Deep Interpolation of Remote Sensing Land Surface Temperature Data with Partial Convolutions

A new land surface temperature fusion strategy based on cumulative distribution function matching and multiresolution Kalman filtering

A stepwise framework for interpolating land surface temperature under cloudy conditions based on the solar-cloud-satellite geometry

수문학, 기상학 및 기후학 등에서 필수적인 자료중의 하나인 지상기온 자료는 최근 보건, 생물, 환경 등의 다양한 분야로까지 활용영역이 확대되고 있어 그 중요성이 커지고 있으나 지상관측을 통한 지상기온자료의 취득은 시공간적인 제약이 크기 때문에 실측된 기온자료는 시공간 해상도가 낮아 높은 해상도가 요구되는 연구 분야에서는 활용성에 큰 제약을 갖게 된다. 이를 극복하기 위한 하나의 대안으로 상대적으로 높은 시공간 해상도를 가지고 있는 위성영상자료에서 얻을 수 있는 지표면온도 자료를 이용하여 지상기온을 추정하는 많은 연구들이 수행되어 왔다. 본 연구는 이러한 연구의 일환으로써 기상청에서 제공하고 있는 AWS(Automatic Weather Station)에서 취득된 2010년 지상 온도 자료(AWS data)를 바탕으로 대표적인 지표면 온도자료인 MODIS Land Surface temperature(LST data:MOD11A1)와 지상기온에 영향을 미칠 수 있는 Land Cover Data, DEM(digital elevation model) 등의 보조 자료와 함께 다양한 지구통계 기법들을 이용하여 남한 지역의 지상기온을 추정하였다. 추정 전 2010년 전체(365일) LST자료와 AWS자료와의 차이에 대한 RMSE(Root Mean Square Error)값의 계절별 피복별 분석결과 계절에 따른 RMSE값의 변동계수는 0.86으로 나타났으나 피복에 따른 변동계수는 0.00746으로 나타나 계절별 차이가 피복별 차이보다 큰 것으로 분석 되었다. 계절별 RMSE 값은 겨울철이 가장 낮은 것으로 나타났으며 AWS자료와 LST자료와 보조자료를 이용한 선형 회귀분석결과에서도 겨울철의 결정 계수가 가장 높은 0.818로 나타났으며, 여름철의 경우에는 0.078로 나타나 계절별 차이가 매우 크게 나타났다. 이러한 결과를 바탕으로 지구통계 기법들의 대표적인 방법론인 크리깅 방법 중 일반적으로 많이 사용되고 있는 정규 크리깅, 일반 크리깅, 공동 크리킹, 회귀 크리깅을 이용하여 지상기온을 추정한 후 모델의 정확도를 판단할 수 있는 교차 검증을 실시한 결과 정규 크리깅과 일반 크리깅에 의한 RMSE 값은 1.71, 공동 크리깅과 회귀 크리깅에 의한 RMSE 값은 각각 1.848, 1.63으로 나타나 회귀 크리깅 방법에 의한 추정의 정확도가 가장 높은 것으로 분석되었다. Near surface air temperature data which are one of the essential factors in hydrology, meteorology and climatology, have drawn a substantial amount of attention from various academic domains and societies. Meteorological observations, however, have high spatio-temporal constraints with the limits in the number and distribution over the earth surface. To overcome such limits, many studies have sought to estimate the near surface air temperature from satellite image data at a regional or continental scale with simple regression methods. Alternatively, we applied various Kriging methods such as ordinary Kriging, universal Kriging, Cokriging, Regression Kriging in search of an optimal estimation method based on near surface air temperature data observed from automatic weather stations (AWS) in South Korea throughout 2010 (365 days) and MODIS land surface temperature (LST) data (MOD11A1, 365 images). Due to high spatial heterogeneity, auxiliary data have been also analyzed such as land cover, DEM (digital elevation model) to consider factors that can affect near surface air temperature. Prior to the main estimation, we calculated root mean square error (RMSE) of temperature differences from the 365-days LST and AWS data by season and landcover. The results show that the coefficient of variation (CV) of RMSE by season is 0.86, but the equivalent value of CV by landcover is 0.00746. Seasonal differences between LST and AWS data were greater than that those by landcover. Seasonal RMSE was the lowest in winter (3.72). The results from a linear regression analysis for examining the relationship among AWS, LST, and auxiliary data show that the coefficient of determination was the highest in winter (0.818) but the lowest in summer (0.078), thereby indicating a significant level of seasonal variation. Based on these results, we utilized a variety of Kriging techniques to estimate the surface temperature. The results of cross-validation in each Kriging model show that the measure of model accuracy was 1.71, 1.71, 1.848, and 1.630 for universal Kriging, ordinary Kriging, cokriging, and regression Kriging, respectively. The estimates from regression Kriging thus proved to be the most accurate among the Kriging methods compared.

