Assessment of Retrieved GMI Emissivity over Land, Snow, and Sea Ice in the GEOS System
1. Introduction
Measurements from microwave sounders and imagers provide a valuable source of information including atmospheric temperature and water vapor in numerical weather prediction (NWP) systems that assimilate these observations directly over water surfaces (oceans and other large water bodies). In a recent decadal survey, targeted observables in the planetary boundary layer (PBL) were cited as a key need for future observations (NASEM 2018). Microwave instruments, which sense in the PBL are currently available; however, utilizing surface-sensitive microwave observations for atmospheric data assimilation remains a challenge over land, snow, and sea ice. This is in part due to the inability of surface emissivity models used by NWP data assimilation systems to simulate observations with sufficient accuracy. The GEOS-ADAS (Todling and el Akkraoui 2018), which utilizes the Community Radiative Transfer Model (CRTM) (Han 2006; Chen et al. 2008), is no exception.
The ECMWF system has retrieved instantaneous surface emissivity from surface-sensitive channels for SSMIS and MHS radiance observations, and applies these estimates to the closest channels higher in frequency (Baordo and Geer 2016) in the calculation of simulated radiances. This approach currently is also being tested in the GEOS-ADAS for AMSU-A and ATMS radiances (Zhu et al. 2021). No or minimal emissivity spectral variability has been assumed in the abovementioned studies. This is generally supported by retrievals in Munchak et al. (2020, hereby referred to as M2020) where the spectral variability was nominally 0.03 between 89 and 166 GHz except over shallow snow cover. Recently, work by M2020 provided a new database for emissivity over land, snow, and sea ice retrieved from the NASA Global Precipitation Measurement (GPM) mission. Compared with Tool to Estimate Land Surface Emissivities at Microwave (TELSEM2; Wang et al. 2017), M2020 provides emissivities for more frequencies (i.e., 10.7 GHz V/H). Moreover, this database is unique in that it utilizes both active and passive data to retrieve surface emissivity and normalized radar cross section. While the emissivity values may be useful for other sensors, they are most applicable to the GPM Microwave Imager (GMI). In this work the GEOS-ADAS is modified to utilize emissivity values from M2020 in place of values used by CRTM. Presently, only GMI radiances over ocean are used in the operational GEOS-ADAS. This study will focus on the GMI radiances over land, snow, and ice, as a first attempt to evaluate GMI radiances over these nonwater surface types. Two cases are then presented, one with 1 week of observation minus background departures using the modified GEOS-ADAS, and one utilizing the original GEOS-ADAS. It should be noted that the surface emissivity models in CRTM are not state of the art and are scheduled to be replaced by the Community Surface Emissivity Module (CSEM; Chen and Weng 2016). Simulations using default CRTM emissivity values are used merely as reference comparing against M2020, and is not a thorough comparison against other more state of the art modules such as CSEM.
This paper is organized as follows: First, in section 2 the microwave emissivity databases and models used are discussed; next, in section 3 an evaluation of observation minus background values produced by the GEOS-ADAS are presented. Finally, the implications for assimilating GMI surface sensitive channels and the potential benefits in NWP including better representation of the PBL are discussed.
2. Microwave surface emissivity databases and models
There has been significant progress made in the last few years with respect to microwave emissivity databases. Some recently developed packages include TELSEM2 and CSEM (Chen and Weng 2016). TELSEM2 is currently distributed with RTTOV (Saunders et al. 2020), while CSEM will soon be included with CRTM. The surface models (land, snow, and ice) currently available in the CRTM are the physical model LandEm (Weng et al. 2001) for land surfaces, LandEm or an empirical/semiempirical models for specific sensors over snow covered surfaces, or empirical/semiempirical model for specific sensors over ice surfaces. A database designed specifically for GMI has been recently made available in M2020. This database is unique in that it uses active microwave data from the Dual-Frequency Precipitation Radar (DPR) on GPM, along with passive microwave data from GMI. The details of both M2020 and emissivity models present in the CRTM’s default configuration, which are relevant to GMI are discussed.
