Comparing different profiles to characterize the atmosphere for three MODIS TIR bands
Introduction
Surface radiance measurements taken by remote sensing instruments aboard satellites are affected by the atmosphere. In the particular case of the thermal infrared region (3–14 μm), there exists two specific spectral ranges located at 3.7–4.1 μm and 8–14 μm, where the atmosphere shows the lowest radiative effect, mainly due to the water vapor content (W).
The land surface temperature (LST) is a variable of great interest in numerous meteorological and climatological studies, and its accurate retrieval is of prime interest to estimate energy and water fluxes budgets between the surface and atmosphere (Sánchez et al., 2011, Sánchez et al., 2014). The main concerns in retrieving LST from satellite data are the atmospheric and emissivity correction (Jiang and Liu, 2014). Currently LST can be obtained from several algorithms dependent on the sensor specifications (Zhou et al., 2012). Two of these techniques are the single-channel (SC; Vlassova et al., 2014) and the temperature and emissivity separation (TES; Hulley et al., 2014) methods. Both algorithms derive the LST from inversion of the radiative transfer equation (RTE), which at the same time needs the previous knowledge of a characterized atmosphere, among other factors. The radiation measured with a radiometric sensor is composed of a double contribution: first, the radiation directly emitted by the surface, and secondly, the radiation reflected in the surface, coming from the surroundings and the atmosphere. In addition, if the measurement is taken from satellite, the atmosphere contributes in two different ways: on the one hand, absorbing part of the surface radiation and in the other, emitting radiation directly to the sensor. RTE connects the at-sensor radiation and the radiation emitted by the surface through an energy balance defined as:where τi is the atmospheric transmissivity at the spectral range i, εi is the surface spectral emissivity, Bi(T) is the Planck function for black body spectral radiance at temperature T, Li,hem↓ is the downwelling hemispheric radiance, Li↑(θ) is the upwelling path radiance at zenith angle θ, and LiTOA is the radiance measured by the sensor at the top of the atmosphere. All the surface and atmospheric factors exposed in Eq. (1) can be found explicitly defined in the reviewing publication of García-Santos et al. (2010). The W present in the atmosphere affects directly to the value of these atmospheric parameters (τi, Li,hem↓, Li↑(θ)).
Nowadays, the most suitable way to characterize the atmosphere is by using a radiative transfer code (RTC; Berk et al., 2011), which calculates the atmospheric factors from introducing vertical profiles of pressure, air temperature and humidity at different levels of altitude. Probably, the best representative profile of the corresponding atmosphere is obtained from radiosounding data, acquired with a launched balloon. However, this data is rarely available at the time and location of the measurements acquisition. As an alternative, there exists the possibility to obtain an atmospheric profile derived from the spectral features of overpassing satellite sensor as well as with interpolating models (in space and time) based on radiosounding data acquired close to a selected point (Jiménez-Muñoz et al., 2010).
The objective of this paper is to analyze differences on the atmospheric variables, τi, Li,hem↓, Li↑(θ), after applying on MODTRAN RTC (Berk et al., 2011) different vertical profiles of cloudless days. These profiles are obtained from three different sources: a) modeled MODIS spectral measured radiances (MOD07; Borbas et al., 2011) profiles, b) spatial and temporal interpolated National Centers for Environmental Prediction (NCEP) atmospheric profiles (Barsi et al., 2005), c) radiosounding data measured by balloons launched by AEMET and showed on the web of the Department of Atmospheric Science in the University of Wyoming (WYO).
In previous studies carried out by Coll et al. (2012) and Li et al. (2013), the LST obtained from satellite data using SC methods were compared with in situ ground LST measurements, after applying different atmospheric profiles obtained from the NCEP and the MOD07 product. In both cases, these studies obtained good results for the LST comparison, but MOD07 introduced greater errors in the LST retrievals (± 1.0 K for NCEP and MOD07 in Coll et al., 2012 and ± 1.1 K for NCEP and ± 1.2 K for MOD07 in Li et al., 2013). This paper faces the same objective pursued by Coll et al. (2012) and Li et al. (2013) but in a different and profuse manner. The most important contribution in this paper is that the comparison of atmospheric parameters obtained from NCEP and MOD07 profiles with in situ radiosounding data has been done for a period of 4 years (from 2010 to 2013), and for three different sites with differences in the heights above sea level as well as in the distances from the corners used by NCEP to interpolate the profiles, yielding to more representative statistical results. In addition, results and statistics from MODIS band 29 (8.4–8.7 μm) are included. This band, which is not analyzed before in the other papers, is added to the study. It is important an accurate atmospheric correction between 8.4 and 8.7 μm, which is required in mineralogical and geological research purposes. For instance, mapping geological presence of minerals, like quartz, based on measurements in the reststrahlen region 8–9 μm, where the emissivity of quartz decreases in a very pronounced manner.
