Davide Consoli
Davide Consoli has a total of 82 co-authors. Here are the top co-authors by number of shared publications:
- Leandro Parente
OpenGeoHub Foundation, Doorwerth, The Netherlands
15 shared publications - Tomislav Hengl
OpenGeoHub Foundation, Doorwerth, The Netherlands
12 shared publications - Yu-Feng Ho
OpenGeoHub Foundation, Doorwerth, The Netherlands
7 shared publications Murat Şahin
6 shared publications- Mustafa Serkan Isik
OpenGeoHub Foundation, Doorwerth, The Netherlands
6 shared publications - Laerte Guimarães Ferreira
Remote Sensing and GIS Laboratory (LAPIG), Federal University of Goiás (UFG), Goiania 74001-970, Brazil
5 shared publications
Latest Publications
- Top Cited PapersOpen Access
- 6 February 2026
- Published by Copernicus GmbH in Earth System Science Data
- Vol. 18 (2) , 989-1036
OpenLandMap-soildb: global soil information at 30 m spatial resolution for 2000–2022+ based on spatiotemporal Machine Learning and harmonized legacy soil samples and observations
There is increasing interest in global dynamic soil information with changes in soil properties mapped over time and at high spatial resolution. Thanks to long-term, multi-temporal, and fine- and medium-resolution satellite missions such as Landsat, MODIS, Copernicus Sentinel and similar, it is possible to produce globally consistent predictions of key soil variables that match other 10–30 m spatial resolution global data sets. This paper describes data preparation, modeling, and production of OpenLandMap-soildb: global dynamic predictions of soil organic carbon content, soil organic carbon density, bulk density, soil pH in H2O, soil texture fractions (clay, sand and silt) and USDA subgroup soil types (USDA soil taxonomy subgroups) at 30 m spatial resolution based on spatiotemporal Machine Learning (Quantile Regression Random Forest with output predictions showing the mean plus the 68 % probability lower and upper prediction intervals). To train the models, a large compilation of soil samples imported from legacy soil projects was used: 216 000 soil samples with soil carbon density (kg m−3), 408 000 soil samples with soil carbon content (g kg−1), 272 000 soil samples with soil pH in H2O, 363 000 soil samples with clay, silt and sand content (%) and 134 000 samples with bulk density oven dry (t m−3). Soil carbon and soil pH were mapped with 5-year time-intervals; soil texture fractions, bulk density, and soil types were mapped for recent years only. The cross-validation results indicate Root Mean Square Error (RMSE) of 17.7 (kg m−3; 0.486 in log-scale) and Concordance Correlation Coefficient (CCC) of 0.88 for SOC density, RMSE of 51.3 (g kg−1; 0.574 in log-scale) and CCC of 0.87 for SOC content, RMSE of 0.15 (t m−3) and CCC of 0.92 for bulk density of fine-earth, RMSE of 0.51 and CCC of 0.91 for soil pH, RMSE of 8.4 % and CCC of 0.87 for soil clay content, and RMSE of 12.6 % and CCC of 0.84 for soil sand content respectively. The most important variables for predicting soil organic carbon density (kg m−3) were: soil depth, Landsat-based uncalibrated Gross Primary Productivity (GPP), Normalized Difference Vegetation Index (NDVI) and CHELSA bioclimatic indices. The global distribution of soil pH can be primarily explained by the CHELSA Aridity Index (long-term), annual precipitation, and salinity grade. The global stocks for 2020–2022+ period for 0–30 cm depth interval are estimated at 461 Pg (Peta grams); the results further indicate that, in the last 25 years, the world has lost at least 11 Pg of SOC in the top soil. Suggestions are made on how to set up global permanent monitoring stations to accurately track land degradation and enable land restoration projects. The training data set is available at https://doi.org/10.5281/zenodo.4748499 (Hengl and Gupta, 2025), while the resulting data products can be accessed at https://doi.org/10.5281/zenodo.15470431 (Consoli et al., 2025) and https://world.soils.app (OpenGeoHub Foundation, 2026). Both datasets are released under a CC-BY license.
