. Meteorol. Soc. 1985, 111, 839–855. [CrossRef] 17. Norman, J.M.; Kustas, W.P.; Humes, K.S. Source approach for estimating soil and vegetation energy fluxes in observations of directional radiometric surface temperature. Agric. For. Meteorol. 1995, 77, 263–293. [CrossRef] 18. Chen, S.; Liu, W.H.; Ye, T. Dataset of trend-preserving bias-corrected daily temperature, precipitation and wind from NEX-GDDP and cmIP5 over the Qinghai-Tibet Plateau. Data Brief. 2020, 31, 105733. [CrossRef] 19. Chen, X.L.; Su, Z.; Ma, Y.M.; Yang, K.; Wang, B. Estimation of surface energy fluxes under complex terrain of Mt. Qomolangma over the Tibetan Plateau. Hydrol. Earth Syst. Sci. 2013, 17, 1607–1618. [CrossRef] 20. Brutsaert, W. On a derivable formula for long-wave radiation from clear skies. Water Resour. Res. 1975, 11, 742–744. [CrossRef] 21. Liu, R.; Wen, J.; Wang, X.; Zhang, Y. Evapotranspiration estimated by using datasets from the Chinese FengYun-2D geostationary meteorological satellite over the Yellow River source area. Adv. Space Res. 2015, 55, 60–71. [CrossRef] 22. Li, X.Q.; Yang, H.; You, R. Remote sensing typhoon Songda’s rainfall structure based on Microwave Radiation Imager of FY-3B satellite. Chin. J. Geophys. 2013, 28, 875–889. (In Chinese) 23. Han, C.; Ma, Y.; Chen, X.; Su, Z. Trends of land surface heat fluxes on the Tibetan Plateau from 2001 to 2012. Int. J. Clim. 2017, 37. [CrossRef] 24. Gao, Q.; Wang, S.; Yang, X. Estimation of surface air specific humidity and air–sea latent heat flux using FY-3C microwave observations. Remote Sens. 2019, 11, 466. [CrossRef] 25. Niu, Z.; Zou, X.; Ray, P.S. Development and testing of a clear-sky data selection algorithm for FY-3C/D microwave temperature sounder-2. Remote Sens. 2020, 12, 1478. [CrossRef] 26. Xu, Z.; Liu, S.; Xu, T. The observation and calculation method of soil heat flux and its impact on the energy balance closure. Adv. Earth Sci. 2013, 28.[CrossRef] 27. Shuttleworth, W.J.; Gurney, R.J.; Hsu, A.Y. FIFE: The variation in energy partition at surface flux sites. Remote Sens. Large Scale Glob. Process. 1989, 186, 67–74.