Radiometric Calibration Analysis for Thermal Infrared Data From MERSI-LL Onboard the Dust-Dawn Orbiting Satellite FY3E

FengYun-3E (FY3E) is the world's first dust-dawn orbiting meteorological satellite for civil use, which has filled the vacancy of global early-morning-orbit satellite observation by working together with FengYun-3C and FengYun-3D satellites. The Medium Resolution Spectral Imager-Low Light (MERSI-LL) sensor carried by FY3E can detect surface temperature variation. The radiometric calibration conditions of MERSI-LL thermal infrared bands were evaluated using the collected field measurements and atmospheric transfer simulations during 5–23 December, 2022, at Lake Erhai. A thermal infrared radiometer equipped on an unmanned surface vehicle was used to continuously collect surface-emitted radiance and water temperature. Atmospheric conditions, surface emissivity, and aerosol optical depth measured near-field experiment site were adopted by atmospheric radiative transfer code to calculate the influence of atmosphere on long-wave radiation during the propagation from land surface to satellite aperture. The good in-orbit operational status could be detected according to our calibration experiments, and the calibration analysis suggested that the differences between the simulated brightness temperature and satellite-based brightness temperature are 0.71 K with an RMSE of 0.79 K and 0.11 K with an RMSE of 0.47 K for channels 6 and 7, respectively, which achieved better or similar calibration accuracy by comparing with TIR channels of other FengYun satellites.


I. INTRODUCTION
A S THE second generation of Chinese Sun-synchronous meteorological satellites, Fengyun3 (FY3) carried several key payloads to detect meteorological parameters used in weather analysis, numerical weather forecasting, climate prediction, and environment and disaster [1], [2].As the important operational meteorological satellites of China's secondgeneration polar-orbiting FengYun meteorological satellite series, FengYun-3C (FY3C), FengYun-3D (FY3D), and FengYun-3E (FY3E) have been playing important roles in observing global weather/climate conditions and related land surface processes [3], [4], [5], [6], [7].FY3 satellites have provided land surface temperature (LST) scientific products with spatial resolution greater than 1 km, and the time series of data products greater than 10 years [3], [5].In order to improve the application of scientific datasets in weather prediction, climate monitoring, and environmental research, the radiometric recalibration work has been conducted and applied in the instruments to create the long-term consistent historical datasets for climate analysis [8], [9], [10].The absolute radiometric calibration process and the evaluation of in-orbit operational calibration condition of the satellite-based sensors have already become the most important steps for their quantitative application [11], [12], [13].
LST is one of important parameters in examining surface thermal environment, energy budget, and climate change [14], [15], [16], [17].Along with the increasing frequency of extreme heat events under global warming, it is necessary to accurately examine their spatial-temporal change using consistent temperature records from satellite-based thermal infrared (IR) sensors [18], [19].Due to the specification differences of thermal channels of FengYun satellites, it is necessary to examine their radiometric calibration conditions for understanding the possible uncertainties of LST products.The Medium Resolution Spectral Imager (MERSI) onboard FY3A, FY3B, and FY3C is designed as the next generation instruments to replace the visible infrared radiometer (VIRR), which is similar to MODIS with 20 visible and IR channels [1].As the first operational meteorological satellite on dawn-dusk orbit, the ground track of FY3E is near the terminator line that distinguishes the illuminated face of the sunlit Earth and its night side [2].Therefore, FY3E could contribute to the completeness of the baseline configuration of the core polar operational constellation, and the observed meteorological parameters by FY3E would be greatly helpful in improving the accuracy of numerical weather prediction.The key meteorological parameters, including atmospheric precipitable water and surface temperature, could be detected using the IR channels from the Medium Resolution Spectral Imager-Low Light (MERSI-LL) (see Table I) [2], [7].As the collaborative observation performed by FY3C, FY3D, and FY3E, they are also suitable for detecting diurnal variation and distribution evolution of extreme climate events with their similar spatial and spectral specifications of the popular Sun-synchronous meteorological sensors [6], [9], [20].MERSI-LL has improved its capability in detecting key parameters of atmosphere, land, and ocean, and the spatial resolution of thermal IR channels increased to 250 m.The thermal IR split-window channels, 10.3-11.3 and 11.5-12.5 μm, could be used to accurately examine the conditions of LST and sea surface temperature with a noise equivalent temperature difference of 0.4 K [2].
The periodic radiometric calibration evaluation is an important operational procedures in improving the accuracy and consistence of thermal IR datasets, further identifying the possible uncertainties to understand the spatial-temporal characterization of surface temperature [21], [22].The radiometric calibration analysis is usually conducted in different phases to improve the thermal IR data quality, including laboratory calibration, onboard calibration, and in-flight calibration [23], [24], [25].The prelaunch calibration conducted in laboratory could thoroughly quantify the efficiency of the photoelectric structure and spectral variation of each element of the detector, and then, the reasonable adjustments could be applied to improve the consistence of different elements and produce high quality satellite images [6], [26].An onboard coefficient of radiometric calibration is derived from the calibration system by the scheduled calibration procedure, such as the observation for blackbody and deep space, then the instrument responses can be quantified, which further used to produce the calibrated datasets by converting the digital numbers (DN) into TOA radiance or reflectance [13], [25].In-flight absolute radiometric calibrations are usually conducted in the large and homogeneous calibration site, such as Lake Titicaca, Lake Qinghai, and Lake Tahoe [11], [23], [27], [28].During the concurrent field experiments, surface radiance, atmospheric conditions, and aerosol optical depth (AOD) are measured by field instruments, those parameters can be adopted in the atmospheric radiative transfer simulation to predict the radiance at the sensor's aperture and evaluate radiometric calibration conditions [11], [23], [29], [30].In order to improve the accuracy of the measured surface parameters, the automatic or semiautomatic instruments have been used in the field experiments to provide more data samples for calibration analysis, such as the lake buoys or ocean buoys, the field spectroradiometer, or the automatic TIR radiometer observation deployed the unmanned surface vehicles (USV) [23], [28], [31].Most vicarious calibration investigations for satellite TIR sensors suggested that the calibration accuracy is mainly limited to the weather conditions (the water vapor contents) and the uncertainties of field measurements [11], [32].The simulation using atmospheric transfer model could examine the possible influence of atmosphere condition on long wave radiation, which can be used to evaluate the in-orbit radiometric calibration conditions of TIR sensors [24], [33].
FY-3 satellite series will continuously support the studies related to climate change and land surface processes in the next decades, and the derived surface temperature products have already played important roles in understanding surface thermal environments, urban climate change, and Earth energy budget.MERSI-LL onboard FY3E can fill the observation gap of polarorbiting meteorological satellite within the 6-h assimilation window on dusk-dawn orbit, and improve the observation capability of LST by comparing with the existing polar-orbiting satellites.Therefore, it is necessary to monitor its in-orbit calibration status and identify the possible uncertainty sources of LST datasets by using accurate radiometric calibration analysis.We conducted field campaign at Laker Erhai in December 2022 to examine the properties of surface temperature variation and atmospheric conditions.Then, the measurement datasets and atmospheric propagation simulation adopted in calibration analysis were used to calculate the TOA radiance, and evaluated the calibration accuracy using the bias between satellite-measured radiance and predicted radiance at the time of satellite overpass, further analyzed the status of in-orbit radiometric calibration of MERSI-LL TIR channels.

