Open Access
Add to BibSonomyAbstract
Net surface shortwave radiation (NSSR) significantly affects regional and global climate change, and is an important aspect of research on surface radiation budget balance. Many previous studies have proposed methods for estimating NSSR. This study proposes a method to calculate NSSR using FY-2D short-wave channel data. Firstly, a linear regression model is established between the top-of-atmosphere (TOA) broadband albedo (r) and the narrowband reflectivity (ρ1), based on data simulated with MODTRAN 4.2. Secondly, the relationship between surface absorption coefficient (as) and broadband albedo (r) is determined by dividing the surface type into land, sea, or snow&ice, and NSSR can then be calculated. Thirdly, sensitivity analysis is performed for errors associated with sensor noise, vertically integrated atmospheric water content, view zenith angle and solar zenith angle. Finally, validation using ground measurements is performed. Results show that the root mean square error (RMSE) between the estimated and actual r is less than 0.011 for all conditions, and the RMSEs between estimated and real NSSR are 26.60 W/m2, 9.99 W/m2, and 23.40 W/m2, using simulated data for land, sea, and snow&ice surfaces, respectively. This indicates that the proposed method can be used to adequately estimate NSSR. Additionally, we compare field measurements from TaiYuan and ChangWu ecological stations with estimates using corresponding FY-2D data acquired from January to April 2012, on cloud-free days. Results show that the RMSE between the estimated and actual NSSR is 48.56W/m2, with a mean error of −2.23W/m2. Causes of errors also include measurement accuracy and estimations of atmospheric water vertical contents. This method is only suitable for cloudless conditions.
© 2016 Optical Society of America
1. Introduction
Net surface shortwave radiation (NSSR) is defined as the difference between total solar radiation and the radiation of surface reflection, which is the main part of surface net radiation and the main source of surface energy, as well as an important driving force in energy exchange between the land and the atmosphere. Accurately estimating NSSR is important for research on the surface radiation budget balance, the distribution and change of surface energy, regional and global climate change and estimations of land surface temperature under cloudy sky conditions [1–4].
NSSR retrieval through remote sensing was common in the 1980s. Pinker and Corio (1984) made direct estimations of NSSR using NOAA5 satellite observation data, and showed that a high correlation exists between top-of-atmosphere (TOA) and ground level NSSR observations [5]. Some research also estimated NSSR using geostationary satellite data [6,7]. Since then, most studies have focused on ground NSSR retrieval using the TOA NSSR and air/ground parameters. Cess and Vulis (1989) found that a linear relationship exists between TOA and surface NSSR [8]. Furthermore, this linear relationship is stable under different atmospheric conditions and solar zenith angles (SZA), so it can be used to estimate NSSR. Cess et al. (1991) further analyzed the other factors affecting this linear relationship, and illustrated that clouds related to atmospheric absorption affect the linear relationship, resulting in errors [9]. Li et al. (1993a) proposed a new method for estimating NSSR directly, using SZA and planet albedos, under different atmospheric conditions [10]. However, this method did not consider the effects caused by surface pressure, ozone, the amounts and types of aerosol, or the heights and types of cloud. Masuda et al. (1995) considered these factors in an improved model [11,12]. Hollmann et al. (2002) estimated NSSR by scanning radiation balance and using the method proposed by Li et al. (1993a) [13]. However, the results were poor due to the coarse resolution.
Considering that moderate-resolution imaging spectroradiometer (MODIS) provides higher spatial resolution images, progressively more studies are calculating net surface longwave radiation and NSSR using MODIS data [14–17]. Kim and Liang (2010) proposed a direct hybrid algorithm to estimate NSSR based on physical principles [18]. They improved the accuracy of the algorithm by adding elevation and water vapor to the model, resulting in the estimation of NSSR with a much higher spatial distribution.
With the launch of the FengYun (FY) series of Chinese geostationary meteorological satellites, the potential has arisen for the study of multi-temporal weather states and global climate changeby using FY-2D data to estimate NSSR. The objective of this study is to estimate NSSR using FY-2D VIS channel data. First, we introduce the data used in this study, and then the method used to estimate NSSR from FY-2D data is proposed in section 3. Sensitivity analyses are performed in section 4, the method is applied to actual FY-2D data in section 5, and finally we present our conclusions.