a. Emissivity available in CRTM
The control case in this work utilizes the default emissivity models over land, snow, and sea ice available for GMI in the CRTM. As mentioned previously, the emissivity models utilized by the CRTM are sensor specific; however, in the case of GMI there are no derived empirical or semiempirical models for snow or sea ice. The LandEm physical model provides emissivity values for frequencies below 80 GHz over land and snow-covered surfaces. LandEm is a physical model derived from a three-layer radiative transfer model along with modified Fresnel equations at layer interfaces. The model uses several parameterizations, and variables obtained from the GEOS system such as leaf area index (LAI), snow depth, surface type, along with sensor view geometry parameters such as zenith angle. It was validated with available data at the time from ground-based measurements (Mätzler 1994), and satellite data from AMSU-A. For frequencies above 80 GHz, the CRTM will use a constant value of 0.95 for land and 0.9 for snow. Given there is no empirical model for sea ice available in the case of GMI, the CRTM will use a default value of 0.92. For convenience, the source of emissivity values for each GMI channel are summarized in Table 1. The nomenclature “V” refers to vertical polarization, while “H” refers to horizontal polarization.
The source of emissivity used by the CRTM for each surface type and GMI channel
b. GPM Microwave Imager
Data from the GMI are utilized both in M2020 and in this study. GMI is a 13 channel conically scanning microwave radiometer aboard the GPM mission. Launched in February 2014, GPM is in a non-sun-synchronous orbit with an inclination angle of 65°. The conical scan has a nominal Earth incidence angle of 52.8° for channels at or below 89 GHz and 49.2° for channels at or above 166 GHz (Skofronick-Jackson et al. 2017; Petty and Bennartz 2017). The conical scan geometry with a near-constant Earth incidence angle/zenith viewing makes it possible for more accurate emissivity databases specifically designed for GMI as in M2020, given there is an Earth incidence angle dependence on emissivity. The channels for GMI are shown in Table 1, and range in frequency between 10.6 and 183 GHz. All 13 channels are sensitive to precipitation. The lower-frequency channels (10.6–37 GHz) are most sensitive to emission from liquid precipitation, and higher frequencies (89–183 GHz) are most sensitive to ice scattering. Channels 12 (183 ± 3 GHz V) and 13 (183 ± 7 GHz V) are strongly sensitive to water vapor and have a relatively weak surface sensitivity outside polar regions and high elevations (i.e., anywhere there is a low concentration of water vapor). Additionally, the presence of ice sheets in such regions enhance surface sensitivity. In the absence of scattering due to clouds and precipitation, channel 5 (23 GHz) along with channels 10–11 (166 GHz V/H) also have sensitivity to water vapor in addition to surface sensitivity; however, this sensitivity is weaker than the 183 GHz channels. Channels 10–11 are typically considered surface sensitive when column water vapor is approximately less than 20 mm (M2020). Currently, in the GEOS-ADAS channels are only assimilated over water using 23 GHz, 37 GHz V, 166 GHz V, 183 ± 3 GHz V, and 183 ± 7 GHz V (Kim et al. 2020). In future assimilation work over land, snow, and ice, a different channel selection may be necessary. Any remaining emissivity uncertainty may result in large uncertainty in brightness temperature simulation, which may conflate signals from different geophysical parameters.
c. Emissivity from active–passive microwave land surface database
Recently, M2020 presented an active–passive microwave land surface database that includes monthly average emissivity values for GMI channels 1–11. The average emissivities were derived using 5 years of emissivity retrievals (March 2014–February 2019) using data from GPM, thus providing a climatological emissivity value on a monthly basis. The climatology is constructed taking each month of the year (January–December) and grouping the 5 years of retrievals by month. The data are provided using a 0.25° × 0.25° global (67°S–67°N) grid, with an average value of emissivity at each grid cell over land, snow, and ice (with a fill value for no retrieval). The dataset is unique in that it utilizes the DPR on GMI to both filter out precipitation-contaminated observations, along with retrieval diagnostics and ancillary data from MERRA-2 (Gelaro et al. 2017). The dataset also contains a surface classification based on the spectral emissivity and radar backscatter cross-section characteristics. The retrieval of emissivity uses GMI brightness temperatures taken from the level-1CR data product, surface normalized radar cross section (σ0) from the DPR level 2A data product, along with data from MERRA-2, which are used as the a priori atmospheric profiles and surface temperature for the retrieval of emissivity from brightness temperatures.