A. Field Campaign Overview
Lake Erhai in Yunnan Province, China, was selected as the calibration site in winter to conduct field campaigns.Lake Erhai located on the Yunnan-Guizhou Plateau, and it has an area of 251 km 2 with an average elevation 1972 m above sea level.The average width of Lake Erhai is 6.4 km with the maximum depth of 21 m.It has a low-latitude plateau monsoon climate with a mean air temperature of approximately 15 °C and an average annual precipitation of 1028 mm [34].There is a distinct dry season from November to April of the following year and a typical rainy season from May to October, mainly in July and August.We conducted ground-based calibration experiments at Lake Erhai from 5 to 23 December, 2022 (see Table II).A high precision and portable field seven-channel IR radiometer (CE312) was equipped on a solar-powered USV to collect surface-emitted radiance and brightness temperature (BT) records.CE312 situated on the USV at a height of 2 m above water surface, and its two channels with center wavelength at 10.6 and 11.3 μm were used in the calibration analysis.USV cruised along a programmed navigation route from local time 10:00 to 20:00 under automatic navigation mode (see Fig. 1), and the measured records were transferred to the real-time monitoring computer.A ground-based microwave radiometer (QFW6000) near the calibration site was used to collect atmospheric profile information (air temperature, relative humidity, pressure, and water vapor).The diurnal changes of AOD at the calibration site were detected by a Sun Sky Lunar Multispectral Photometer (CE318).Weather conditions and lake surface conditions are also continuously monitored by USV sensors, including air temperature, relative humidity, water temperature, pressure, and wind speed.Finally, the surface measurements and atmospheric conditions were used to calculate the radiance at sensor aperture using atmospheric propagation simulations.The at-sensor thermal IR radiance comprises three components: the thermal radiance emitted from water surface and attenuated by the atmosphere, the upward atmospheric thermal radiance, and the radiance reflected by water surface attenuated by atmosphere [28], [33].Due to the relatively consistent spectral emissivity of a waterbody, the spectral radiance at TIR senor's aperture can be expressed as [29], [33] where λ is the wavelength, τ λ is atmospheric transmittance, B λ is the Planck function for BB spectral radiance, R ↑ λ is path radiance, and R ↓ λ is the atmospheric downwelling radiance.The at-sensor radiance is convolved with spectral response functions (f λ ) to obtain the equivalent radiance of a specific sensor channel, which can be expressed as (2)

