Satellite Soil Observation (SatSoil): extraction of bare soil reflectance for soil organic carbon mapping on Google Earth Engine

Temporal mosaicking approaches of Sentinel-2 images for extending topsoil organic carbon content mapping in croplands

Spatial information on soil organic carbon (SOC) in the Indian Himalayan region is vital for sustainable land management and conservation efforts, as it helps to identify areas vulnerable to erosion and degradation. However, the difficult Himalayan terrain is a major roadblock to the very high spatial resolution mapping of SOC. Previous studies mapped SOC in Himalayan region with resolutions varying from 250 to 90 m. The increase in spatial resolution definitely enhances the data quality and supports towards decision making for sustainable soil management. The present study tried to overcome this challenge and mapping of SOC was done at a resolution of 30 m by integrating various machine learning (ML) techniques i.e. random forest regression (RF), support vector regression (SVR) and extreme gradient boosting (XGB).Surface soil samples were strategically collected from 421 georeferenced locations representing the dominant elevation zones, geology and land use land cover (LULC) types to develop spatial models for predicting SOC. Environmental covariates representing various pedogenic factors, namely climatic variables, terrain attributes, spectral indices, LULC as well as lithological information were generated using diverse data sources employing geospatial data analysis and google earth engine (GEE) platform. A feature ranking and variable selection protocol was used for the selection of optimal set of covariates prior to model development. The three validated models were used for mapping the spatial distribution of SOC in the study area. The pixel wise uncertainty in SOC prediction by different models were also spatially mapped by generating the lower (LL) and upper limits (UL) of the 90% prediction interval.Results revealed that the RF model (R2train: 0.92, R2test: 0.72) performed better compared to XGB (R2train: 0.59, R2test: 0.37) and SVR (R2train: 0.53, R2test: 0.35) models during both training and testing phases indicated by various evaluation metrics. Covariates representing vegetation, climate as well as topography were found to equally dominate (03 nos. each) among the first 10 important predictors governing the better prediction performance of RF model. The SOC maps indicated soils in the north, north-east and north-western parts of the study area were exhibited comparatively higher SOC contents than rest of the study area. Among the predominant land use types, evergreen forest was found to exhibit the highest SOC values (2.88%) compared to others.The study demonstrated the potential of digital soil mapping (DSM) techniques, enhanced by RS and ML, for mapping SOC in the hilly tracts of Himalayas with a 30 m spatial resolution. This detailed database can prove beneficial in devising effective land management and resource conservation strategies in the fragile mountain ecosystems of Indian Himalayas.

Since the onset of agriculture, soils have lost their organic carbon to such an extent that the soil functions of many croplands are threatened. Hence, there is a strong demand for mapping and monitoring critical soil properties and in particular soil organic carbon (SOC). Pilot studies have demonstrated the potential for remote sensing techniques for SOC mapping in croplands. It has, however, been shown that the assessment of SOC may be hampered by the condition of the soil surface. While growing vegetation can be readily detected by means of the well-known Normalized Difference Vegetation Index (NDVI), the distinction between bare soil and crop residues is expressed in the shortwave infrared region (SWIR), which is only covered by two broad bands in Landsat or Sentinel-2 imagery. Here we tested the effect of thresholds for the Cellulose Absorption Index (CAI), on the performance of SOC prediction models for cropland soils. Airborne Prism Experiment (APEX) hyperspectral images covering an area of 240 km2 in the Belgian Loam Belt were used together with a local soil dataset. We used the partial least square regression (PLSR) model to estimate the SOC content based on 104 georeferenced calibration samples (NDVI < 0.26), firstly without setting a CAI threshold, and obtained a satisfactory result (coefficient of determination (R2) = 0.49, Ratio of Performance to Deviation (RPD) = 1.4 and Root Mean Square Error (RMSE) = 2.13 g kgC−1 for cross-validation). However, a cross comparison of the estimated SOC values to grid-based measurements of SOC content within three fields revealed a systematic overestimation for fields with high residue cover. We then tested different CAI thresholds in order to mask pixels with high residue cover. The best model was obtained for a CAI threshold of 0.75 (R2 = 0.59, RPD = 1.5 and RMSE = 1.75 g kgC−1 for cross-validation). These results reveal that the purity of the pixels needs to be assessed aforehand in order to produce reliable SOC maps. The Normalized Burn Ratio (NBR2) index based on the SWIR bands of the MSI Sentinel 2 sensor extracted from images collected nine days before the APEX flight campaign correlates well with the CAI index of the APEX imagery. However, the NBR2 index calculated from Sentinel 2 images under moist conditions is poorly correlated with residue cover. This can be explained by the sensitivity of the NBR2 index to both soil moisture and residues.