Figure 1 shows GMI observations simulated and compared in this study. It should be noted that there are no retrieved 166 GHz emissivities over portions of Africa and South America in the M2020 dataset due to insufficient sensitivity of this channel to the surface in these regions due to high water vapor amounts. To simplify the implementation, no GMI observations are considered over these regions (in white). The classification of land, snow, and ice are taken from the GEOS model surface type. The surface classification in GEOS is derived from the catchment based model of Koster et al. (2000) over land, the Multilayer snow model of Stieglitz et al. (2001), and the Operational Sea Surface Temperature and Ice Analysis system (OSTIA; Donlon et al. 2012).
Locations of all GMI observations over land, snow, and ice used in this study.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
3. Evaluation of active–passive microwave land surface database using GEOS
In this work the emissivity models available in the GEOS-ADAS are compared against the climatological emissivity values available in M2020. GMI has 13 channels, with varied sensitivity to the surface. The two water vapor channels at 183 GHz having less surface sensitivity than others outside polar regions, or regions with ice sheets. Currently, the GEOS-ADAS only assimilates 23 GHz, 37 GHz V, 166 GHz V, and two vertically polarized water vapor channels at 183 GHz over ocean. No GMI radiances over land, snow, and ice are used. This resulted in some slight modifications to the GEOS-ADAS along with some other quality control decisions. These are described in section 3a, and in section 3b simulated GMI observations using the default GEOS-ADAS emissivity models are compared against that of M2020.
a. Evaluation method using the GEOS-ADAS
The comparisons made in this study use two slightly modified versions of the GEOS-ADAS version 5.27.1. Most of the code changes are common among both systems. First, version 5.27.1 of the GEOS-ADAS rejects GMI observations over land, snow, ice, and mixed surfaces. This check is modified to only reject mixed surfaces for both cases. While having the ability to simulate mixed surfaces is desirable, as a first step to evaluate the emissivity database of M2020, it is best to compare surfaces without the complexity of accounting for surface cover fractions, which may introduce more representativeness error. Next, a quality control check that rejects GMI observations north of 55° latitude and south of −55° latitude is removed. This check was originally added to the GEOS-ADAS to avoid sea ice. A goal of this study is to investigate the ice emissivity available in M2020 and compare it to the default values in CRTM for comparison; therefore, the latitude check is removed for both simulations. Finally, to avoid the complexities of comparing regions affected by rain or clouds a check is added to flag GMI observations with a total sum of ice and liquid water content of 0.15 kg m−2 and remove them from consideration.
For the case using M2020, there were a few steps necessary to ingest emissivity into the GEOS-ADAS. First, a netCDF file was generated based on data provided in M2020. The emissivity dataset is a five-dimensional array indexed by channel, latitude, longitude, month, and surface type (land, snow, or sea ice). Each observation is interpolated using a nearest neighbor approach to the M2020 latitude and longitude grid. The month of the observation is used as an index along with the surface type given by the GEOS-ADAS. For the two 183 GHz channels, there are no retrieved values available from M2020. Instead, the emissivity values from the 166 GHz V channel are used to approximate the emissivity value. A similar assumption was made in Baordo and Geer (2016), using 89 GHz emissivity values to estimate 183 GHz emissivity for the SSMIS instrument.
Radiative transfer calculations are performed by running the GSI in a stand-alone mode in place of a full 4D-EnVar experiment. In stand-alone mode, the background fields are taken from an existing run of the GEOS-ADAS. The run is an experimental version of the GEOS-Forward Processing system using the version 5.27.1 of the GEOS-ADAS, which assimilates the full suite of infrared, microwave, and conventional observations including an all-sky assimilation of GMI over ocean surfaces. This allows a quick method to produce simulated observations that are effectively offline simulations for comparison of the emissivity models, and since the background fields are constant in the comparisons, the only differences are due to the changes in emissivity in the radiative transfer calculations. The background fields are on hourly intervals and are interpolated in space and time to the observation location. One week of GMI observations are simulated for 1–7 December 2020 for four synoptic time windows at 6 h intervals. In all comparisons to follow the observation minus background values are used to indicate whether the emissivity model improves the simulation. A smaller magnitude (absolute value) observation minus background indicates the simulation is closer, and therefore considered an improvement. In all comparisons, no bias correction is applied to the simulated values.