B. Calibration Analysis
Generally, DN values from the thermal IR images of MERSI-LL can be converted to radiances according to the calibration equation released by National Satellite Meteorological Center (NSMC), Chinese Meteorological Administration [35].The radiance values were used to derive at-sensor BTs using the inverse Planck equation [23], and then, it was used to perform concurrent ground-based accuracy calibration analysis around the satellite overpass.According to the user instructions of MERSI-LL released by NSMC, the radiance values (mW/m2 •sr•cm −1 ) of each pixels in TIR images have been included in the Level 1 products, and the radiance values could be converted to at-sensor BT using the inverse Plank equation where T e is the derived BT using the inverse Planck equation (K), h is the Planck constant (6.626 × 10 −34 J•s), c is the speed of light (2.9979 × 10 8 m/s), k is the Boltzmann constant (1.3806 × 10 −23 J/K), λ is the center wavelength of each band, and L is the at-sensor radiance.Here, the center wavelengths of bands 6 and 7 of FY3E/MERSI-LL are 10.8 and 12 μm, respectively.Further, the correction calculation for TIR channels from NSMC was used to correct the derived BT where T bb is the final at-sensor BT, A and B are the correction coefficient, and 1.00121 and -0.2810 are used in the BT calculation process of band 6, while 1.00113 and -0.2286 are adopted for the BT calculation of band 7.
The TIR radiometer deployed on USV can continuously record surface radiance with 10-s intervals along a programmed navigation route.We selected the observation records within 20 min around the satellite overpass to calculate the averaged lake surface radiance to reduce the possible uncertainties from single observation.We found mismatches of relative spectral response (RSR) of TIR channels between MERSI-LL and CE312, which would cause the calibration errors induced by surface temperature differences for the same observed target from different TIR instruments (see Fig. 2).Here, we conduct spectrum matching to eliminate the influence of those disagreements on radiometric calibration based on a series of numerical experiments of the Moderate Resolution Atmospheric Transmission Model (MODTRAN) [23], [35].Lots of simulation combinations were used to generate TOA spectral radiance, including nine different targets (cloud cover, desert, dry grass, field, forest, fresh snow, maple, ocean, and wet grass), two satellite view angles (0°and 10°), six built-in atmospheric profiles of MODTRAN (Tropical, Mid-latitude Summer/Winter, Sub-Arctic Summer/Winter, U.S. Standard), and nine boundary temperatures (278 K, 283 K, 288 K, 293 K, 298 K, 303 K, 208 K, 212 K, and 320 K).The TOA radiances were convolved with spectral response functions of the selected TIR channels to derive the unified channel equivalent radiance according to the following equation: where R CE and R TIS are unified equivalent spectral radiance for CE312 and MERSI-LL, L mod,λ is the top-of-atmosphere spectral radiance simulated by MODTRAN, f CE,λ and f TIS,λ are the spectral response functions of the selected TIR bands from CE312 and MERSI-LL.Finally, the spectral matching factors were derived from the 972 pairs of unified channel equivalent radiance using linear regression analysis.Generally, our simulated TOA radiance series have covered different atmospheric conditions and land cover types, which would create the reasonable coefficient of spectrum matching for TIR sensors (see Fig. 3).
We calculated the at-sensor radiance by eliminating the influence of the atmosphere on the at-sensor radiance during the radiative propagation process from the lake surface to satellite aperture.MODTRAN model was used to simulate the atmospheric absorption and scattering effects with regard to thermal radiation.QFW6000 instrument around the satellite overpass at calibration site detected the atmospheric parameters at altitudes ranging from land surface to 10 km, and upper-air atmospheric parameters collected from National Centers for Environmental Prediction (NCEP)-Department of Energy (DOE) Reanalysis 2 datasets 1 to provide entire tropospheric layer information for atmospheric transfer simulations (see Fig. 4).Atmospheric aerosols' measurements by an automatic Sun tracking photometer (CE318) at the calibration site were used to calculate the optical depth at 550 nm and the visibility distance.These results were adopted in MODTRAN to derive the path thermal radiance and spectral transmittance during the propagation  process from the lake surface to the satellite aperture.Then, the simulation results were convolved with spectral response functions of FY3E/MERSI-LL to obtain band-averaged values of the atmospheric parameters and further predict the equivalent at-senor radiance.