2. Data
In this paper, three types of data are used to build the model and validate the method. The first is simulated data and the others are VIS data from FY-2D, which is the Chinese first generation geostationary meteorological satellite, and in situ data from both TaiYuan XiaoDian meteorological station, located in Shanxi Province, and ChangWu ecological experiment station, located in Shaanxi province.
2.1 Simulated data
It is important to simulate radiative transfer for a wide range of atmospheric and surface conditions [19]. In this work, data was simulated using an atmospheric radiative transfer model (MODTRAN4), under six standard atmospheric model conditions, provided by MODTRAN4 (tropical atmosphere, middle latitude summer atmosphere, mid latitude winter atmosphere, subarctic summer atmosphere, subarctic winter atmosphere, and United States 1976 standard atmosphere), and nine types of surface cover including soil, vegetation canopy, grassland, wetland, city, desert, ocean surface, new snow and sea ice.In addition, different geometric observation angles are taken into consideration, including seven viewing zenith angles (VZAs, 0°, 15°, 30.37°, 39.79°, 45.56°, 49.5°, 52.37°), three relative azimuth angles (RAAs, 0°, 60° and 120°), and seven solar zenith angles (SZAs, 0°, 10°, 20°, 30°, 40°, 50° and 60°). Meanwhile, four types of aerosol (countryside, oceans, cities and troposphere) and four different visibilities (10, 15, 23, 30 km) are also considered. In total, 56448 different cases are simulated. The simulated radiances are integrated to TOA radiance (L1) in the FY-2D VIS channel, using the corresponding spectral response curve, and upward broadband fluxes are simultaneously calculated.
2.2 Satellite data
In this study, FY-2D data are downloaded from the China Meteorological website (http://satellite.cma.gov.cn/PortalSite/Data/Satellite.aspx). FY-2D, a geostationary meteorological satellite developed by Shanghai Academy of Space Flight Technology and China Academy of Space Technology, was launched on 8 December 2006 and is located above the Equator at longitude 86.5° E and a distance of 35,800 km. The satellite aims to provide meteorological information for the Asia-Pacific region. The upgraded Stretched-Visible and Infrared Spin-Scan radiometer (S-VISSR) is the main sensor loaded on the satellite, and it can acquire one full disc image covering the Earth’s surface, from 60°N to 60°S latitude and 45°E to 165°E longitude, taking 30 min per acquisition during the flood season. S-VISSR consists of 5 channels, including a VIS channel and 4 infrared channels (Table 1). The downloaded disc image file provides radiances in the VIS channel. In addition, latitude, longitude, view zenith angle (VZA), solar zenith angle (SZA), and relative azimuth angle (RAA) are also provided. Considering the method is only suitable for clear sky conditions, cloud cover is also used to set the conditions.
Table 1. Main technical index of FY2-VISSR radiometer
2.3 Field measurement data
To validate the proposed method, field measurements from grassland were collected from TaiYuan XiaoDian meteorological station (elevation 780m), located in Shanxi Province (112°33′E and 37°47′N). Incoming and reflected solar radiation was measured by TBQ-2. Data for January 24, February 7, 16 and 17, March 24 and April 25, 2012, collected at 1-h intervals, were used for validation in this study. The other in situ data were collected from the ChangWu ecological station located in Shaanxi Province (107°40′E and 35°12′N), which joined the Chinese Ecosystem Research Network in 1991. The site is wheat and corn dominated field with an elevation of 1300m, located in the vicinity of the meteorological station. Incoming and reflected solar radiation were measured by CM11 and CM6B. Data was collected at 1-h intervals on February 16 and 17, March 24 and April 25, 2012.
3. Method
3.1 Transformation from narrowband reflectivity to broadband albedo
Tang et al. (2006) originally developed a narrowband-to-broadband albedo conversion model, which links the narrowband apparent reflectance at the top of the atmosphere (TOA ρ1) to shortwave broadband albedo (r) for clear and cloudy skies [15]. For FY-2D data, TOA r can be obtained using the following equation (Eq. (1)):
where r is defined bywhere E0 is the TOA solar irradiance at one astronomical unit (W/m2), θs is SZA, d is Earth–Sun distance in a stronomical units, and Fu is the TOA upward flux (W/m2).ρ1 represents the TOA reflectivity of the VIS band, which can be estimated on the basis of simulated TOA radiances (L1), as described in section 2.1,using Eq. (3):
Where Ēband_i is the mean exoatmospheric irradiance of the VIS band and µs is the cosine of the SZA. Parameter r is obtained from simulated upward fluxes using Eq. (2). The coefficients in Eq. (1) can be estimated using the minimization procedure method based on ρ1 and r, mentioned above.According to Tang et al.(2006) [15], the coefficients in Eq. (1) are related to VZA and SZA, therefore a relationship is established from the simulations between coefficient bi and viewing zenith angle, for a given SZA and RAA, using the least squares method. The formula is as follows:
where c1i~c3i are constants for a given SZA and RAA. A very good fitting result for SZA = 0° is shown in (Fig. 1). The correlations are over 0.98 in all cases and most are larger than 0.99.