All-sky 1 km land surface temperature (LST) data are urgently needed. Two widely applied approaches to derive such LST data are merging thermal infrared remote sensing (TIR)-passive microwave remote sensing (PMW) observations and merging TIR reanalysis data. However, as only the Moderate Resolution Imaging Spectroradiometer (MODIS) is adopted as the TIR source for merging, current 1 km all-sky LST products are limited to the MODIS observation time. Therefore, a gap still remains in terms of all-sky LST data with a higher temporal resolution or at other times (e.g., dawn-dusk time). Under this background, this study merged the observations of the Medium Resolution Spectrum Imager (MERSI-LL) on board the dusk-dawn-orbit Fengyun (FY)-3E satellite and Global Land Data Assimilation System (GLDAS) data to estimate dawn-dusk 1 km all-sky LST using a random forest-based method (RFRTM). The results showed that the model had good robustness, with an STD of 0.62-0.86 K of the RFRTM LST, compared with the original MERSI-LL LST. Validation against in situ LST showed that the estimated LST had an accuracy of 1.34-3.71 K under all-sky conditions. In addition, compared with the dawn-dusk LST merged from MERSI-LL and the Special Sensor Microwave Imager/Sounder (SSMI/S), the RFRTM LST showed better performance in accuracy and image quality. This study's findings are beneficial for filling the gap in all-sky LST at high spatiotemporal resolutions for associated applications.

The widespread availability of Land Surface Temperature (LST) data from various sources presents a contemporary challenge for urban climate studies: how to efficiently compare these data with the results of traditional methods of temperature monitoring, which typically assume measurements at 2 m under sheltered conditions. In this line, the current study is based primarily on data extracted from a network of 31 points of hourly temperature monitoring at the 2 m level (Tair2m), in use between 2019 and 2021, in the city of Suceava in north-eastern Romania. These data allowed a detailed mapping for each hourly time step through multiple regression, adjusted by IDW, which was identified as the best interpolation method of Tair2m. These data were analyzed in parallel with LST data derived from Landsat imagery available in the analyzed period for 35 summer days with no or low cloud cover. The mapping results of both the Tair2m and LST data describe the main characteristics of the Suceava urban agglomeration (SvMA) heat island, which presents polynuclear features with intensities—as expressed by the temperature difference between the cores of the heat island and the surrounding rural areas—spanning during the summer noontime between 3.0 °C based on Tair2m and 7.1 °C on LST, respectively. The values of the Tair2m–LST differences were 0.68 °C on average, ranging from 5.33 to −19.17 °C, directly proportional to the imperviousness ratio (IMD) values, reaching the highest values in the local climate zones (LCZs) with a high built-up ratio (up to −19.17 °C) and the lowest (0.5 ÷ −0.5 °C) for those with bare soils, with isolated bushes and trees, with few or no buildings. The study results could serve as a tool to downscale the LST data to the level of Tair2m, which is useful for interpretation of the data derived from these commonly used tools in urban climate monitoring.