b. Results of evaluation
The simulated brightness temperatures (also commonly referred to as the background) using M2020 and using the default CRTM are compared against GMI observations for the first week of December 2020. In Fig. 2, scatterplots observation minus background (OMB) are shown for the 13 GMI channels. One feature that clearly stands out is ice (in orange) and snow surfaces (in blue) exhibit the largest scatter in OMB values in both in the M2020 simulation and default CRTM simulation relative to land (in green). Next, departures over land are significantly smaller for M2020 in Figs. 2a–g. Similarly, large departures of OMBs over ice are reduced by using M2020 in place of the default CRTM value of 0.92 emissivity. This is especially true for horizontal polarization channels (shown in Figs. 2b,d,g,i). Finally, it is clear the default CRTM exhibits a bias over ice in Figs. 2a–i and over snow in Figs. 2h–m (values not centered around zero), whereas in the M2020 simulation the values are closely centered around zero. For reference, all observation points and their classification as either over land, snow, or ice are shown in Fig. 1.
Scatterplots of GEOS-ADAS calculated values of observation minus background using M2020 vs default CRTM values.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
Next, OMB values are averaged spatially on a 2.5° × 2.5° averaging grid over the week and plotted spatially, while all observation locations are used to compute histograms in Figs. 3 and 4. In each panel, the 2D map plot in the upper part shows the averaged OMB at each grid box, and the plot in the lower part displays histograms for all data (shaded area), data over land (green line with squares), over ice (gray line with triangle), and over snow (purple line with circle), respectively. In lieu of plotting all GMI channels, only channels currently assimilated operationally over ocean in the GEOS-ADAS are plotted in Figs. 3 and 4. The remaining channels are plotted for reference in the appendix. Significant improvements can be observed in Fig. 4 (M2020 simulation) versus Fig. 3 (CRTM default) spatially. First, there are far fewer points in yellow indicating values outside the ±10 K range. Next, there are many more regions that are in the ±5 K OMB range in Fig. 4 versus Fig. 3. Moreover, when CRTM default emissivities are used, the OMB values of the channels with horizontal polarization are much worse than those vertical polarization; but when M2020 emissivities are used, the OMB differences between horizontal and vertical polarization are decreased. These results are in agreement with those observed in Fig. 2. Additionally, there are some regional improvements that can be seen comparing Figs. 3 and 4. Simulated brightness temperatures are much closer to observed brightness temperatures at 23 and 37 GHz (V/H) in North Africa comparing Figs. 3a–c to Figs. 4a–c. Regions over ice show an improvement over regions near Antarctica in Fig. 4 versus Fig. 3. For the 23 and 37 GHz (V/H) channels, large regions exceeding 10 K departures are reduced around Antarctica, along with departures closer to zero for the 183 GHz channels. In northern Asia, snow covered regions show improvements most clearly comparing Figs. 3 and 4 for the 166 and 183 GHz channels where departures are closer to zero.
GEOS-ADAS calculated values of observation minus background using default CRTM values.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
GEOS-ADAS calculated values of observation minus background using M2020.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
Observing the histograms of OMB values in Figs. 3 and 4, in Fig. 4 there is a clear Gaussian pattern nearly centered around zero for all channels. Figure 3 only has a Gaussian pattern for frequencies at 166 GHz and above (Figs. 3d–f). The two 183 GHz channels have little sensitivity to surface, especially the 183 ± 3 GHz channel, which has a sensitivity that peaks higher in the atmosphere. The improvements are further shown in Fig. 5, which contains spatial plots and histograms of the difference between the absolute values of OMB simulations using M2020 minus the absolute values of OMB using the default CRTM. Negative values indicate the M2020 simulation is closer to the observation (blue); positive values (red) indicate it is further away from the observation. Overall, with exception to the 183 ± 3 GHz channel (Fig. 5e), there are many points where the OMB indicates the M2020 simulation is closer to the observation.