<p>Since the onset of agriculture, soils have lost their organic carbon to such an extent that the soil functions of many croplands are threatened and there is therefore a strong demand for mapping and monitoring critical soil properties and in particular soil organic carbon (SOC). Pilot studies have demonstrated the potential for remote sensing techniques for SOC mapping in croplands, given their large spatial coverage and high temporal resolution. It has however been shown that the assessment of SOC may be hampered by crop residues. In this study we tested the effect of the threshold for the cellulose absorption index (CAI), on the performance of SOC prediction models for bare cropland soils. Airborne Prism Experiment (APEX) hyperspectral images covering an area of 230 km<sup>2</sup> in the Belgian Loam Belt were used together with a local soil dataset. We used the partial least square regression (PLSR) model to estimate the SOC content based on 104 georeferenced calibration samples, firstly without setting a CAI threshold, and obtained a satisfactory result (R²=0.49, RPD=1.4 and RMSE=2.14 g kgC<sup>-1</sup> for cross-validation). However, a cross comparison of the estimated SOC values to grid-based measurements of SOC content within three fields revealed a systematic overestimation for fields with high residue cover. We then tested different thresholds of CAI in order to mask pixels with high residue cover, by eliminating calibration samples used in the PLSR model based on this threshold. The best model was obtained for CAI threshold of 0.8 (R²=0.59, RPD=1.5 and RMSE=1.76 g kgC<sup>-1</sup> for cross-validation). These results reveal that the purity of the pixels needs to be assessed aforehand in order to produce reliable SOC maps. Preliminary results indicate that an index based on the SWIR bands of the MSI Sentinel 2 sensor is also capable of detecting crop residues. However, the application under moist conditions and for different types of residues needs to be confirmed.</p>

Digital soil organic carbon (SOC) mapping is used for ecological protection and addressing global climate change. Sentinel-1 (S-1) microwave radar remote sensing data offer critical insights into SOC dynamics through tracking variations in soil moisture and vegetation characteristics. Despite extensive studies using S-1 data for SOC mapping, most focus on either single or multi-date periods without achieving satisfactory results. Few studies have investigated the potential of time-series S-1 data for high-accuracy SOC mapping. This study utilized S-1 data from 2017 to 2021 to analyze temporal variations in the correlation between SOC and time-series S-1 data in southern Xinjiang, China. The primary objective was to determine the optimal monitoring period for SOC. Within this period, optimal feature subsets were extracted using variable selection algorithms. The performance of the partial least squares regression, random forest, and convolutional neural network–long short-term memory (CNN-LSTM) models was evaluated using a 10-fold cross-validation approach. The findings revealed the following: (1) The correlation between time-series S-1 data and SOC exhibited both interannual and monthly variations, with the optimal monitoring period from July to October. The data volume was reduced by 73.27% relative to the initial time-series dataset when the optimal monitoring period was determined. (2) Introducing time-series S-1 data into SOC mapping significantly improved CNN-LSTM model performance (R2 = 0.80, RPD = 2.24, RMSE = 1.11 g kg⁻1). Compared to models using single-date (R2 = 0.23) and multi-date (R2 = 0.33) data, the R2 increased by 0.57 and 0.47, respectively. (3) The newly developed vertical–horizontal maximum and mean annual cumulative indices made a significant contribution (17.93%) to mapping SOC. Therefore, integrating the optimal monitoring period, feature selection, and deep learning model offers significant potential for enhancing the accuracy of digital SOC mapping.