Fig. 1 The relationship between coefficient bi (i = 0,1) in Eq. (1) and VZAs for SZA = 0° and RAA = 60°.
Meanwhile, comparison of the actual total TOA shortwave albedos with those estimated using Eq. (1), based on the conversion coefficients b0 and b1, is shown in Fig. 2. The root mean square error (RMSE) of 0.0069 is relatively small and scattered points are distributed near the 1:1 line, indicating a good fit.
Fig. 2 Comparison of the actual TOA total shortwave broadband albedos with those estimated using Eqs. (1) and (4) under different conditions.
3.2 Retrieval of NSSR
Li et al. (1993, 1995) [10, 11] pointed out that the surface absorption coefficient (as), which is defined as the ratio of the net surface absorbed radiation to the outgoing flux at the TOA, can be expressed as a linear relationship with broadband albedo (r):
where,The parameters in Eq. (6) are the same as in Eq. (2). The fitting coefficients α’ and β’ can be expressed as:
where: µ = cos(SZA) and ω is atmospheric water vertical content (g/cm2), which is 4.11, 2.92, 0.85, 2.08, 0.42 and 1.42 g/cm2, respectively, for the 6 standard atmospheric profiles used in the simulation. On the basis of the simulated TOA upward flux, as can be calculated using Eq. (6) and r is determined as described in section 3.1. Parameters x, y, z, and a1~a7 can be determined by solving Eqs. (7) and (8), using nonlinear least square fitting for the given SZA.In the process of fitting the 7 parameters we categorize the surfaces into three groups: land surface, ocean surface and snow&ice surface. Coefficients for the land surface type are obtained using 42336 pairs of as and r simulated from six atmospheric profiles, six land surfaces, three aerosols, four types of visibility, seven SZAs, seven VZAs, and three RAAs. Similarly, different sets of coefficients are computed for the other surfaces.
NSSR can be estimated using Eqs. (1) and (4)-(8) with coefficient ci and the seven parameters. Comparison of actual and estimated NSSR is shown in Fig. 3. Along with Table 2, this shows that the RMSE varies from 9.99 W/m2 to 26.60 W/m2 with the different surface types. Among the three types, the minimum RMSE occurs with an ocean surface and the maximum RMSE occurs with a land surface. Most scatter points are near the 1:1 line, especially for the ocean surface, but a few points lie far from the 1:1 line for land and snow&ice surfaces, resulting in larger RMSEs. Due to the complex nature of the land surface, the error range, which varies from −36.97 to 147.01 W/m2, is largest for this surface. It is smallest for the ocean, ranging from −26.52 to 29.58 W/m2. However, the absolute mean error is smallest for the land surface, at 0.34 W/m2.
Table 2. Errors of estimated NSSR and actual NSSR (W/m2)
4. Sensitivity analysis
NSSR values can be obtained using TOA ρ1, w, RAA, VZA, and SZA. The accuracy of ρ1 depends on, among other things, instrument calibration and navigation. Furthermore, w estimated using remote sensing data, as well as SZA and VZA values, can all contain some bias. As a result, the estimation errors on NSSR, introduced by several error sources, must be taken into account and evaluated.
4.1 Sensitivity analysis of water vertical content
Regarding the errors in NSSR introduced by atmospheric water vertical content (w), sensitivity analysis is performed by multiplying w by 0.9, 0.95, 1.05, or 1.1 i.e. a percentage error of −10, −5, 5, or 10%. Table 3 gives the RMSEs and mean errors of the estimated and measured NSSR, including w error, for the different surface type. Figure 4 shows the uncertainty in NSSR retrieval values as a result of the same errors in w. It shows that NSSR uncertainty introduced by the w errors is generally smallest for the land surface type. In Fig. 4, we found that RMSEs increase with absolute percentage error in w for both land and snow&ice surfaces, whilst RMSEs decrease with absolute percentage error in w for ocean surfaces.
