Efficient hyper-parameter determination for regularised linear BRDF parameter retrieval

ABSTRACT

Linear kernel driven models of the surface Bidirectional Reflectance Distribution Function (BRDF) are valuable tools for exploiting Earth observation data acquired at different sun–sensor geometries. Here we present a method that efficiently determines linear BRDF model weights using Tikhonov smoothing where the smoothing parameter λ is determined via a Generalized Singular Value Decomposition with the root mean square error prescribed depending on the MODIS band. We applied this method to twenty-six different deciduous broadleaf sites across an entire year using the MODIS Terra and Aqua reflectance data products. Kernel weights and white sky albedo derived from this GSVD method were generally consistent with those provided by the MCD43 data products. The GSVD derived results had less sample variability compared to the MCD43 data products, attributable to the assumed smoothness between kernel weights in the Tikhonov smoothing method. The GSVD technique consistently outperforms MCD43 in the reconstruction of observed MODIS reflectance data, of which retrievals from this method will do a better job of estimating albedo and normalizing data to specified geometries.     

Validation of Narrow-band Surface Albedo Retrieved from FY-3C MERSI Satellite Data

Surface albedo is one of the driving factors in surface radiant energy balance and surface-atmosphere interaction.It is widely used in surface energy balance, medium and long-term weather forecasting and global change research.This study aims to validate the surface albedo retrieved from FY-3C MERSI. This paper selected four regions in Africa and North America as study areas to validate the retrieved albedo from the reflectance data and angle data of FY-3C MERSI at 250 m resolution in 2014. The semi-empirical kernel-driven BRDF(bidirectional reflectance distribution function) model RossThick-LiSparseR and least squares fitting method were used to calculate the parameter of BRDF. Then four narrow-band black-sky albedos and four narrow-band white-sky albedos can be obtained by angle integration. Finally, the cross-validation of FY-3C surface narrow-band albedo products with MODIS albedo products (MCD43A3) was carried out. The results show that theRoot Mean Square Error(RMSE) between the FY-3C narrow-band albedo and the corresponding MODIS narrow-band albedo is in the range of 0.01 ~ 0.04, and the Mean Bias (MBIAS) is 0.09. FY-3C narrow-band albedo has good consistency with the corresponding MODIS narrow-band albedo in the visible and near-infrared bands. So, the methodologyof using the BRDF model to invert the surface albedo of FY-3C medium resolution imaging spectrometer data is feasible and reliable. The further improvement of the inversion accuracy of FY3C-MERSI surface albedo also depends on the improvement of basic data processing quality, including image geometric correction, calibration, and strict data quality control.     

FY-3C中分辨率成像光谱仪数据的窄波段地表反照率验证研究

地表反照率数据对地表能量平衡和全球变化研究具有重要意义。基于2014年FY-3C卫星250 m分辨率的反射率数据和角度数据,选取非洲及北美洲的4个区域作为研究区,采用RossThick-LiSparseR模型作为BRDF(Bidirectional Reflectance Distribution Function)核模型反演了地表窄波段反照率,得到250 m分辨率的4个窄波段黑空、白空反照率。将反演得到的FY-3C地表窄波反照率产品与MODIS反照率产品（MCD43A3）数据进行了交叉验证,结果表明：FY-3C窄波段反照率与对应MODIS窄波段反照率对比的均方根误差在0.01~0.04,平均偏差(MBIAS)为0.09,FY-3C窄波段反照率与对应的MODIS窄波段反照率在可见光波段、近红外波段有较好的一致性。本研究提升了国产风云极轨卫星的应用范围,可为FY-3C地表反照率业务化产品提供算法支撑。     

Evaluation of the MODIS (MOD10A1) daily snow albedo product over the Greenland ice sheet