A. Lake Surface Temperature Variation
We investigated the BT variation observed by CE312 deployed on the USV during the field experiments to examine the temperature consistence of the lake surface (see Fig. 5).Generally, the water temperature around the calibration site ranged from 13.27 °C to 15.31 °C, which indicated that the temporal variation of lake surface temperature was about 2 K.Meanwhile, we found that there are different patterns of lake temperature change in the morning and the afternoon according to the BT observation from local time 10:00 to 19:30 on 12 December, 2022.Lake surface temperature around the calibration site slowly increased from 13.27 °C to 14.26 °C in the morning, and this trend will maintain before 13:30 with the maximum temperature of 15.31 °C, then it had a slight decrease before 14:00, and then maintained a relative consistent temperature values.The variation of lake surface temperature may be mainly altered by the local climate conditions, and the multivariate correlation investigation suggested that lake surface temperature change mainly influenced by air temperature and long-wave radiation with their correlation coefficients greater than 0.97 [34].The temperature variation in the afternoon (mainly after 14:00) could maintain the homogeneous and consistent conditions with  the averaged BT 14.48 ± 0.17 °C.The further investigation about the temperature variations during the satellite overpass indicated that it less than 0.09 K during 20 min before or after the satellite overpass.Therefore, our selected period for calibration experiment is reasonable for calculate the averaged radiance values around satellite overpass for radiometric calibration analysis.Certainly, we also noted that Erhai a weak by comparing with Lake Qinghai from the selection of ideal calibration site for TIR instruments because lake temperature variation of Lake Qinghai is less than 1 K [28], [32].This situation may be mainly influenced by the alteration of heat capacity on lake temperature variation, and the relative larger variation of Lake Erhai due to its relative small area and the depth.Therefore, it is needed to select the appropriate period to conduct an accurate radiometric calibration analysis when Lake Erhai is selected as a calibration site for in-flight radiometric calibration evaluation of TIR instrument.

B. BT Bias Analysis
Water surface radiances observed from field experiments, the upward spectral transmittance, and path radiance were convoluted with the RSR function of FY3E/MERSI-LL band 6 and 7 to predict at-sensor radiance.In order to reduce the uncertainty of spectrum mismatch between TIR bands of satellite sensor and field sensor, the spectrum-matching factor was used to correct the radiance.At-sensor radiances were converted to BT by using the inverse Planck equation.Then, we can evaluate the radiometric calibration conditions of MERSI-LL using the BT bias.Generally, the calibration analysis from 5 to 23 December, 2022, indicated that almost all records about BT bias change were less than 1 K except for BT bias on 23 December (1.02 K for band 6) (see Fig. 6).The average temperature biases reached 0.71 K with an RMSE of 0.79 K for band 6 and 0.11 K with an RMSE of 0.47 K for band 7, respectively, which indicated that band 7 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.Fig. 7. At-sensor radiance differences (W/m2•sr•µm) and at-sensor BT difference (K) of TIR bands calculated using the atmospheric profile parameters from QFW6000 measurements and NCEP reanalysis data. of MERSI-LL has better radiometric calibration conditions than that of band 6.The relative larger BT bias on December 16 and December 20 (0.938 and 1.020 K for band 6) may be influenced by the water vapor contents (see Fig. 4).By contrast, lower water vapor contents (0.224 cm) are detected with the better calibration results happened on 5 December with BT bias of 0.594 and 0.115 K for bands 6 and 7, respectively.The calibration accuracy well satisfied the requirement of in-flight calibration accuracy for most popular satellite-based TIR sensors (less than 1 K), such as ASTER, FY4A/AGRI, FY3D/MERSI-II, and Landsat [23], [30], [35], [36], [37].
According to the diurnal variation of lake surface temperature, we mainly selected the observation parameters captured at dusk, about local time 19:00, to conduct the concurrent calibration experiment and calibration analysis (see Table II).As a new operational meteorological instrument of FY3 series, the radiometric calibration of FY3E/MERSI-LL has been improved by comparing with other instruments, such as the calibration accuracy by 1.5 and 2.1 K in bands 4 and 5 of FY2C/MSR [38], the calibration accuracy of -0.46 and -1.04 K for bands 24 and 25 of FY3D/MERIS-II [35], and the calibration accuracy for bands 12 and 13 of FY4A/AGRI by 0.34 and 0.65 K. Regarding to the calibration results from vicarious calibration experiments, the radiometric calibration conditions of TIR channels of FY3E/MERSI-LL are similar to that of other popular satellite-base TIR instruments, such as 0.6 K for band 6 of Landsat 5 according to the long term calibration analysis [22], 0.87 and 1.67 K for bands 10 and 11 of Landsat 8/TIRS [30], and 1 K for ASTER thermal bands using vicarious calibration analysis from 2000 to 2013 [39].

