In satellite remote sensing applications, enhancing the precision of level 2 (L2) algorithms relies heavily on the accurate estimation of the surface reflectance across the ultraviolet (UV) to visible (VIS) spectrum. However, the mutual dependence between the L2 algorithms and the surface reflectance retrieval poses challenges, necessitating an alternative approach. To address this issue, many satellite algorithms generate Lambertian-equivalent reflectivity (LER) products as a priori surface reflectance data; however, this often results in an underestimation of these data. This study is the first to assess the applicability of background surface reflectance (BSR), derived using a semi-empirical bidirectional reflectance distribution function (BRDF) model, in an operational environmental satellite algorithm. This study pioneered the application of the BRDF model to hyperspectral satellite data at 440 nm, aiming to provide more realistic preliminary surface reflectance data. In this study, the Geostationary Environment Monitoring Spectrometer (GEMS) data were used, and a comparative analysis of the GEMS BSR and GEMS LER retrieved in this study revealed an improvement in the relative root mean squared error (rRMSE) accuracy of 3 %. Additionally, a time series analysis across diverse land types indicated a greater stability exhibited by the BSR than by the LER. For further validation, the BSR was compared with other LER databases using ground-truth data, yielding superior simulation performance. These findings present a promising avenue for enhancing the accuracy of surface reflectance retrieval from hyperspectral satellite data, thereby advancing the practical applications of satellite remote sensing algorithms.

Rapid warming of the Arctic has resulted in widespread sea ice loss. Sea ice radiative forcing (SIRF) is the instantaneous perturbation of Earth’s radiation at the top of the atmosphere (TOA) caused by sea ice. Previous studies focused only on the role of albedo on SIRF. Skin temperature is also closely related to sea ice changes and is one of the main factors in Arctic amplification. In this study, we estimated SIRF considering both surface albedo and skin temperature using radiative kernels. The annual average net-SIRF, which consists of the sum of albedo-SIRF and temperature-SIRF, was calculated as −54.57 ± 3.84 W/m² for the period 1982–2015. In the net-SIRF calculation, albedo-SIRF and temperature-SIRF made similar contributions. However, the albedo-SIRF changed over the study period by 0.12 ± 0.07 W/m² per year, while the temperature-SIRF changed by 0.22 ± 0.07 W/m² per year. The SIRFs for each factor had different patterns depending on the season and region. In summer, rapid changes in the albedo-SIRF occurred in the Kara and Barents regions. In winter, only a temperature-SIRF was observed, and there was little difference between regions compared to the variations in albedo-SIRF. Based on the results of the study, it was concluded that the overall temperature-SIRF is changing more rapidly than the albedo-SIRF. This study indicates that skin temperatures may have a greater impact on the Arctic than albedo in terms of sea ice surface changes.

Snow and cloud discrimination is a main factor contributing to errors in satellite-based snow cover. To address the error, satellite-based snow cover performs snow reclassification tests on the cloud pixels of the cloud mask, but the error still remains. Machine Learning (ML) has recently been applied to remote sensing to calculate satellite-based meteorological data, and its utility has been demonstrated. In this study, snow and cloud discrimination errors were analyzed for GK-2A/AMI snow cover, and ML models (Random Forest and Deep Neural Network) were applied to accurately distinguish snow and clouds. The ML-based snow reclassified was integrated with the GK-2A/AMI snow cover through post-processing. We used the S-NPP/VIIRS snow cover and ASOS in situ snow observation data, which are satellite-based snow cover and ground truth data, as validation data to evaluate whether the snow/cloud discrimination is improved. The ML-based integrated snow cover detected 33–53% more snow compared to the GK-2A/AMI snow cover. In terms of performance, the F1-score and overall accuracy of the GK-2A/AMI snow cover was 73.06% and 89.99%, respectively, and those of the integrated snow cover were 76.78–78.28% and 90.93–91.26%, respectively.