The recently launched Landsat-9 has an important mission of working together with Landsat-8 to reduce the revisit period of Landsat Earth observations to eight days. This requires the data of Landsat-9 to be highly consistent with that of Landsat-8 to avoid bias caused by data inconsistency when the two satellites are simultaneously used. Therefore, this study evaluated the consistency of the surface reflectance (SR) and land surface temperature (LST) data between Landsat-8 and Landsat-9 based on five test sites from different parts of the world using synchronized underfly image pairs of both satellites. Previous cross-comparisons have demonstrated high consistency between the spectral bands of Landsat-8 and Landsat-9, with differences of around 1%. However, it is unclear whether this low deviation will be amplified in subsequent multiband calculations. It is also necessary to determine whether the difference is consistent across different land cover types. Therefore, this study used a three-level cross-comparison approach to specifically examine these concerns. Besides the commonly used band-by-band comparison, which served as the first-level comparison in this study, this approach included a second-level comparison based on the calculations of several indicators and a third-level comparison based on a composite index calculated from the indicators obtained in the second-level comparison. This three-level approach will examine whether the difference found in the first-level per-band comparison would change after the subsequent calculations in the second- and third-level comparisons. The Remote Sensing based Ecological Index (RSEI) was used for this approach because it is a composite index integrating four indicators. The results of this three-level comparison show that the first-level per-band comparison exhibited high consistency between the two satellites’ SR data, with an average absolute percent change (PC) of 1.88% and an average R 2 of 0.957 across six bands in the five test sites. This deviation increased to 2.21% in the third-level composite index-based comparison, with R 2 decreasing to 0.956. This indicates that after complex calculations, the deviation between the bands of the two satellites was amplified to some extent. However, when analyzing specific land cover types, notable differences emerged between the two satellites for the water category, with an average absolute PC ranging from 18% to 35% and an R 2 of lower than 0.6. Additionally, there were also nearly 5% differences for the built-up land category, with an average R 2 value of lower than 0.7. The comparison of LST data between both satellites also reveals that the Landsat-9 LST is on average 0.24°C lower than Landsat-8 LST across the five test areas but can be 0.58°C lower in built-up land-dominated areas and 0.42°C higher in desert environments. Overall, the SR and LST data between Landsat-8 and Landsat-9 are consistent. However, their performance varies depending on different land cover types. Caution is needed particularly for water-related research when utilizing both satellites simultaneously. Significant discrepancies may also arise in the areas characterized by deserts and built-up lands.

The relationship between remotely sensed land surface temperature (LST) data and ground-measured air temperature is important for a number of applications. This article details a pilot project over the summer (June, July, August) 2010 using Moderate Resolution Imaging Spectroradiometer (MODIS) LST data and air temperature data from a custom network of data loggers across the conurbation of Birmingham, UK. The results show that at night-time air temperature is consistently higher than the satellite-measured LST, but significant station-specific variability exists.

Dengue, one of the main global threats to public health, has experienced a significant increase in its incidence due to the geographic expansion of the Aedes aegypti and Aedes albopictus vectors. This phenomenon is driven by factors such as climate change, urbanization and the emergence of urban heat islands (UHIs), which lead to elevated temperatures that promote the development and transmission of the virus. This study evaluates the spatial and temporal variations of land surface temperature (LST) in Córdoba, Argentina, using remote sensors such as MODIS and those on board the Landsat missions, with the objective of analyzing their relationship with UHIs and Ae. aegypti activity. Besides, we analyzed LST dynamics across data from 2014 to 2024 in a recently urbanized area in order to assess the thermal impact of land cover change. A key finding was that maximum in LST preceded peaks in Ae. aegypti oviposition by about six-seven weeks, revealing a clear lagged temperature-oviposition relationship. Spatial analysis revealed up to $5^{\circ} \mathrm{C}$ differences in LST within the city, with significantly higher values in areas with reduced vegetation. In a specific case, the conversion of a vegetated area into a public park led to an average LST increase of $4^{\circ} \mathrm{C}$ over six years, showing greater thermal intensification than a nearby stable vegetated site. Cluster analysis grouped mosquito trap locations based on LST levels, revealing consistent seasonal trends and lower normalized difference vegetation index (NDVI) values in hotter areas. Our findings confirm the influence of urban land cover changes on thermal variations and its potential to affect Ae. aegypti dynamics. As urban expansion continues, integrating thermal landscape monitoring into vector surveillance systems may help anticipate shifts in mosquito activity and dengue risk. Additionally, our results highlight the importance of using LST data at higher temporal frequency for enhanced vector risk assessment.

