Spatial Downscaling of Land Surface Temperature over Heterogeneous Regions Using Random Forest Regression Considering Spatial Features

Estimation of 100 m root zone soil moisture by downscaling 1 km soil water index with machine learning and multiple geodata

Root zone soil moisture (RZSM) is crucial for agricultural water management and land surface processes. The 1 km soil water index (SWI) dataset from Copernicus Global Land services, with eight fixed characteristic time lengths (T), requires root zone depth optimization (Topt) and is limited in use due to its low spatial resolution. To estimate RZSM at 100-m resolution, we integrate the depth specificity of SWI and employed random forest (RF) downscaling. Topographic synthetic aperture radar (SAR) and optical datasets were utilized to develop three RF models (RF1: SAR, RF2: optical, RF3: SAR + optical). At the DEMMIN experimental site in northeastern Germany, Topt (in days) varies from 20 to 60 for depths of 10 to 30 cm, increasing to 100 for 40-60 cm. RF3 outperformed other models with 1 km test data. Following residual correction, all high-resolution predictions exhibited strong spatial accuracy (R ≥ 0.94). Both products (1 km and 100 m) agreed well with observed RZSM during summer but overestimated in winter. Mean R between observed RZSM and 1 km (100 m; RF1, RF2, and RF3) SWI ranges from 0.74 (0.67, 0.76, and 0.68) to 0.90 (0.88, 0.81, and 0.82), with the lowest and highest R achieved at 10 cm and 30 cm depths, respectively. The average RMSE using 1 km (100 m; RF1, RF2, and RF3) SWI increased from 2.20 Vol.% (2.28, 2.28, and 2.35) at 30 cm to 3.40 Vol.% (3.50, 3.70, and 3.60) at 60 cm. These negligible accuracy differences underpin the potential of the proposed method to estimate RZSM for precise local applications, e.g., irrigation management.

HiMIC-Monthly: A 1 km high-resolution atmospheric moisture index collection over China, 2003-2020

Near-surface atmospheric moisture is a key environmental and hydro-climatic variable that has significant implications for the natural and human systems. However, high-resolution moisture data are severely lacking for fine-scale studies. Here, we develop the first 1 km high spatial resolution dataset of monthly moisture index collection in China (HiMIC-Monthly) over a long period of 2003~2020. HiMIC-Monthly is generated by the light gradient boosting machine algorithm (LightGBM) based on observations at 2,419 weather stations and multiple covariates, including land surface temperature, vapor pressure, land cover, impervious surface proportion, population density, and topography. This collection includes six commonly used moisture indices, enabling fine-scale assessment of moisture conditions from different perspectives. Results show that the HiMIC-Monthly dataset has a good performance, with R2 values for all six moisture indices exceeding 0.96 and root mean square error and mean absolute error values within a reasonable range. The dataset exhibits high consistency with in situ observations over various spatial and temporal regimes, demonstrating broad applicability and strong reliability.

Reconstruction of Gap-Free Land Surface Temperature at a 100 m Spatial Resolution from Multidimensional Data: A Case in Wuhan, China

Land surface temperatures (LST) are vital parameters in land surface-atmosphere interactions. Constrained by technology and atmospheric interferences, LST retrievals from various satellite sensors usually return missing data, thus negatively impacting analyses. Reconstructing missing data is important for acquiring gap-free datasets. However, the current reconstruction methods are limited for maintaining spatial details and high accuracies. We developed a new gap-free algorithm termed the spatial feature-considered random forest regression (SFRFR) model; it builds stable nonlinear relationships to connect the LST with related parameters, including terrain elements, land coverage types, spectral indexes, surface reflectance data, and the spatial feature of the LST, to reconstruct the missing LST data. The SFRFR model reconstructed gap-free LST data retrieved from the Landsat 8 satellite on 27 July 2017 in Wuhan. The results show that the SFRFR model exhibits the best performance according to the various evaluation metrics among the SFRFR, random forest regression and spline interpolation, with a coefficient of determination (R²) reaching 0.96, root-mean-square error (RMSE) of 0.55, and mean absolute error (MAE) of 0.55. Then, we reconstructed gap-free LST data gathered in Wuhan from 2016 to 2021 to analyze urban thermal environment changes and found that 2020 presented the coolest temperatures. The SFRFR model still displayed satisfactory results, with an average R² of 0.91 and an MAE of 0.63. We further discuss and discover the factors affecting the visual performance of SFRFR and identify the research priority to circumvent these disadvantages. Overall, this study provides a simple, practical method for acquiring gap-free LST data to help us better understand the spatiotemporal LST variation process.