Towards spatially continuous mapping of soil organic carbon in croplands using multitemporal Sentinel-2 remote sensing

The Namibian Soil Profile Database contains 4960 entries, all samples with geographic coordinates. Each soil property presents a different number of observations, which decreases with depth. To perform the Digital Soil Map of Soil Organic Carbon (SOC) up to 30 cm, 1298 sample points were used. The covariates used in the model were composed from land cover, geology, terrain characteristics extracted from the digital elevation model and remote sensing data. Most of the covariates carry 30 m of spatial resolution. The Random Forest model implemented in Google Earth Engine (GEE) was applied with an external validation split of 80/20 %. As the second validation layer, samples from the Namibian tier of the Soils4Africa project were applied. The use of GEE facilitated the generation of a SOC distribution map for Namibia at a spatial resolution of 30 × 30 m. The highest amounts of SOC are stored in the central region of Namibia, with SOC values ranging from 0.1 to 1.9 %. The map is suitable for national and regional decision-making, offering a baseline for determining SOC stocks, monitoring changes in SOC, assessing the effects of bush encroachment/thickening and bush control, and for agriculture implications. Mapping of soil properties in other depths are scheduled for the near future.

Soil organic carbon (SOC) plays a pivotal role in the functioning of terrestrial ecosystems, has the potential to mitigate climate change and provides several benefits for soil health. Understanding the spatial distribution of SOC can possibly help formulate sustainable soil management practices. Conventional soil surveys are often limited by their spatio-temporal resolution and high cost, necessitating the development of innovative techniques that can capture the intricate variability of SOC across landscapes. In response to this need, digital soil mapping (DSM) has emerged as a powerful approach that uses advanced geospatial technologies and statistical methods to predict soil properties across large areas. Predictor variables for DSM include climate data, topographical features, geological attributes, legacy soil maps, land management practices, spatial information and remote sensing data. The spectral response of bare soil, measured by multispectral satellite sensors, can be an adequate predictor of SOC and texture at the field scale and in small regions, but its use for the assessment of soil properties at large scale (thousands of km²) has been less explored 1. In this study, bare soil spectra derived from Sentinel-2 were used to estimate SOC and texture across agricultural parcels in Flanders, northern Belgium (n=169-175). Five different machine learning models were tested: generalized linear regression (GLM), partial least squares regression (PLSR), random forest (RF), cubist regression (CR) and gradient boosting machine (GBM). The SOC prediction of a DSM model using bare soil spectra was compared with that of a DSM model using environmental covariates: topography (elevation, slope and compound topographic index), climate (average annual temperature, total annual precipitation, average annual evapotranspiration), texture (sand, silt and clay content), vegetation (proportion of the year the soil is covered by vegetation) and location. The predictive performances of these models were compared to a DSM model that included both the bare soil spectra and the environmental covariates. Soil texture (sand, silt, clay) was adequately predicted using the bare soil spectra from the spring seedbed (R²: 0.53-0.71; RPD: 1.54-2.18; RPIQ: 1.36-2.41), but the predictive performance for SOC was poor (R²: 0.19; RPD: 1.07; RPIQ: 1.45). All three DSM models showed poor predictive performance for SOC, with the best performance for the model including all covariates (R²: 0.26; RPD: 1.25; RPIQ: 1.68). For the DSM model from bare soil spectra (PLSR), all Sentinel-2 spectral bands showed high relative importance except bands 2 (blue) and 3 (green). For the DSM model from environmental covariates (GBM), vegetation cover and topography explained most of the variation in SOC. The DSM model including all variables (GBM) showed a low influence of bare soil spectral bands, but a high influence of previous vegetation cover and topography. These results showed the importance of terrain characteristics and vegetation for assessing large scale SOC distribution. The overall low predictive performance for SOC obtained in this study indicates the complex nature of factors influencing SOC distribution across a large region and highlights the need for more in-depth high resolution studies.1 https://doi.org/10.3390/rs14122917