Background and Objectives: The outlines a study that investigates the variations in land surface temperature (LST) and land use in Phrae Province, Thailand, during two distinct climate events: The El Niño event in 2020 and the La Niña event in 2024. The study highlights the impact of climate change, including the increasing frequency of these events, on the environment and agriculture. It emphasizes the importance of geospatial technology in monitoring these changes and utilizing satellite imagery to understand land temperature trends for effective disaster prevention planning. The study begins by acknowledging the role of climate change, both natural and human-induced, in contributing to global warming and exacerbating the frequency of El Niño and La Niña events. These climate phenomena lead to severe weather changes that directly impact sectors such as agriculture, often resulting in natural disasters. The objective of this research is twofold: 1) Compare Land Surface Temperature (LST) Variations: The study aims to analyze and compare the variations in LST during two distinct periods, the El Niño event of 2020 and the La Niña event of 2024, within Phrae Province.2) Examine the Relationship between LST and Land Use Types: It seeks to explore how different types of land use—such as forest land, agricultural land, and urban areas—affect LST during these events. Methodology: The methodology involves the use of Landsat 8 satellite imagery from both 2020 and 2024 to gather the data needed for the analysis. The images were processed in the following steps: 1) Composite Band Images: Landsat imagery was used to create composite band images, a process that allows the extraction of relevant data for analysis. 2) Supervised Classification: The images were classified using the Maximum Likelihood Classification algorithm. Training areas were designated to guide the classification, which aimed to categorize land use into three primary types: forest land, agricultural land, and urban/built-up land. 3) Land Surface Temperature (LST) Analysis: The next step was converting the satellite image’s Digital Number (DN) values into radiance, which were then translated into brightness temperature. To adjust for the surface emissivity, the brightness temperature was further refined to provide an accurate measure of LST. 4) Categorizing LST by Land Use Type: After deriving the LST values, the study categorized the temperature ranges according to the different land use types, helping to identify temperature patterns related to land use changes. Main Results: The study presents several key findings related to both land use and LST: 1) land use classification (2020 vs. 2024): In 2020, forest land made up 66.3% of Phrae Province’s total land area, followed by agricultural land (25.7%) and urban/built-up areas (8%). By 2024, the land use pattern had changed slightly. Forest land decreased to 63.5%, while agricultural land increased to 26.9%, and urban/built-up land grew to 9.6%. 2) Land Surface Temperature (LST) Data (2020 and 2024) showed that in 2020, the highest recorded temperature was 44.2°C in March, and the lowest was 23.8°C in August. While, in 2024, the highest recorded temperature was 43.7°C in April, and the lowest was 24.2°C in August. The Monthly average temperature comparison between two events expressed that in 2020, the highest and lowest average temperature found in March (34°C) and August (10°C), respectively. While, in 2024, it was found in April (34°C) and August (10.8°C), respectively. The LST varied among land use types. In 2020, the average LST for forest areas was lowest (21.5°C) and followed by urban/built-up land (32.3°C) and agricultural land (36.2°C), respectively. While, in 2024, it still be found as the same trend as in 2022 but the forest areas shown increased LST (24.4°C), while agricultural land and urban/built-up areas experienced a slight decrease, 34.5°C and 32.3°C, respectively. Conclusion: This study concludes that land surface temperature (LST) variations are significantly influenced by the El Niño and La Niña climate phenomena, with the effects modulated by land use type. Key findings reveal that: (1) During the El Niño event in 2020, agricultural and urban/built-up areas recorded elevated LSTs. This is likely attributable to the intensified heat and drier conditions commonly associated with El Niño episodes. (2) In contrast, during the La Niña event in 2024, forested areas exhibited comparatively higher LSTs than agricultural and urban zones. Although La Niña typically brings cooler and wetter weather, it can also produce localized temperature anomalies, particularly in dense forest regions with limited capacity for heat dissipation. The study underscores the crucial role of geospatial technologies in tracking and analyzing climate-induced variations in land use and surface temperatures. Such insights are essential for local governments and policymakers to design informed, spatially targeted strategies for disaster risk reduction and climate adaptation. This is especially pertinent in agriculturally reliant regions such as Phrae Province, where climate variability poses direct risks to livelihoods and food security. Moreover, the findings contribute to the expanding literature on climate-land interactions, emphasizing the complexity of climate impacts across different ecosystems. The application of satellite data in this context proves to be a powerful tool for environmental monitoring, enabling more accurate and timely assessments of climatic effects. Ultimately, this research advocates for enhanced integration of geospatial analysis in climate resilience planning, ensuring that both human and natural systems are better equipped to navigate future environmental challenges.