Deducing Leading Factors of Spatial Distribution of Carbon Reserves in Nanjing Metropolitan Area Based on Random Forest Model

Improving carbon reserves is considered to be an important way to alleviate global warming. However, there is a lack of research work based on the perspective of metropolitan area, and there is also a lack of analysis on the leading influencing factors of spatial distribution of carbon storage in subregions of metropolitan area. In this study, Nanjing metropolitan area (NMA) is taken as the research area, and the InVEST model is used to calculate the spatial distribution of regional carbon reserves, and the evolution of carbon reserves distribution in recent 20 years is analyzed. Then, based on the random forest (RF) model, taking the whole study area and subareas as the research scope, a regression model of each selected impact factor and carbon reserves is established, and the leading factors of spatial distribution of carbon reserves in NMA are obtained. The results show that the overall carbon reserves level in the study area is in a downward trend. Through the application of the RF model, the leading factors of the spatial distribution of carbon reserves in NMA and its subareas are derived. The research proves that the application of the RF model in the analysis is helpful for city planners and governments to make plans and improve regional carbon storage more effectively.

A Global 250-m Downscaled NDVI Product from 1982 to 2018

Satellite-based normalized difference vegetation index (NDVI) time series data are useful for monitoring the changes in vegetation ecosystems in the context of global climate change. However, most of the current NDVI products cannot effectively reconcile high spatial resolution and continuous observations in time. Here, to produce a global-scale, long-term, and high-resolution NDVI database, we developed a simple and new data downscaling approach. The downscaling algorithm considers the pixel-wise ratios of the coefficient of variation (CV) between the coarse- and fine-resolution NDVI data and relative changes in the NDVI against a baseline period. The algorithm successfully created a worldwide monthly NDVI database with 250 m resolution from 1982 to 2018 by translating the fine spatial information from MODIS (Moderate-resolution Imaging Spectroradiometer) data and the long-term temporal information from AVHRR (Advanced Very High Resolution Radiometer) data. We employed the evaluation indices of root mean square error (RMSE), mean absolute error (MAE), and Pearson’s correlation coefficient (Pearson’s R) to assess the accuracy of the downscaled data against the MODIS NDVI. Both the RMSE and MAE values at the regional and global scales are typically between 0 and 0.2, whereas the Pearson’s R values are mostly above 0.7, which implies that the downscaled NDVI product is similar to the MODIS NDVI product. We then used the downscaled data to monitor the NDVI changes in different plant types and places with significant vegetation heterogeneity, as well as to investigate global vegetation trends over the last four decades. The Google Earth Engine platform was used for all the data downscaling processes, and here we provide a code for users to easily acquire data corresponding to any part of the world. The downscaled global-scale NDVI time series has high potential for the monitoring of the long-term temporal and spatial dynamics of terrestrial ecosystems under changing environments.

Geospatial Artificial Intelligence (GeoAI) in the Integrated Hydrological and Fluvial Systems Modeling: Review of Current Applications and Trends

This paper reviews the current GeoAI and machine learning applications in hydrological and hydraulic modeling, hydrological optimization problems, water quality modeling, and fluvial geomorphic and morphodynamic mapping. GeoAI effectively harnesses the vast amount of spatial and non-spatial data collected with the new automatic technologies. The fast development of GeoAI provides multiple methods and techniques, although it also makes comparisons between different methods challenging. Overall, selecting a particular GeoAI method depends on the application’s objective, data availability, and user expertise. GeoAI has shown advantages in non-linear modeling, computational efficiency, integration of multiple data sources, high accurate prediction capability, and the unraveling of new hydrological patterns and processes. A major drawback in most GeoAI models is the adequate model setting and low physical interpretability, explainability, and model generalization. The most recent research on hydrological GeoAI has focused on integrating the physical-based models’ principles with the GeoAI methods and on the progress towards autonomous prediction and forecasting systems.