Table 3. RMSEs of the estimated and measured NSSR, including w errors, for different surface types (W/m2), Mean errors are shown in brackets
Fig. 4 Difference between estimated and real NSSR after including w errors for Land, Ocean, and Snow&ice surfaces.
4.2 Sensitivity analysis of TOA narrowband reflectivity
Noise from the TOA reflectivity sensor is less than 0.02 for FY-2D. NSSR uncertainty due to TOA reflectivity errors is investigated by adding −0.02, −0.01, 0.01, and 0.02 to the measured TOA reflectivity. Table 4 gives the RMSEs and mean errors of the estimated and measured NSSR, after including TOA reflectivity errors, for the different surface types. Analysis of the mean NSSR errors shows that most NSSR values are overestimated when the TOA reflectivity error is negative, and on the contrary, most NSSR values are underestimated when the TOA reflectivity error is positive. Figure 5 illustrates the uncertainty in NSSR retrieval caused by errors due to reflectivity noise. It shows that the uncertainty caused by reflectivity errors has the smallest influence on the land surface type. Additionally, RMSEs of NSSR increase with an increase in TOA reflectivity absolute error.
Table 4. RMSEs of the estimated and measured NSSR after adding reflectivity noise for different surface types (W/m2). Mean errors are shown in brackets.
Fig. 5 Difference between estimated and real NSSR after including real TOA reflectivity errors for Land, Ocean, and Snow &ice surfaces.
4.3 Sensitivity analysis of SZA
Considering the errors in NSSR introduced by SZA, sensitivity analysis is performed by adding −20°, −10°, 10° and 20° to the actual SZA. Table 5 shows the RMSEs and mean errors of the estimated and measured NSSR after adding errors in SZA for the different surface types. We find that NSSRs are typically overestimated when the SZA is lower than the actual SZA. On the contrary, most NSSRs are underestimated when SZA is greater than the actual SZA. Figure 6 shows the uncertainty in the retrieved NSSR as a result of SZA error. It shows that when SZA is lower than the actual SZA, the uncertainty introduced by SZA errors is less than when SZA is greater than actual SZA. Furthermore, the RMSEs increase with an increase in SZA absolute error.
Table 5. RMSEs of the estimated and measured NSSR with SZA errors included for different surface types (W/m2). Mean errors are shown in brackets.
Fig. 6 Difference between estimated and real NSSR after including SZA errors for Land, Ocean, and Snow&ice surfaces.
4.4 Sensitivity analysis of VZA
In order to analyze the errors in NSSR values introduced by VZA, sensitivity analysis is performed by adding −10°, −8°, −6°, −4°, −2°, 2°, 4°, 6°, 8° and 10° to the actual VZA. Table 6 gives the RMSEs and mean errors of the estimated and measured NSSR, including VZA error, by different surface type. For land and ocean surfaces, most NSSRs are overestimated whether the VZA is bigger or smaller than the actual VZA. Figure 7 shows the uncertainty in the retrieved NSSR introduced by errors in VZA. It shows that the uncertainty introduced by VZA errors has less influence on land surfaces than on ocean and snow&ice surfaces. Additionally, the RMSEs increase with an increase in VZA absolute error for snow&ice surfaces.
Table 6. RMSEs of the estimated and measured NSSR with VZA errors for different surface type (W/m−2). Mean errors are shown in brackets.
Fig. 7 Difference between estimated and real NSSR after including VZA errors for Land, Ocean, and Snow&ice surfaces.
5. Application to FY-2D data
5.1 FY-2D data processing
FY-2D daytime data were downloaded for January 24, February 7, February 16, February 17, and April 25, 2012, corresponding to the dates of field measurements. As this method is only suitable for clear sky conditions, pixels are deleted if the cloud cover is recorded as larger than 0.1.