This study evaluates the performance of the beta-test MODIS (MOD10A1) daily albedo product using in situ data collected in Greenland during summer 2004. Results indicate the beta-test product tracks the general seasonal variability in albedo but exhibits significant more temporal variability than observed at the stations. This may indicate problems with the cloud detection algorithm, and/or failure of the BRDF model to adequately model the bidirectional reflectance of snow. Comparisons with in situ observations at five automatic weather stations in Greenland indicate an overall RMSE of 0.067 for the Terra instrument and an RMSE of 0.075 on Aqua. The Terra-retrieved-albedo are slightly better correlated with the in situ data than the Aqua retrievals (r = 0.79 versus r = 0.77). Comparisons were also made between the MODIS daily albedo product and the MODIS 16-day albedo product (MOD43B3). Results indicate general correspondence between the two products, with better agreement found using the Terra-retrieved-albedo than the Aqua-retrieved albedo. The reason for the differences in albedo between the Aqua and Terra satellites remains unclear. At the stations examined, both the Terra and Aqua retrievals were made at nearly the same time of the day and therefore the differences in albedo between the satellites cannot be explained by differences in solar illumination. Finally, the albedo derived using MODIS data and the direct estimation algorithm (DEA) was also compared with 2004 Greenland in situ data. Results from this comparison suggest that the DEA performs well as long as the solar zenith angle of the observation is not greater than 70°.     

Validation of the Climate-SAF surface broadband albedo product: Comparisons with in situ observations over Greenland and the ice-covered Arctic Ocean

This paper describes a validation study performed by comparing the Climate-SAF Surface Albedo Product (SAL) to ground truth observations over Greenland and the ice-covered Arctic Ocean. We compare Advanced Very High Resolution Radiometer (AVHRR)-based albedo retrievals to data from the Greenland Climate Network (GCN) weather stations and the floating ice station Tara for polar summer 2007. The AVHRR dataset consists of 2755 overpasses. The overpasses are matched to in situ observations spatially and temporally. The SAL algorithm presented here derives the surface broadband albedo from AVHRR channels 1 and 2 using an atmospheric correction, temporal sampling of an empirical Bidirectional Reflectance Distribution Function (BRDF), and a narrow-to-broadband conversion algorithm. The satellite product contains algorithms for snow, sea ice, vegetation, bare soil, and water albedo. At the Summit and DYE-2 stations on the Greenland ice sheet, instantaneous SAL RMSE is 0.073. The heterogeneous surface conditions at satellite pixel scale over the stations near the Greenland west coast increase RMSE to > 0.12. Over Tara, the instantaneous SAL RMSE is 0.069. The BRDF sampling approach reduces RMSE over the ice sheet to 0.053, and to 0.045 over Tara. Taking into account various sources of uncertainty for both satellite retrievals and in situ observations, we conclude that SAL agrees with in situ observations within their limits of accuracy and spatial representativeness.     

Comparison between in situ and MODIS-derived spectral reflectances of snow and sea ice in the Amundsen Sea,Antarctica

The spectral albedo and directional reflectance of snow and sea ice were measured on sea ice of various types, including nilas, grey ice, pancake ice, multi-year pack ice, and land-fast ice in the Ross, Amundsen and Bellingshausen seas during a summer cruise in February through March 2000. Measurements were made using a spectroradiometer that has 512 channels in the visible and near-infrared (VNIR) region in which 16 of the 36 bands of the Moderate Resolution Imaging Spectroradiometer (MODIS) are covered. Directional reflectance is also retrieved from the MODIS radiometrically calibrated data (Level 1B) concurrently acquired from the first National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) satellite, Terra. The locations of the ground ice stations are identified accurately on the MODIS images, and the spectral albedo and directional reflectance values at the 16 VNIR MODIS bands are extracted for those pixel locations. MODIS-derived reflectance is then corrected for the intervening atmosphere whose parameters are retrieved from the MODIS atmospheric profiles product (MOD07_L2) for the same granule. The corresponding spectral albedo and directional reflectance with the same viewing geometry as MODIS are derived from our ground-based spectroradiometer measurements. Because the footprint of the ground spectroradiometer is much smaller than the pixel sizes of MODIS images, the averaged spectral reflectance and albedo in the vicinity of each ice station are simulated for the corresponding MODIS pixel from the ground spectral measurements by weighting over different surface types (various ice types and open water). An accurate determination of ice concentration is important in deriving ground reflectance of a simulated pixel from in situ measurements. The best agreement between the in situ and MODIS measurements was found when the ground had 10/10 ice concentration (discrepancy range 0.2–11.69%, average 4.8%) or was oneice-type dominant (discrepancy range 0.8–16.9%, average 6.2%). The more homogeneous the ground surface and the less variable the ground topography, the more comparable between the in situ and satellite-derived reflectance is expected.     