Difference in GEOS-ADAS calculated absolute values of observation minus background using M2020 minus values calculated with default CRTM.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
Next, the relative improvement over land, snow, and ice are considered. Comparing both spatially and viewing histograms in Figs. 1, 3, and 4, there are clear improvements over land. Comparing the histograms over land (green squares), there is a clear Gaussian pattern with a nearly centered around zero for all channels for M2020 (Fig. 4), whereas in Fig. 3 there are cases where there is a Gaussian pattern for the default CRTM case; however, for all channels there is an improvement for the M2020 case (Fig. 4). For points identified as snow covered for the default CRTM and M2020 simulations in Figs. 3 and 4, improvements are still noticeable, they are far less dramatic with long tails remaining in the distribution for Fig. 4. Comparing histograms and spatial distributions of OMB over ice in Figs. 3 and 4, there are improvements; however, the improvements are not as dramatic as either over land or snow. One notable exception is for the 183 ± 7 GHz channel when comparing Figs. 3f and 4f, the gray histograms (indicating observations over ice) have a Gaussian pattern more closely centered around zero in Fig. 4 indicating simulations using M2020 are closer to the GMI observations. Improvements relative to the default CRTM configuration over ice are somewhat expected, especially for channels over ice, which utilize a default value of 0.92 by default in CRTM versus using the emissivity provided by M2020. Similar expectations would apply to frequencies of 89 GHz and above where only constant values of emissivity are used over land and snow surfaces (see Table 1). Improvements relative to the default CRTM configuration over land and snow below 89 GHz indicate that the LandEm of Weng et al. (2001) does not perform as well as M2020.
4. Summary and future work
In this work it has been shown that using emissivity values from M2020 can improve simulation of brightness temperatures from GMI using the GEOS-ADAS. The improvements are most noticeable over land with a dramatic improvement, followed by snow and ice with a less dramatic improvement. With such a dramatic improvement over land, it may be possible to attempt assimilating surface sensitive GMI channels over land, or at the very least as a first guess for an in-line retrieval or adding emissivity to the GEOS-ADAS control vector. This may be necessary given M2020 is a climatology, and variations in surface properties such as soil moisture may result in significant departures especially at lower-frequency channels. Assimilating surface sensitive channels over land could provide information regarding temperature and moisture in the planetary boundary layer in the GEOS-ADAS. This was noted as a key need by the decadal survey (NASEM 2018), and is an ongoing effort at the Global Modeling and Assimilation Office. Additionally, it may be possible to utilize M2020 for similar sensors such as AMSR-2, which has an Earth incidence angle of 55° (Maeda et al. 2016); however, this would require rigorous testing. Full observing system experiments will be conducted using surface sensitive channels over land, snow, and sea ice, which could provide more data in the planetary boundary layer, thus improving its representation in the GEOS system.
Acknowledgments.
The authors thank the NASA GSFC Science Task Group (STG) on “Evaluation and Improvement of Surface Emissivity for Enhanced Earth and Planetary Remote Sensing” for support and funding to conduct this study.
Data availability statement.
Data used for this study include L1C calibrated brightness temperatures freely available from https://doi.org/10.5067/GPM/GMI/GPM/1C/05. Data from M2020 are available from https://doi.org/10.21227/fypd-zj65.
APPENDIX
Spatial Differences for Additional Channels
In section 3b Figs. 3–5 were introduced to highlight differences between using the default emissivity model present in CRTM and simulations performed using CRTM with the M2020 emissivity database. To simplify the discussion, only channels currently assimilated over ocean in the GEOS-ADAS were included. Other centers may have different channel selections; thus, the remaining channels for GMI are plotted following Figs. 3–5 in Figs A1–A3.
GEOS-ADAS calculated values of observation minus background using default CRTM values following Fig. 3.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
GEOS-ADAS calculated values of observation minus background using M2020 following Fig. 4.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
Difference in GEOS-ADAS calculated absolute values of observation minus background using M2020 minus values calculated with default CRTM, following Fig. 5.
Citation: Journal of Atmospheric and Oceanic Technology 39, 10; 10.1175/JTECH-D-21-0187.1
REFERENCES
Baordo, F., and A. J. Geer, 2016: Assimilation of SSMIS humidity-sounding channels in all-sky conditions over land using a dynamic emissivity retrieval. Quart. J. Roy. Meteor. Soc., 142, 2854–2866, https://doi.org/10.1002/qj.2873.
Chen, M., and F. Weng, 2016: Modeling land surface roughness effect on soil microwave emission in Community Surface Emissivity Model. IEEE Trans. Geosci. Remote Sens., 54, 1716–1726, https://doi.org/10.1109/TGRS.2015.2487885.
Chen, Y., F. Weng, Y. Han, and Q. Liu, 2008: Validation of the Community Radiative Transfer Model by using CloudSat data. J. Geophys. Res., 113, D00A03, https://doi.org/10.1029/2007JD009561.