C. Uncertainty Analysis
The uncertainties of radiometric calibration analysis of satellite-based sensors TIR by using vicarious calibration method are mainly influenced by field measurements, data processing, and calibration method selection [24], [32], [36].Many studies have investigated the surface radiance observed at calibration site, atmospheric parameters measurements, the simulation process of the atmospheric radiative transfer model, Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.and the selection of the concurrent pixel [23], [32], [35], [38].According to our investigation at Lake Erhai during field experiment, this study mainly analyzed the possible uncertainties from the field observation, atmospheric profiles differences, and spectral matching processes, and the overall calibration uncertainty caused by each parameter can be estimated by combining those different factors.
1) Observation Uncertainty: Due to the emissivity of lake water highly approaching 1, water temperature records collected by USV were considered as the reference temperature.They can be used to identify the difference between water surface BT measured by CE312 and water temperature from USV for understanding the observation uncertainty.There are about 480 records from local time 10:00 to 19:30 on 12 December, 2022, and the temperatures range from 286.36 to 288.55 K.In order to accurately examine the temperature difference under relative homogeneous conditions, we mainly selected the field experiment period from local time 14:00 to 19:00 with temperature range less than 0.9 K.The results indicated that the averaged absolute differences between surface BTs and surface temperatures reach about 0.54 K with a root mean square (rms) of 0.55 K for band 3, while the averaged absolute differences increase to 0.64 K with an rms of 0.65 K for band 1.Our results were greater than about 1 K comparing with the previous investigation at Lake Qinghai using the automated hydrometeorological buoy system (0.44 K) [38].
2) Uncertainty Induced by Atmospheric Profile Parameters: Due to the limitation of QFW6000 instrument, we can get atmospheric profile parameters at altitudes ranging from land surface to 10 km, which almost cover the entire troposphere at the calibration site.There are still slight amount of water vapor in upper-air atmosphere (less than 5%).Those atmospheric profile parameters, including temperature, altitude, relative humidity, and air pressure data, were adopted in the MODTRAN simulation to calculate the at-sensor radiances.In order to identify the uncertainties induced by the selection of atmospheric profile parameters, we compare the differences of the derived at-sensor BT using the two different atmospheric profiles, including atmospheric parameters from QFW6000 and NCEP reanalysis datasets.Fig. 7 showed that the differences between at-sensor radiance (W/m 2 •sr•μm) and at-sensor BT (K) of MERSI-LL bands 6 and 7 calculated using the atmospheric profile parameters from QFW6000 measurements and NCEP reanalysis data, and the results showed that the difference of at-sensor radiance ranged from 0.0003 to 0.065 W/m 2 •sr•μm with an rms of 0.043 W/m 2 •sr•μm for MERSI-LL band 6, while the difference of at-sensor radiance ranged from 0.002 to 0.101 W/m 2 •sr•μm with an rms of 0.068 W/m 2 •sr•μm for MERSI-LL band7.The difference of at-sensor BT ranged from 0.003 to 0.505 K with an rms of 0.339 K for MERSI-LL band 6, while the difference of at-sensor BT ranged from 0.027 to 0.931 K with an rms of 0.627 K for MERSI-LL band 7. Other studies also claimed that the variation of atmospheric profile has introduces errors in the at-sensor radiance propagation, such as a 10% change of water vapor profile caused at-sensor BT bias of 0.1 K for MODIS channel 31, and a 1.0 g/m3 variation of water vapor resulted in 1 and 0.6 K BT bias in FY3D channel 24 and channel 25, respectively [35], [37].Water vapor absorption distributed at TIR spectrum is the most important factor in altering the calibration accuracy during experiment analysis, and the real-time atmospheric profiles could provide the atmospheric conditions for the atmospheric radiative transfer simulation.Our study has adopted the atmospheric profile parameters from QFW6000 with high temporal resolution (2-min interval), and it could well monitor the near surface atmospheric conditions change.
3) Uncertainty Induced by Spectral Mismatch: In order to quantify the possible uncertainty induced by the spectral mismatch between satellite TIR sensor and field TIR sensor, we mainly compared and analyzed the difference of the simulated at-sensor BT before and after spectral matching (see Fig. 8).Generally, the results showed that the spectral mismatch between CE312 and MERSI-LL caused the difference of at-sensor radiance ranged from 0.050 to 0.055 W/m 2 •sr•μm with an rms of 0.016 W/m 2 •sr•μm for band 6, while the range of the difference would increase to 0.347 to 0.365 W/m 2 •sr•μm with an rms of 0.111 W/m 2 •sr•μm for band7.Therefore, the difference of at-sensor BT for band 6 ranged from 0.397 to 0.425 K with an rms of 0.411 K, while it increased to 3.086 to 3.281 K with an rms of 3.192 K for band 7. The results indicated that the spectral matching process is very important for radiometric calibration analysis, especially for those band pairs used for calibration analysis with obvious disagree in the spectrum range, such as the MERSI-LL band 7 and CE312 band 1 (see Fig. 2).
In fact, there are some details needed to be paid attention during the calibration analysis.The selection of the concurrent pixels from TIR images during the satellite overpass is also important to derive the reasonable satellite-based BT.Due to the limitation of field campaigns experiments and weather conditions, calibration experiments may experience some weak climate, such as part of the calibration site influenced by the cloud coverage.These situations make it difficult to select and calculate the at-sensor radiance.Previous study found that the derived BT have a decrease trend when the selected pixels near the cloud edge, and 5-8 pixels would be still contaminated by clouds with lower temperatures than normal clear sky conditions from TIR images of FY4A/AGRI [23].