基于中国通量网的MODIS短波反照率验证与分析

遥感地表反照率产品的验证与分析是将其应用于环境研究的基础。采用中国通量网的地表实测短波反照率数据对MODIS反照率产品进行对比和分析,针对选取的8个地面站点,提取了2004年的MODIS反照率产品并进行验证。这些站点的植被覆盖情况涵盖了草地、森林和农业用地。结果显示MODIS在多数情况下能提供准确的地表反照率产品。针对各个站点的误差、均方根误差、相关系数分析都显示了这个结果,总体反演误差在0.002左右。较大的误差出现在有冰雪影响的时候,排除受积雪影响的数据,总体均方根误差可达0.028。分析了引起误差的原因并提出了改进意见。     

Mapping bottom depth and albedo in coastal waters of the South China Sea islands and reefs using Landsat TM and ETM+ data

Optical models for the retrieval of shallow water bottom depth and albedo using multispectral data usually require in situ water depth data to tune the model parameters. In the South China Sea (SCS), however, such in situ data are often lacking or obsolete (perhaps from half a century ago) for most coastal waters around its islands and reefs. Here, we combine multispectral data collected by MODIS and Landsat to estimate bottom depth and albedo for four coral reef regions in the SCS, with results partially validated by some scarce in situ data. The waters in these remote regions are oligotrophic whose optical properties can be well derived from Moderate Resolution Imaging Spectroradiometer (MODIS) measurements when the waters are optically deep. The MODIS-derived optical properties are used to estimate the water column attenuation to the Landsat measurements over shallow waters, thus eliminating the requirement of model tuning using field measured water depths. The model is applied to four Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) images covering Pratas Atoll, Woody Island, Scarborough Shoal, and North Danger Reefs. The retrieved bathymetry around Pratas Atoll and North Danger Reefs are validated with some in situ data between 1 and 25 m. The relative difference and root mean square difference between the two measurements were 17% and 1.6 m, for Pratas Atoll and 11% and 1.1 m for North Danger Reefs, respectively. These results suggest that the approach developed here may be extended to other shallow, clear waters in the SCS.     

Accuracy assessment of the MODIS 16-day albedo product for snow: comparisons with Greenland in situ measurements

The accuracy of the Moderate Resolution Imaging Spectroradiometer (MODIS) 16-day albedo product (MOD43) is assessed using ground-based albedo observations from automatic weather stations (AWS) over spatially homogeneous snow and semihomogeneous ice-covered surfaces on the Greenland ice sheet. Data from 16 AWS locations, spanning the years 2000-2003, were used for this assessment. In situ reflected shortwave data were corrected for a systematic positive spectral sensitivity bias of between 0.01 and 0.09 on a site-by-site basis using precise optical black radiometer data. Results indicate that the MOD43 albedo product retrieves snow albedo with an average root mean square error (RMSE) of ±0.07 as compared to the station measurements, which have ±0.035 RMSE uncertainty. If we eliminate all satellite retrievals that rely on the backup algorithm and consider only the highest quality results from the primary bidirectional reflectance distribution function (BRDF) algorithm, the MODIS albedo RMSE is ±0.04, slightly larger than the in situ measurement uncertainty. There is general agreement between MODIS and in situ observations for albedo <0.7, while near the upper limit, a −0.05 MODIS albedo bias is evident from the scatter of the 16-site composite.     