Air temperature (Ta), defined as the temperature 2 m above the land’s surface, is one of the most important factors for environment and climate studies. Ta can be measured by obtaining the land surface temperature (LST) which can be retrieved with the 11- and 12-µm bands from satellite imagery over a large area, and LST is highly correlated with Ta. To measure the Ta in a broad area, we studied a Ta retrieval method through Deep Neural Network (DNN) using in-situ data and satellite data of South Korea from 2014 to 2017. To retrieve accurate Ta, we selected proper input variables and conditions of a DNN model. As a result, Normalized Difference Vegetation Index, Normalized Difference Water Index, and 11- and 12-µm band data were applied to the DNN model as input variables. And we also selected proper condition of the DNN model with test various conditions of the model. In validation result in the DNN model, the best accuracy of the retrieved Ta showed an correlation coefficient value of 0.98 and a root mean square error (RMSE) of 2.19 K. And then we additional 3 analysis to validate accuracy which are spatial representativeness, seasonal analysis and time series analysis. We tested the spatial representativeness of the retrieved Ta. Results for window sizes less than 132 × 132 showed high accuracy, with a correlation coefficient of over 0.97 and a RMSE of 1.96 K and a bias of −0.00856 K. And in seasonal analysis, the spring season showed the lowest accuracy, 2.82 K RMSE value, other seasons showed high accuracy under 2K RMSE value. We also analyzed a time series of six the Automated Synoptic Observing System (ASOS) points (i.e., locations) using data obtained from 2018 to 2019; all of the individual correlation coefficient values were over 0.97 and the RMSE values were under 2.41 K. With these analysis, we confirm accuracy of the DNN model was higher than previous studies. And we thought the retrieved Ta can be used in other studies or climate model to conduct urban problems like urban heat islands and to analyze effects of arctic oscillation.

Because ice surface temperature (IST) controls snow melt, sea ice growth and air–ocean heat exchange, it is a key parameter in analyses of the Arctic climate system. However, optical satellite-based IST data often have missing values due to the satellite orbit, clouds and polar night. In this study, we estimated IST using a deep neural network (DNN) algorithm based on meteorological, sea ice and geometric variables that are strongly correlated with IST. Moderate-Resolution Imaging Spectroradiometer (MODIS)/Terra IST data were used, and the input data were 2 m air temperature (Ta), 30-year averaged Ta (Ta climatology), total column water vapour (TCWV), solar zenith angle (SZA), local solar noon angle (LSN) and latitude. The data were classified into six cases according to sea ice age (SIA) and Ta to create an efficient DNN model that supplemented the IST data missing from the existing dataset. Model validation based on the MODIS IST data revealed a correlation coefficient of 0.94, root mean square error of 3.54 K and relative root mean square error of 1.35%, showing high accuracy.

Specific humidity (SH) which means the amount of water vapor in 1 kg of air, is used as an indicator of energy exchange between the atmosphere and the Earth’s surface. SH is typically computed using microwave satellites. However, the spatial resolution of data for microwave satellite is too low. To overcome this disadvantage, we introduced new methods that applied data collected by the Landsat-8 satellite with high spatial resolution (30 m), a meteorological model, and observation data for South Korea in 2016–2017 to 4 machine learning techniques to develop an optimized technique for computing SH. Among the 4 machine learning techniques, the random forest-based method had the highest accuracy, with a coefficient of determination (R) of 0.98, Root Mean Square Error (RMSE) of 0.001, bias of 0, and Relative Root Mean Square Error (RRMSE) of 11.16%. We applied this model to compute land surface SH using data from 2018 to 2019 and found that it had high accuracy (R = 0.927, RMSE = 0.002, bias = 0, RRMSE = 28.35%). Although the data used in this study were limited, the model was able to accurately represent a small region based on an ensemble of satellite and model data, demonstrating its potential to address important issues related to SH measurements from satellites.