- 6 February 2026
- Published by Copernicus GmbH in Earth System Science Data
- Vol. 18 (2) , 989-1036
Top Cited PapersOpen AccessOpenLandMap-soildb: global soil information at 30 m spatial resolution for 2000–2022+ based on spatiotemporal Machine Learning and harmonized legacy soil samples and observations
There is increasing interest in global dynamic soil information with changes in soil properties mapped over time and at high spatial resolution. Thanks to long-term, multi-temporal, and fine- and medium-resolution satellite missions such as Landsat, MODIS, Copernicus Sentinel and similar, it is possible to produce globally consistent predictions of key soil variables that match other 10–30 m spatial resolution global data sets. This paper describes data preparation, modeling, and production of OpenLandMap-soildb: global dynamic predictions of soil organic carbon content, soil organic carbon density, bulk density, soil pH in H2O, soil texture fractions (clay, sand and silt) and USDA subgroup soil types (USDA soil taxonomy subgroups) at 30 m spatial resolution based on spatiotemporal Machine Learning (Quantile Regression Random Forest with output predictions showing the mean plus the 68 % probability lower and upper prediction intervals). To train the models, a large compilation of soil samples imported from legacy soil projects was used: 216 000 soil samples with soil carbon density (kg m−3), 408 000 soil samples with soil carbon content (g kg−1), 272 000 soil samples with soil pH in H2O, 363 000 soil samples with clay, silt and sand content (%) and 134 000 samples with bulk density oven dry (t m−3). Soil carbon and soil pH were mapped with 5-year time-intervals; soil texture fractions, bulk density, and soil types were mapped for recent years only. The cross-validation results indicate Root Mean Square Error (RMSE) of 17.7 (kg m−3; 0.486 in log-scale) and Concordance Correlation Coefficient (CCC) of 0.88 for SOC density, RMSE of 51.3 (g kg−1; 0.574 in log-scale) and CCC of 0.87 for SOC content, RMSE of 0.15 (t m−3) and CCC of 0.92 for bulk density of fine-earth, RMSE of 0.51 and CCC of 0.91 for soil pH, RMSE of 8.4 % and CCC of 0.87 for soil clay content, and RMSE of 12.6 % and CCC of 0.84 for soil sand content respectively. The most important variables for predicting soil organic carbon density (kg m−3) were: soil depth, Landsat-based uncalibrated Gross Primary Productivity (GPP), Normalized Difference Vegetation Index (NDVI) and CHELSA bioclimatic indices. The global distribution of soil pH can be primarily explained by the CHELSA Aridity Index (long-term), annual precipitation, and salinity grade. The global stocks for 2020–2022+ period for 0–30 cm depth interval are estimated at 461 Pg (Peta grams); the results further indicate that, in the last 25 years, the world has lost at least 11 Pg of SOC in the top soil. Suggestions are made on how to set up global permanent monitoring stations to accurately track land degradation and enable land restoration projects. The training data set is available at https://doi.org/10.5281/zenodo.4748499 (Hengl and Gupta, 2025), while the resulting data products can be accessed at https://doi.org/10.5281/zenodo.15470431 (Consoli et al., 2025) and https://world.soils.app (OpenGeoHub Foundation, 2026). Both datasets are released under a CC-BY license.
- Open Access
- 21 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Imagined Flickering Patterns: A Visual Imagery Paradigm for BCI Strategies
Steady state visually evoked potentials (SSVEPs) are a popular type of control signals in brain-computer interfaces (BCIs), in which they are typically elicited by observing a visual stimulus flashing at a specific frequency. For some patients, using SSVEP as control signal for a BCI can be difficult, for instance if they are unable to focus their gaze over the visual stimuli. To address this issue, some approaches were presented to design a gaze-independent SSVEP-controlled BCI but some difficulties have been reported, for instance for patients suffering from locked-in syndrome. In this work we employ a visual imagery (VI) signal, in which the visual stimulus is imagined instead of observed, to drive a BCI system and offer an alternative for patients that encounter issues with standard SSVEP approaches. We tested the proposed approach with 20 untrained subjects within a 3-classes BCI resulting in an offline classification accuracy of 60.93% These results demonstrate how this gaze-independent BCI can be used by inexperienced BCI users.