5.2 Determination of the atmospheric water vertical content
According to Li et al. (2003) [20], the atmospheric water vertical content (WVC) can be derived from the transmittance ratio of split-window channels (assuming emissivity of the two channels isidentical), calculated by:
andwhere c1 and c2 can be derived as functions of VZA, using Eqs. (12) and (13), as proposed by Tang et al. (2015) [21]. Parameters i and j are the split-window channels, the subscript k denotes pixel k, and T refers to the TOA mean (or median) channel brightness temperatures of the N neighboring pixels considered for channels i and j, respectively.5.3 Results of the NSSR
The objective of this work is to estimate instantaneous NSSR using the method proposed in section 3. To validate our proposed method, in situ measurements were made, described in section 2.3.
Comparison of measured NSSR with calculated NSSR is shown in Fig. 8. For the TaiYuan XiaoDian meteorological station RMSE is 48.56 W/m2, with a mean error of −2.23 W/m2. For the ChangWu ecological station RMSE is 74 W/m2, with a mean error of 13.6 W/m2. Points are distributed closely around the 1:1 line, indicating a good fit. As discussed in section 4, the estimation error of atmospheric water vertical content and the noise in TOA reflectivity can cause error in the NSSR calculation. Additionally, error in NSSR can also be caused by the algorithm and field measurements. Figure 9 shows an example of retrieved NSSR in the Beijing area, China, during FY-2D satellite scanning on January 24, 2012 at 14:30 local time. The white regions do not show any data due to cloud cover.
Fig. 9 Retrieved NSSR from the Beijing area in China, during FY-2D satellite scanning on January 24, 2012, at 14:30 local time.
6. Conclusion
In this study, we propose a simple statistical regression method for estimating instantaneous net short solar radiation (NSSR) using FY-2D data. Results from the simulation analysis show that NSSR can be estimated using the proposed method with a root mean square error (RMSE) of 26.60 W/m2, 9.99 W/m2, and 23.40 W/m2 for land, sea and snow&ice surfaces, respectively.
This method uses top-of-atmosphere reflectivity in the VIS band and atmospheric water vertical content (WVC). Considering the errors in estimated NSSR due to reflectivity noise and errors in WVC and view/solar zenith angle (VZA/SZA), a sensitivity analysis regarding uncertainty introduced by these errors was performed. Results show that the accuracy of NSSR retrieval can be altered by 0.24, 0.039, 0.098, and 0.31 W/m2 for land surfaces, when adding −10, −5, 5, and 10% error, respectively, to measured WVC. Uncertainty introduced by reflectivity errors is approximately 5.64, 1.45, 1.71, and 6.08 W/m2 for land surfaces, when adding errors of −0.02, −0.01, 0.01, and 0.02, respectively, to measured reflectivity values. The influence of VZA errors results in uncertainties of less than 3 W/m2 for land surfaces.
In order to validate the retrieval accuracy of the proposed method, we used field measurements made at a meteorological station in TaiYuan XiaoDian district, Shanxi province and ChangWu ecological station, Shaanxi province, between January and April 2012, under cloud-free conditions. RMSEs of the calculated and measured NSSR are 48.56 W/m2, with a mean error of −2.23W/m2, for the TaiYuan XiaoDian meteorological station, and 73.9 W/m2, with a mean error of 13.6 W/m2, for the ChangWu station. The narrow distribution of points around the 1:1 line indicates goodness of fit.
It should be noted that the method was only validated for mid-latitude regions under clear sky conditions. Errors in NSSR can also be caused by the algorithm accuracy, uncertainty in atmospheric water vertical content, and noise in the FY-2D top-of-atmosphere reflectivity signal and solar/view zenith angle errors et al.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (NSFC) under Grant 41271381 and 41161066. We also thank the National Satellite Meteorological Centre for providing the FY-2D data and ChangWu ecological station for providing the field data.