Multi-temporal MODIS-Landsat data fusion for relative radiometric normalization, gap filling, and prediction of Landsat data

A semi-physical fusion approach that uses the MODIS BRDF/Albedo land surface characterization product and Landsat ETM+ data to predict ETM+ reflectance on the same, an antecedent, or subsequent date is presented. The method may be used for ETM+ cloud/cloud shadow and SLC-off gap filling and for relative radiometric normalization. It is demonstrated over three study sites, one in Africa and two in the U.S. (Oregon and Idaho) that were selected to encompass a range of land cover land use types and temporal variations in solar illumination, land cover, land use, and phenology. Specifically, the 30 m ETM+ spectral reflectance is predicted for a desired date as the product of observed ETM+ reflectance and the ratio of the 500 m surface reflectance modeled using the MODIS BRDF spectral model parameters and the sun-sensor geometry on the predicted and observed Landsat dates. The difference between the predicted and observed ETM+ reflectance (prediction residual) is compared with the difference between the ETM+ reflectance observed on the two dates (temporal residual) and with respect to the MODIS BRDF model parameter quality. For all three scenes, and all but the shortest wavelength band, the mean prediction residual is smaller than the mean temporal residual, by up to a factor of three. The accuracy is typically higher at ETM+ pixel locations where the MODIS BRDF model parameters are derived using the best quality inversions. The method is most accurate for the ETM+ near-infrared (NIR) band; mean NIR prediction residuals are 9%, 12% and 14% of the mean NIR scene reflectance of the African, Oregon and Idaho sites respectively. The developed fusion approach may be applied to any high spatial resolution satellite data, does not require any tuning parameters and so may be automated, is applied on a per-pixel basis and is unaffected by the presence of missing or contaminated neighboring Landsat pixels, accommodates for temporal variations due to surface changes (e.g., phenological, land cover/land use variations) observable at the 500 m MODIS BRDF/Albedo product resolution, and allows for future improvements through BRDF model refinement and error assessment.     

基于ART模型的MODIS积雪反照率反演研究

积雪反照率是研究局地或全球的能量收支平衡和气候变化中的重要参数,遥感反演为积雪反照率的获取提供了便利的手段。积雪反照率大小主要取决于积雪的自身物理属性(雪粒径、形状和污染物等因子)以及天气状况,遥感反演反照率大多基于双向反射模型(BRDF),积雪BRDF模型常使用积雪辐射传输模型获得。采用考虑了雪粒径、粒子形状以及污染物影响的渐进辐射传输理论(ART)模型,建立了MODIS积雪反照率反演算法,得到了MODIS 8d合成积雪反照率产品。将此算法应用于具有均一积雪地表的格陵兰岛地区,并使用GC-Net实测数据进行了验证,反演的总均方根误差(RMSE)为0.018,相关系数(r)为0.83,结果表明考虑了积雪特性的ART模型能够较好地反演积雪反照率,而且反演需要的参数较少。     

Comparison of MODIS broadband albedo over an agricultural site with ground measurements and values derived from Earth observation data at a range of spatial scales

Land surface albedo is dependent on atmospheric state and hence is difficult to validate. Over the UK persistent cloud cover and land cover heterogeneity at moderate (km-scale) spatial resolution can also complicate comparison of field-measured albedo with that derived from instruments such as the Moderate Resolution Imaging Spectrometer (MODIS). A practical method of comparing moderate resolution satellite-derived albedo with ground-based measurements over an agricultural site in the UK is presented. Point measurements of albedo made on the ground are scaled up to the MODIS resolution (1 km) through reflectance data obtained at a range of spatial scales. The point measurements of albedo agreed in magnitude with MODIS values over the test site to within a few per cent, despite problems such as persistent cloud cover and the difficulties of comparing measurements made during different years. Albedo values derived from airborne and field-measured data were generally lower than the corresponding satellite-derived values. This is thought to be due to assumptions made regarding the ratio of direct to diffuse illumination used when calculating albedo from reflectance. Measurements of albedo calculated for specific times fitted closely to the trajectories of temporal albedo derived from both Systeme pour l'Observation de la Terre (SPOT) Vegetation (VGT) and MODIS instruments.     