Donlon, C. J., M. Martin, J. Stark, J. Roberts-Jones, E. Fiedler, and W. Wimmer, 2012: The Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) system. Remote Sens. Environ., 116, 140–158, https://doi.org/10.1016/j.rse.2010.10.017.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Han, Y., P. van Delst, Q. Liu, F. Weng, B. Yan, R. Treadon, and J. Derber, 2006: JCSDA Community Radiative Transfer Model (CRTM)-version 1. NOAA Tech. Rep. NESDIS 122, 40 pp., https://repository.library.noaa.gov/view/noaa/1157.
Kim, M. J., J. Jin, A. E. L. Akkraoui, W. McCarty, R. Todling, G. U. Wei, and R. Gelaro, 2020: The framework for assimilating all-sky GPM microwave imager brightness temperature data in the NASA GEOS data assimilation system. Mon. Wea. Rev., 148, 2433–2455, https://doi.org/10.1175/MWR-D-19-0100.1.
Koster, R. D., M. J. Suarez, A. Ducharne, M. Stieglitz, and P. Kumar, 2000: A catchment-based approach to modeling land surface processes in a general circulation model 1. Model structure. J. Geophys. Res., 105, 24 809–24 822, https://doi.org/10.1029/2000JD900327.
Maeda, T., Y. Taniguchi, and K. Imaoka, 2016: GCOM-W1 AMSR2 level 1R product: Dataset of brightness temperature modified using the antenna pattern matching technique. IEEE Trans. Geosci. Remote Sens., 54, 770–782, https://doi.org/10.1109/TGRS.2015.2465170.
Mätzler, C., 1994: Passive microwave signatures of landscapes in winter. Meteor. Atmos. Phys., 54, 241–260, https://doi.org/10.1007/BF01030063.
Munchak, S. J., S. Ringerud, L. Brucker, Y. You, I. de Gelis, and C. Prigent, 2020: An active–passive microwave land surface database from GPM. IEEE Trans. Geosci. Remote Sens., 58, 6224–6242, https://doi.org/10.1109/TGRS.2020.2975477.
NASEM, 2018: Thriving on Our Changing Planet: A Decadal Strategy for Earth Observation from Space. National Academies Press, 716 pp., https://doi.org/10.17226/24938.
Petty, G. W., and R. Bennartz, 2017: Field-of-view characteristics and resolution matching for the Global Precipitation Measurement (GPM) Microwave Imager (GMI). Atmos. Meas. Tech., 10, 745–758, https://doi.org/10.5194/amt-10-745-2017.
Saunders, R., and Coauthors, 2020: RTTOV-13 science and validation report. EUMETSAT Rep. NWPSAF-MO-TV-046, 106 pp., https://nwp-saf.eumetsat.int/site/download/documentation/rtm/docs_rttov13/rttov13_svr.pdf.
Skofronick-Jackson, G., and Coauthors, 2017: The Global Precipitation Measurement (GPM) mission for science and society. Bull. Amer. Meteor. Soc., 98, 1679–1695, https://doi.org/10.1175/BAMS-D-15-00306.1.
Stieglitz, M., A. Ducharne, R. Koster, and M. Suarez, 2001: The impact of detailed snow physics on the simulation of snow cover and subsurface thermodynamics at continental scales. J. Hydrometeor., 2, 228–242, https://doi.org/10.1175/1525-7541(2001)002<0228:TIODSP>2.0.CO;2.
Todling, R., and A. El Akkraoui, 2018: The GMAO Hybrid Ensemble-Variational Atmospheric Data Assimilation System: Version 2.0. NASA Tech. Rep. NASA/TM-2018-104606, 184 pp., https://gmao.gsfc.nasa.gov/pubs/docs/Todling1019.pdf.
Wang, D., and Coauthors, 2017: Surface emissivity at microwaves to millimeter waves over polar regions: Parameterization and evaluation with aircraft experiments. J. Atmos. Oceanic Technol., 34, 1039–1059, https://doi.org/10.1175/JTECH-D-16-0188.1.
Weng, F., B. Yan, and N. C. Grody, 2001: A microwave land emissivity model. J. Geophys. Res., 106, 20 115–20 123, https://doi.org/10.1029/2001JD900019.
Zhu, Y., R. Todling, and J. Jin, 2021: Improving the use of surface-sensitive radiances in the GMAO GEOS system. 23rd Int. TOVS Study Conf., Online, ITWG, http://cimss.ssec.wisc.edu/itwg/itsc/itsc23/agendas/posters/poster.2p.20.zhu.pdf.