Step-By-Step Downscaling of Land Surface Temperature Considering Urban Spatial Morphological Parameters

Land surface temperature (LST) is one of the most important parameters in urban thermal environmental studies. Compared to natural surfaces, the surface of urban areas is more complex, and the spatial variability of LST is higher. Therefore, it is important to obtain a high-spatial-resolution LST for urban thermal environmental research. At present, downscaling studies are mostly performed from a low spatial resolution directly to another high resolution, which often results in lower accuracy with a larger scale span. First, a step-by-step random forest downscaling LST model (SSRFD) is proposed in this study. In our work, the 900-m resolution Sentinel-3 LST was sequentially downscaled to 450 m, 150 m and 30 m by SSRFD. Then, urban spatial morphological parameters were introduced into SSRFD, abbreviated as SSRFD-M, to compensate for the deficiency of remote-sensing indices as driving factors in urban downscaling LST. The results showed that the RMSE value of the SSRFD results was reduced from 2.6 °C to 1.66 °C compared to the direct random forest downscaling model (DRFD); the RMSE value of the SSRFD-M results in built-up areas, such as Gulou and Qinhuai District, was reduced by approximately 0.5 °C. We also found that the underestimation of LST caused by considering only remote-sensing indices in places such as flowerbeds and streets was improved in the SSRFD-M results.

Spatiotemporal Analysis of Urban Heat Islands in Relation to Urban Development, in the Vicinity of the Atacama Desert

Near the Atacama Desert, Tacna city in Peru is among the largest arid cities with constant urban development, thus understanding of the urban surface thermal pattern is needed. We propose a comprehensive study of the urban heat island phenomenon, with the objective of (1) determining the spatial and temporal variations of the urban heat islands (UHIs), in the period 1985 to 2020; (2) analyzing the relationship between the UHI and influencing factors such as vegetation, urban area, and population, using indices calculated with satellite images. The Google Earth Engine repository was used to evaluate the corrected images from the years 1985 to 2020. The coincidence between the normalized difference vegetation index (NDVI) and normalized difference built-up index (NDBI) was good, negative between NDVI and the land surface temperature (LST), attributable to dense vegetation, and negative and very high (−0.81) between NDBI and NDVI, as massive urbanization leads to the reduction in the vegetal surface. The NDBI has a high impact on the LST; a coefficient of connections is recorded as 0.46. Tacna is a very arid region, and an increase in the time of the LST occurred with the increase in industrialization and urbanization. The land use/cover change (LUCC) evidences change in the climate in the city of Tacna; temperatures of 24.2 °C to 44.2 °C are observed in the built-up areas. In vegetated areas, the temperature remains below 24 °C, which is associated with a high rate of potential evapotranspiration. Thus, this study shows that variations in urban form and growth have produced the development of intraurban surface thermal patterns.

A Framework for Generating High Spatiotemporal Resolution Land Surface Temperature in Heterogeneous Areas

Land surface temperature (LST) is a crucial biophysical parameter related closely to the land–atmosphere interface. Satellite thermal infrared measurement provides an effective method to derive LST on regional and global scales, but it is very hard to acquire simultaneously high spatiotemporal resolution LST due to its limitation in the sensor design. Recently, many LST downscaling and spatiotemporal image fusion methods have been widely proposed to solve this problem. However, most methods ignored the spatial heterogeneity of LST distribution, and there are inconsistent image textures and LST values over heterogeneous regions. Thus, this study aims to propose one framework to derive high spatiotemporal resolution LSTs in heterogeneous areas by considering the optimal selection of LST predictors, the downscaling of MODIS LST, and the spatiotemporal fusion of Landsat 8 LST. A total of eight periods of MODIS and Landsat 8 data were used to predict the 100-m resolution LST at prediction time tp in Zhangye and Beijing of China. Further, the predicted LST at tp was quantitatively contrasted with the LSTs predicted by the regression-then-fusion strategy, STARFM-based fusion, and random forest-based regression, and was validated with the actual Landsat 8 LST product at tp. Results indicated that the proposed framework performed better in characterizing LST texture than the referenced three methods, and the root mean square error (RMSE) varied from 0.85 K to 2.29 K, and relative RMSE varied from 0.18 K to 0.69 K, where the correlation coefficients were all greater than 0.84. Furthermore, the distribution error analysis indicated the proposed new framework generated the most area proportion at 0~1 K in some heterogeneous regions, especially in artificial impermeable surfaces and bare lands. This means that this framework can provide a set of LST dataset with reasonable accuracy and a high spatiotemporal resolution over heterogeneous areas.