References and links
1. S. B. Duan, Z. L. Li, N. Wang, H. Wu, and B. H. Tang, “Evaluation of six land-surface diurnal temperature cycle models using clear-sky in situ and satellite data,” Remote Sens. Environ. 124, 15–25 (2012). [CrossRef]
2. S. B. Duan, Z. L. Li, B. H. Tang, H. Wu, and R. Tang, “Generation of a time-consistent land surface temperature product from MODIS data,” Remote Sens. Environ. 140, 339–349 (2014). [CrossRef]
3. S. B. Duan, Z. L. Li, B. H. Tang, H. Wu, and R. Tang, “Direct estimation of land-surface diurnal temperature cycle model parameters from MSG-SEVIRI brightness temperatures under clear sky conditions,” Remote Sens. Environ. 150, 34–43 (2014). [CrossRef]
4. X. Y. Zhang, J. Pang, and L. L. Li, “Estimation of land surface temperature under cloudy skies using combined diurnal solar radiation and surface temperature evolution,” Remote Sens. 7(1), 905–921 (2015). [CrossRef]
5. R. T. Pinker and L. Corio, “Surface radiation budget from satellites,” Mon. Weather Rev. 112(1), 209–215 (1984). [CrossRef]
6. R. T. Pinker and J. A. Ewing, “Modeling surface solar radiation: model formulation and validation,” J. Clim. Appl. Meteorol. 24(5), 389–401 (1985). [CrossRef]
7. R. T. Pinker and J. D. Tarpley, “The relationship between the planetary and surface net radiation: an update,” J. Appl. Meteorol. 27(8), 957–964 (1988). [CrossRef]
8. R. D. Cess and I. L. Vulis, “Inferring surface-solar absorption from broadband satellite measurements,” J. Clim. 2(9), 974–985 (1989). [CrossRef]
9. R. D. Cess, E. G. Dutton, J. J. DeLuisi, and F. Jiang, “Determining surface solar absorption from broadband satellite measurements for clear skies: comparisons with surface measurements,” J. Clim. 4(2), 236–247 (1991). [CrossRef]
10. Z. Q. Li, H. G. Leighton, K. Masuda, and T. Takashima, “Estimation of shortwave flux absorbed at the surface from TOA reflected flux,” J. Clim. 6(2), 317–330 (1993). [CrossRef]
11. K. Masuda, H. G. Leighton, and Z. Q. Li, “A new parameterization for the determination of solar flux absorbed at the surface from satellite measurements,” J. Clim. 8(6), 1615–1629 (1995). [CrossRef]
12. W. B. Rossow and Y. C. Zhang, “Calculation of surface and top-of-atmosphere radiative fluxes from physical quantities based on ISCCP data sets. 2: validation and first results,” J. Geophys. Res. 100(D1), 1167–1198 (1995). [CrossRef]
13. R. Hollmann, A. Bodas, A. Gratzki, K. Dammann, and R. Stuhlmann, “The surface shortwave net flux from the scanner for radiation budget (SCARAB),” Adv. Space Res. 30(11), 2363–2369 (2002). [CrossRef]
14. Z. Li and A. Trishchenko, “A Study toward an improved understanding of the relationship between visible and shortwave measurements,” J. Atmos. Ocean. Technol. 16(3), 347–360 (1999). [CrossRef]
15. B. H. Tang, Z. L. Li, and R. H. Zhang, “A direct method for estimating net surface shortwave radiation from MODIS data,” Remote Sens. Environ. 103(1), 115–126 (2006). [CrossRef]
16. B. H. Tang and Z. L. Li, “Estimation of instantaneous net surface longwave radiation from MODIS cloud-free data,” Remote Sens. Environ. 112(9), 3482–3492 (2008). [CrossRef]
17. J. Wang, B. H. Tang, X. Y. Zhang, H. Wu, and Z. L. Li, “Estimation of surface longwave radiation over the Tibetan plateau region using MODIS data for cloud-free skies,” IEEE J. Sel. Top. Appl. Earth Observ. Rem. Sens. 7(9), 3695–3730 (2014).
18. H. Y. Kim and S. L. Liang, “Development of a hybrid method for estimating land surface shortwave net radiation from MODIS data,” Remote Sens. Environ. 114(11), 2393–2402 (2010). [CrossRef]
19. G. H. Huang, S. M. Liu, and S. L. Liang, “Estimation of net surface shortwave radiation from MODIS data,” Int. J. Remote Sens. 33(3), 804–825 (2012). [CrossRef]
20. Z. L. Li, L. Jia, Z. B. Su, Z. M. Wan, and R. H. Zhang, “A new approach for retrieving precipitable water from ATSR2 split-window channel data over land,” Int. J. Remote Sens. 24(24), 5095–5117 (2003). [CrossRef]
21. B. H. Tang, K. Shao, Z. L. Li, H. Wu, F. Nerry, and G. Q. Zhou, “Estimation and validation of land surface temperature from Chinese second generation polar-orbiting FY-3A VIRR data,” Remote Sens. 7(3), 3250–3273 (2015). [CrossRef]