Near real-time land surface temperature reconstruction from FY-4A satellite using spatio-temporal attention network

The estimation of spatially resolved near-surface air temperature (NSAT) has been extensively performed in previous studies using satellite-derived land surface temperature (LST) from MODIS. However, there remains a need for estimating daily NSAT based on LST data from other satellites, which has important implications for integrating multi-source LST in estimating NSAT and ensuring the continuity of satellite-derived estimates of NSAT over long-term periods. In this study, we conducted a comprehensive comparison of LST derived from Metop with MODIS LST in the modeling and mapping of daily NSAT. The results show that Metop LST achieves consistent predictive performance with MODIS LST in estimating daily NSAT, and models based on Metop LST or MODIS LST have overall predictive performance of about 1.2–1.4 K, 1.5–2.0 K, and 1.8–1.9 K in RMSE for estimating Tavg, Tmax, and Tmin, respectively. Compared to models based on nighttime LST, daytime LST can improve the predictive performance of Tmax by about 0.26–0.28 K, while performance for estimating Tavg or Tmin using different schemes of LST is comparable. Models based on Metop LST also exhibit high consistency with models utilizing MODIS LST in terms of the variability in predictive performance across months, with RMSE of 1.03–1.82 K, 1.3–2.49 K, and 1.26–2.66 K for Tavg, Tmin, and Tmax, respectively. This temporal variability in performance is not due to sampling imbalance across months, which is confirmed by comparing models trained using bootstrapped samples in balance, and our results imply that sampling representativeness, complicated by retrieval gaps in LST, is an important issue when analyzing the variability in predictive performance for estimating NSAT. To fully assess the predictive capability of Metop LST in estimating daily NSAT, more studies need to be performed using different methods across areas with a range of scales and geographical environments.

This study investigated the seasonal variations of daytime urban thermal environment (UTE) based on land surface temperature (LST) in Shenzhen of 2015. The spatial and temporal adaptive reflectance fusion model (STARFM) was used for retrieving seasonal daytime LST at high spatiotemporal resolution by combining MODIS and HJ-1B LST data. Next, the relationship between the land cover and daytime in each season was examined. Finally, daytime LST patterns were classified, and the effects of seasonal variations of high-grade daytime LSTs were analyzed with landscape metrics. The results showed that (1) the STARFM is capable of generating seasonal daytime LST data at high spatiotemporal resolution. (2) Daytime LSTs were generally higher in the western parts of Shenzhen in spring and summer. (3) Daytime LST in each land cover type showed an increasing trend form winter to summer and decreased from summer to autumn. The highest and lowest daytime LSTs in each season were observed in ISAs and water bodies. (4) Landscape metrics provided a quantitative method for describing seasonal variations in daytime LSTs, and it was found that seasons influenced the intensity and extent of daytime LSTs in Shenzhen. These findings may be helpful for urban planners developing regional urban strategies to improve daytime urban thermal comfort conditions.