Commentary on Wang and Zender—MODIS snow albedo bias at high solar zenith angles relative to theory and to in situ observations in Greenland

The recent paper by Wang and Zender [Wang, X., & Zender, C. S. (2010). MODIS snow albedo bias at high solar zenith angles relative to theory and to in situ observations in Greenland. Remote Sensing of Environment.] draws erroneous conclusions about solar zenith angle biases at high latitudes by not making appropriate use of the extensive quality flags available with the MODIS BRDF/Albedo. Coarse resolution MODIS white-sky albedo data are compared with actual blue-sky field albedometer measurements from the Greenland GC-Net. By utilizing large area averages of the MODIS data product that combine both high quality and poor quality data indiscriminately, the authors erroneously conclude that the accuracy deteriorates for solar zenith angle (SZA) > 55° and often becomes physically unrealistic for SZA > 65°. Once the quality flags are considered, however, the comparisons demonstrate that the MODIS product performs quite well out to the recommended limit for product use of 70° SZA. This verifies the conclusions of an earlier more rigorous evaluation performed by Stroeve et al. [Stroeve, J., Box, J. E., Gao, F., Liang, S., Nolin, A., & Schaaf, C. B. (2005). Accuracy assessment of the MODIS 16-day albedo product for snow: comparisons with Greenland in situ measurements. Remote Sensing of Environment.]. With over a decade of observations and products now available from the MODIS instrument, these data are increasingly being used to evaluate and tune climate and biogeochemical models. However, such use should take into account the documented quality and limitations of the satellite-derived product.     

Upscaling of sparse in situ soil moisture observations by integrating auxiliary information from remote sensing

Upscaling of sparse in situ soil moisture (SM) observations is essential for the validation of current and upcoming space-borne SM retrievals, and the successful application of SM observations in hydrological models or data assimilation. In this study, we construct a novel method based on Bayesian data fusion to upscale in situ SM observations to the coarse scale of microwave remote sensing. In the framework of Bayesian theory, the valuable auxiliary information obtained in Moderate Resolution Imaging Spectroradiometer (MODIS) apparent thermal inertia (ATI) is integrated into the upscaling process. The method is validated using SM wireless sensor network data in the Tibetan plateau, which covers an area of approximately 30 × 30 km² with 20 in situ stations. Results confirm that the upscaled SM using the method with randomly selected three stations from the 20 stations is extremely close to the mean of the 20 SMs. The mean root mean square error (RMSE) between the upscaled SM and the mean of the 20 in situ SMs was 0.02 m³ m^?3, and the max RMSE was less than 0.05 m³ m^?3. Furthermore, the sensitivity of the upscaling accuracy to the number of in situ observations is discussed. When the number of in situ observations is greater than nine, the increasing accuracy of the Bayesian method is limited by the uncertainty in the ATI of the remote sensing.     

Validation of the MODIS reflectance product under UK conditions

Surface reflectance obtained from remote-sensing data is the main input to almost all remote-sensing applications. The availability and special characteristics of Moderate Resolution Imaging Spectroradiometer (MODIS) products have led to their use worldwide. Validation of the MODIS reflectance product is then crucial to provid information on uncertainty in the reflectance data, and in other MODIS products and in the applied surface–atmosphere models. Compact Airborne Spectrographic Imager (CASI) and Système Pour l'Observation de la Terre (SPOT) data, collected during the Network for Calibration and Validation in Earth Observation (NCAVEO) 2006 Field Campaign, were applied to validate daily MODIS reflectance data over a site in the southern UK. The difference in the view geometry of at-nadir CASI and SPOT data and off-nadir MODIS data was dealt with using a semi-empirical bidirectional reflectance distribution function (BRDF) model. The validation results showed that for our particular study site, the absolute errors in the MODIS reflectance product were too large to allow the albedo data to be used directly in climate models. The errors were mainly related to the uncertainties in the MODIS atmospheric variables, the BRDF model, and undetected clouds and cloud shadows. More generally, the study highlights the extreme difficulty of achieving pixel-level validation of coarse spatial resolution satellite sensor data in an environment in which the atmosphere is constantly changing, and in which the landscape is characterized by high space–time heterogeneity.     