IV. CONCLUSION
We investigated the absolute radiometric calibration condition of FY3E/MERSI-LL thermal IR band data using field measurements conducted at Lake Erhai on 5-23 December, 2022.Lake surface skin temperature records were continuously collected by a USV equipped with TIR radiometers.Atmospheric parameters and atmospheric profiles were measured near calibration experiment site, and then, they were adopted in the atmospheric radiative transfer code to calculate the absorption of atmosphere on longwave radiation signals propagated from land surface targets to satellite instrument entrance aperture.Generally, ten calibration experiments during the field campaign indicated the calibration accuracy of MERSI-LL bands 6 and 7 reached 0.71 K with an RMSE of 0.79 and 0.11 K with an RMSE of 0.47 K, respectively.MERSI-LL TIR bands have a good functional status due to the absolute radiometric accuracy specified for MERSI-LL bands achieved in bands 6 and 7 with calibration error less than 1 K. Uncertainties may be introduced in the ground-based radiometric calibration analysis, and atmospheric conditions and spectral matching process played more important roles than other factors.Our study also suggests that Lake Erhai has more homogeneous distribution of water temperature in the afternoon than that in the morning, and field experiments conducted at dawn may be not appropriate for radiometric calibration analysis.In order to clearly understand the possible uncertainty of in-orbit TIR sensors, more filed experiments are needed to conduct to collect the reasonable data samples to monitor the operational status of satellite instruments.

Fig. 1 .
Fig. 1.Location of calibration site and the navigation route of USV during field experiments.

Fig. 4 .
Fig. 4. Atmospheric profile parameters, mainly (a) air temperature and (b) relative humidity used in the atmospheric radiative transfer simulation.

Fig. 5 .
Fig. 5. BT change observed by CE312 deployed on the USV during field experiments.Here, the observed records on 12 December, 2022, were selected.

TABLE I SPATIAL
AND SPECTRAL SPECIFICATION OF FY3E/MERSI-LL

TABLE II SELECTED
RADIOMETRIC CALIBRATION EXPERIMENTS DURING FIELD CAMPAIGN