The partitioning of available energy between sensible heat and latent heat is important for precise water resources planning and management in the context of global climate change. Land surface temperature (LST) is a key variable in energy balance process and remotely sensed LST is widely used for estimating surface heat fluxes at regional scale. However, the inequality between LST and aerodynamic surface temperature (Taero) poses a great challenge for regional heat fluxes estimation in one-source energy balance models. To address this issue, we proposed a One-Source Model for Land (OSML) to estimate regional surface heat fluxes without requirements for empirical extra resistance, roughness parameterization and wind velocity. The proposed OSML employs both conceptual VFC/LST trapezoid model and the electrical analog formula of sensible heat flux (H) to analytically estimate the radiometric-convective resistance (rae) via a quartic equation. To evaluate the performance of OSML, the model was applied to the Soil Moisture-Atmosphere Coupling Experiment (SMACEX) in United States and the Multi-Scale Observation Experiment on Evapotranspiration (MUSOEXE) in China, using remotely sensed retrievals as auxiliary data sets at regional scale. Validated against tower-based surface fluxes observations, the root mean square deviation (RMSD) of H and latent heat flux (LE) from OSML are 34.5 W/m2 and 46.5 W/m2 at SMACEX site and 50.1 W/m2 and 67.0 W/m2 at MUSOEXE site. The performance of OSML is very comparable to other published studies. In addition, the proposed OSML model demonstrates similar skills of predicting surface heat fluxes in comparison to SEBS (Surface Energy Balance System). Since OSML does not require specification of aerodynamic surface characteristics, roughness parameterization and meteorological conditions with high spatial variation such as wind speed, this proposed method shows high potential for routinely acquisition of latent heat flux estimation over heterogeneous areas.

Land surface temperature (LST) can fully reflect the water–heat exchange cycle of the earth surface that is important for the study of environmental change. There is little research on LST in the semi-arid region of Abha-Khamis-Mushyet, which has a complex topography. The study used LST data, retrieved from ASTER data in semi-arid mountain areas and discussed its relationship with land use/land cover (LULC), topography and the normalized difference vegetation index (NDVI). The results showed that the LST was significantly influenced by altitude and corresponding LULC type. In the study area, during the summer season, extreme high-temperature zones were observed, possibly due to dense concrete surfaces. LST among different types of land use differed significantly, being the highest in exposed rocky areas and built-up land, and the lowest in dense vegetation. NDVI and LST spatial distributions showed opposite trends. The LST–NDVI feature space showed a unique ABC obtuse-angled triangle shape and showed an overall negative linear correlation. In brief, the LST could be retrieved well by the emissivity derived NDVI TES method, which relied on upwelling, downwelling, and transmittance. In addition, the LST of the semi-arid mountain areas was influenced by elevation, slope zenith angle, aspect and LULC, among which vegetation and elevation played a key role in the overall LST. This research provides a roadmap for land-use planning and environmental conservation in mountainous urban areas.

High-resolution, all-weather land surface temperature (LST) data on a global scale are pivotal for accurately reflecting the thermal feedback from the underlying surface. We developed a novel physically constrained land surface model (LSM) simulated LST downscaling and polar-orbiting satellite LST for reconstructing a time-series framework that is particularly effective in addressing the challenges associated with frequent cloud cover and precipitation in mountainous regions. The framework employs observations from 11 stations and MODIS data, integrating the Noah-MP and the ensemble Kalman filter to construct the data assimilation downscaling framework (DADF). This approach provides a comprehensive solution for acquiring high-resolution, all-weather LST, with both temporal (hourly) and spatial (1 km) precision. To improve the accuracy of LSMs and DADF, we corrected the momentum roughness lengths for each of the six surface categories and soil emissivity based on site observations. Six underlying surface parameters derived from measured data serve as a valuable reference for investigating land surface processes in this region. In a validation of MODIS day and night LST with data from 11 measurements, we observed high precision in high-altitude alpine terrains (Kalasai and Arou) and undulating desert terrains (Tazhong sites A–E). The sand–air admixture in desert terrain, caused by wind, introduces errors in site-based observations. The average RMSE of DADF-LST is 3.85 K (compared to the RMSE of 4.72 K for MODIS LST), and the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> > 0.82 across six categories. The DADF incorporates data assimilation algorithms for LST downscaling, enabling accurate capture of actual LST under cloudy conditions. It provides a physically constrained solution for obtaining LST data with high spatial and temporal resolution globally.