Evaluation of MODIS gross primary productivity and land cover products for the humid tropics using oil palm trees in Peninsular Malaysia and Google Earth imagery

Conducting quantitative studies on the carbon balance or productivity of oil palm is important in understanding the role of this ecosystem in global climate change. In this study, we evaluated the accuracy of MODIS (Moderate Resolution Imaging Spectroradiometer) annual gross primary productivity (GPP) (the product termed MOD-17) and its upstream products, especially the MODIS land cover product (the product termed MOD-12). We used high-resolution Google Earth images to classify the land cover classes and their percentage cover within each 1 km spatial resolution MODIS pixel. We used field-based annual GPP for 2006 to estimate GPP for each pixel based on percentage cover. Both land cover and GPP were then compared to MODIS land cover and GPP products. The results show that for pure pixels that are 100% covered by mature oil palm trees, the RMSE (root mean square error) between MODIS and field-based annual GPP is 18%, and that this is increased to 27% for pixels containing mostly oil palm. Overall, for an area of about 42 km² the RMSE is 26%. We conclude that land cover classification (at 1 km resolution) is one of the main factors for the discrepancy between MODIS and field-based GPP. We also conclude that the accuracy of the MODIS GPP product could be improved significantly by using higher-resolution land cover maps.     

Estimating daily maximum air temperature from MODIS in British Columbia,Canada

Air temperature (T_a) is an important climatological variable for forest research and management. Due to the low density and uneven distribution of weather stations, traditional ground-based observations cannot accurately capture the spatial distribution of T_a, especially in mountainous areas with complex terrain and high local variability. In this paper, the daily maximum T_a in British Columbia, Canada was estimated by satellite remote sensing. Aqua MODIS (Moderate Resolution Imaging Spectroradiometer) data and meteorological data for the summer period (June to August) from 2003 to 2012 were collected to estimate T_a. Nine environmental variables (land surface temperature (LST), normalized difference vegetation index (NDVI), modified normalized difference water index (MNDWI), latitude, longitude, distance to ocean, altitude, albedo, and solar radiation) were selected as predictors. Analysis of the relationship between observed T_a and spatially averaged remotely sensed LST indicated that 7 × 7 pixel size was the optimal window size for statistical models estimating T_a from MODIS data. Two statistical methods (linear regression and random forest) were used to estimate maximum T_a, and their performances were validated with station-by-station cross-validation. Results indicated that the random forest model achieved better accuracy (mean absolute error, MAE = 2.02°C, R² = 0.74) than the linear regression model (MAE = 2.41°C, R² = 0.64). Based on the random forest model at 7 × 7 pixel size, daily maximum T_a at a resolution of 1 km in British Columbia in the summer of 2003–2012 was derived, and the spatial distribution of summer T_a in this area was discussed. The satisfactory results suggest that this modelling approach is appropriate for estimating air temperature in mountainous regions with complex terrain.     