Section 2 describes the methodology applied to reach the fixed objective. Section 3, shows the results obtained and the corresponding discussions. A simulation study was carried out in Section 4 to evaluate the effect in terms of LST when applying SC method with NCEP and MOD07 profiles if in-situ radiosounding data is not available. Finally, the main conclusions of the study are given in Section 5.
Section snippets
Site
The three selected sites for the study are located at Spain (Murcia, Zaragoza and Madrid). They have been chosen because, from the different Wyoming radiosounding data available, they represent different altitudes above sea level.
The site situated at lesser height above the sea level is Murcia (62 m). The radiosoundings are launched at the Territorial Center of AEMET of Murcia, located at 3.5 km from the city (38°N, 1°9′W).
Another region analyzed is focused on the airport of Zaragoza (41°39′N,
Sensitivity analysis
The uncertainty for each of the retrieved parameters is calculated from the uncertainties associated to the different parameters of the profiles commented in Section 2.2. The process to calculate the uncertainty associated to the different parameters that characterize the atmosphere is explained. First, MODTRAN is run with the original profiles, then, MODTRAN is run again with profiles that include the original profiles plus the uncertainty associated to their parameters. Finally, the
Simulation study
In order to analyze the uncertainties in terms of temperature retrieved using the atmospheric parameters calculated in this paper, through different sources, a simulation study was carried out using the SC method (Vlassova et al., 2014). The procedure is as follows:
First, for a defined T (in this study three different temperatures were selected: 273 K, 293 K and 313 K) a Bi(T) is calculated and a LiTOA is retrieved trough Eq. (1) at the three MODIS bands 29, 31 and 32. Atmospheric parameters from
Conclusions
This work has analyzed the differences appeared when characterizing the atmosphere from different atmospheric profiles: NCEP, MOD07_L2 product and radiosoundings from the University of Wyoming (used as reference).
Atmospheric parameters in the TIR region 8–14 μm (upwelling radiance, downwelling hemispheric radiance and atmospheric transmissivity) were obtained for three thermal bands (29, 31 and 32) of sensor MODIS.
It is concluded that both NCEP and MOD07 profiles show an acceptable agreement
Acknowledgments
This study was supported by the Spanish Ministry of Economy and Competitiveness, through projects CGL2013-46862-C2-1-P and CGL2011-30433-C02-02. We also want to thank the Government of Generalitat Valenciana for the support this study through the project PROMETEOII/2014/086.
References (17)
- et al.
Comparison between different sources of atmospheric profiles for land surface temperature retrieval from single channel thermal infrared data
Remote Sens. Environ.
(2012) - et al.
Thermal-based techniques for land cover change detection using a new dynamic MODIS multispectral emissivity product (MOD21)
Remote Sens. Environ.
(2014) - et al.
Atmospheric correction of optical imagery from MODIS and Reanalysis atmospheric products
Remote Sens. Environ.
(2010) - et al.
Synoptic environment, mesoscale configurations and forecast parameters for hailstorms in Southwestern Europe
Atmos. Res.
(2013) - et al.
Assessing crop coefficients of sunflower and canola using two.source energy balance and thermal radiometry
Agric. Water Manag.
(2014) - et al.
Validation of a Web-Based Atmospheric Correction Tool for Single Thermal Band Instruments
- et al.
MODTRAN®5.2.1 USER’S MANUAL
(2011) - et al.
Modis Atmospheric Profile Retrieval Algorithm Theoretical Basis Document, Collection 6, Version 7
(2011)
Cited by (21)
Combined land surface emissivity and temperature estimation from Landsat 8 OLI and TIRS
2020, ISPRS Journal of Photogrammetry and Remote SensingSimultaneous inversion of multiple land surface parameters from MODIS optical–thermal observations
2017, ISPRS Journal of Photogrammetry and Remote Sensing