On the effectiveness of multi-scale landscape metrics in soil organic carbon mapping

National soil organic carbon (SOC) maps are essential to improve greenhouse gas accounting and support climate-smart agriculture. Large-scale SOC models based on wall-to-wall soil information from remote sensing remain a challenge due to the high diversity of natural soil conditions and the difficulty of accounting for the spatial location of the soil samples. In this study, we tested if the implementation of local ensemble models (LEM) can be used to improve the SOC predictions from Landsat-based soil reflectance composites (SRC) for Germany. For this, we divided the research area into 30 times 30 km tiles and calculated local generalized linear models (GLM) based on random, nearby observations. Based on the GLMs, local SOC maps were predicted and aggregated using a moving window approach. The local variable importance was analyzed to identify spatial dependencies in the correlation between the SRC and SOC. For the final SOC map, a Random Forest (RF) model was trained using the aggregated local SOC predictions, the SRC, and a full set of training samples from the agricultural soil inventory. The results show that the LEM was able to improve the accuracy (R2 = 0.68; RMSE = 5.6 g kg−1), compared to the maps based on a single, global model (R2 = 0.52; RMSE = 6.8 g kg−1). The local variable importance of the spectral bands showed clear spatial patterns throughout the research area. Differences can be explained by the local soil conditions, influencing the correlation between SOC and the spectral properties. Compared to the widely adopted integration of distance covariates such as geographical coordinates, the LEM was able the reduce the spatial autocorrelation to a greater extent and to improve the prediction accuracy, especially for underrepresented SOC values. The LEM presents a new method to integrate spatial information and increase the interpretability of DSM models.

Using broadband crop residue angle index to estimate the fractional cover of vegetation, crop residue, and bare soil in cropland systems

Remote estimates of soil organic carbon using multi-temporal synthetic images and the probability hybrid model

Effects of optical and radar satellite observations within Google Earth Engine on soil organic carbon prediction models in Spain

High‐resolution, field‐scale soil organic carbon (SOC) mapping in croplands is crucial for effective and precise agricultural management. Recent developments in unmanned aerial vehicles (UAVs) combined with miniaturized visible–near infrared spectrometers have enabled the rapid and low‐cost field‐scale SOC mapping. However, a field‐specific spectrotransfer model is often needed for such UAV‐based hyperspectral measurements, implying local sampling and model development are still required, and this hampers the widespread application of UAV‐based methods. In this study, we aim to test to what extent SOC prediction models derived from an existing regional soil spectral library (SSL) can be applied to UAV‐based hyperspectral data, without the need for additional field sampling. To this end, an UAV survey was conducted over a bare cropland within the Belgian Loam Belt for field‐scale SOC mapping. We evaluated two calibration approaches, one based on local sampling and model development, and one where we capitalized on an existing (laboratory‐based) regional SSL. For the local calibration approach, we obtained a good prediction performance with RMSE of 0.57 g kg−1 and RPIQ of 2.35. For the regional model, a spectral alignment procedure was needed to resolve the discrepancy between UAV‐ and laboratory‐based measurements. This resulted in a fair SOC prediction accuracy with RMSE of 0.93 g kg−1 and RPIQ of 1.45. The comparison of SOC maps derived from the two approaches, along with an external validation showed a high consistency, indicating that UAV‐based spectral measurements, in combination with SSLs have the potential to improve the efficiency of high‐resolution SOC mapping.

Mapping soil organic carbon content over New South Wales, Australia using local regression kriging