Retrieval of canopy height using moderate-resolution imaging spectroradiometer (MODIS) data

In this study we use the 500 m Moderate Resolution Imaging Spectroradiometer (MODIS) Bidirectional Reflectance Distribution Function (BRDF) product to develop multivariate linear regression models that estimate canopy heights over study sites at Howland Forest, Maine, Harvard Forest, Massachusetts and La Selva Forest, Costa Rica using (1) directional escape probabilities that are spectrally independent and (2) the directional spectral reflectances used to derive the directional escape probabilities. These measures of canopy architecture are compared with canopy height information retrieved from the airborne Laser Vegetation Imaging Sensor (LVIS). Both the escape probability and the directional reflectance approaches achieve good results, with correlation coefficients in the range 0.54-0.82, although escape probability results are usually slightly better. This suggests that MODIS 500 m BRDF data can be used to extrapolate canopy heights observed by widely-spaced satellite LIDAR swaths to larger areas, thus providing wide-area coverage of canopy height.     

An algorithm for the retrieval of 30-m snow-free albedo from Landsat surface reflectance and MODIS BRDF

We present a new methodology to generate 30-m resolution land surface albedo using Landsat surface reflectance and anisotropy information from concurrent MODIS 500-m observations. Albedo information at fine spatial resolution is particularly useful for quantifying climate impacts associated with land use change and ecosystem disturbance. The derived white-sky and black-sky spectral albedos may be used to estimate actual spectral albedos by taking into account the proportion of direct and diffuse solar radiation arriving at the ground. A further spectral-to-broadband conversion based on extensive radiative transfer simulations is applied to produce the broadband albedos at visible, near infrared, and shortwave regimes. The accuracy of this approach has been evaluated using 270 Landsat scenes covering six field stations supported by the SURFace RADiation Budget Network (SURFRAD) and Atmospheric Radiation Measurement Southern Great Plains (ARM/SGP) network. Comparison with field measurements shows that Landsat 30-m snow-free shortwave albedos from all seasons generally achieve an absolute accuracy of ±0.02-0.05 for these validation sites during available clear days in 2003-2005, with a root mean square error less than 0.03 and a bias less than 0.02. This level of accuracy has been regarded as sufficient for driving global and regional climate models. The Landsat-based retrievals have also been compared to the operational 16-day MODIS albedo produced every 8-days from MODIS on Terra and Aqua (MCD43A). The Landsat albedo provides more detailed landscape texture, and achieves better agreement (correlation and dynamic range) with in-situ data at the validation stations, particularly when the stations include a heterogeneous mix of surface covers.     

Evaluation of MODIS land surface temperature with in-situ snow surface temperature from CREST-SAFE

This article presents the procedure and results of a temperature-based validation approach for the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Surface Temperature (LST) product provided by the National Aeronautics and Space Administration Terra and Aqua Earth Observing System satellites using in-situ LST observations recorded at the Cooperative Remote Sensing Science and Technology Center – Snow Analysis and Field Experiment (CREST-SAFE) during the years of 2013 (January–April) and 2014 (February–April). A total of 314 day-and-night clear-sky thermal images, acquired by the Terra and Aqua satellites, were processed and compared to ground-truth data from CREST-SAFE with a frequency of one measurement every 3 min. CREST-SAFE is a synoptic ground station, located in the cold county of Caribou in Maine, USA, with a distinct advantage over most meteorological stations because it provides automated and continuous LST observations via an Apogee Model SI-111 Infrared Radiometer. This article also attempts to answer the question of whether a single pixel (1 km²) or several spatially averaged pixels should be used for satellite LST validation by increasing the MODIS window size to 5 × 5, 9 × 9, and 25 × 25 windows.

Several trends in the MODIS LST data were observed, including the underestimation of daytime values and night-time values. Results indicate that although all the data sets (Terra and Aqua, diurnal and nocturnal) showed high correlation with ground measurements, day values yielded slightly higher accuracy (about 1°C), both suggesting that MODIS LST retrievals are reliable for similar land-cover classes and atmospheric conditions. Increasing the MODIS window size showed an overestimation of in-situ LST and some improvement in the daytime Terra and night-time Aqua biases, with the highest accuracy achieved with the 5 × 5 window. A comparison between MODIS emissivity from bands 31, 32, and in-situ emissivity showed that emissivity errors (relative error = ?0.30%) were insignificant.