- 21 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessImagined Flickering Patterns: A Visual Imagery Paradigm for BCI Strategies
Steady state visually evoked potentials (SSVEPs) are a popular type of control signals in brain-computer interfaces (BCIs), in which they are typically elicited by observing a visual stimulus flashing at a specific frequency. For some patients, using SSVEP as control signal for a BCI can be difficult, for instance if they are unable to focus their gaze over the visual stimuli. To address this issue, some approaches were presented to design a gaze-independent SSVEP-controlled BCI but some difficulties have been reported, for instance for patients suffering from locked-in syndrome. In this work we employ a visual imagery (VI) signal, in which the visual stimulus is imagined instead of observed, to drive a BCI system and offer an alternative for patients that encounter issues with standard SSVEP approaches. We tested the proposed approach with 20 untrained subjects within a 3-classes BCI resulting in an offline classification accuracy of 60.93% These results demonstrate how this gaze-independent BCI can be used by inexperienced BCI users.
- Open Access
- 23 August 2025
- Published by Springer Nature in Scientific Data
- Vol. 12 (1) , 1-17
Global 30-m annual median vegetation height maps (2000–2022) based on ICESat-2 data and Machine Learning
Accurately measuring vegetation height is essential for understanding ecosystem structure, carbon storage, and biodiversity, yet global height models have overwhelmingly focused on forests, excluding ecosystems with shorter herbaceous vegetation or shrubs. To address this gap in vegetation structure data, we developed the first global estimate of median vegetation height annually from 2000–2022 at 30 m resolution, using ICESat-2 satellite Lidar, Landsat cloud free composites, and other Earth Observation raster data. Thirty two (32) million ICESat-2 20 m segments were used within 10 independent draws to build ensemble Gradient Boosted Tree (GBT) models and estimate 90% prediction intervals. Our model achieves a root mean square error (RMSE) of 2.35 m, R2 values of 0.515 and a D2 regression score of 0.62 estimated on the testing set. Comparisons with existing global height products show that our approach increases detail and heterogeneity of height in short vegetation ecosystems. Output maps are publicly available together with reference samples and trained models under CC-BY license.
- 23 August 2025
- Published by Springer Nature in Scientific Data
- Vol. 12 (1) , 1-17
Open AccessGlobal 30-m annual median vegetation height maps (2000–2022) based on ICESat-2 data and Machine Learning
Accurately measuring vegetation height is essential for understanding ecosystem structure, carbon storage, and biodiversity, yet global height models have overwhelmingly focused on forests, excluding ecosystems with shorter herbaceous vegetation or shrubs. To address this gap in vegetation structure data, we developed the first global estimate of median vegetation height annually from 2000–2022 at 30 m resolution, using ICESat-2 satellite Lidar, Landsat cloud free composites, and other Earth Observation raster data. Thirty two (32) million ICESat-2 20 m segments were used within 10 independent draws to build ensemble Gradient Boosted Tree (GBT) models and estimate 90% prediction intervals. Our model achieves a root mean square error (RMSE) of 2.35 m, R2 values of 0.515 and a D2 regression score of 0.62 estimated on the testing set. Comparisons with existing global height products show that our approach increases detail and heterogeneity of height in short vegetation ecosystems. Output maps are publicly available together with reference samples and trained models under CC-BY license.
- Open Access
Light use efficiency (LUE) based bimonthly gross primary productivity (GPP) for global grasslands at 30 m spatial resolution (2000–2022)
The article describes production of a high spatial resolution (30 m) bimonthly light use efficiency (LUE) based gross primary productivity (GPP) data set representing grasslands for the period 2000 to 2022. The data set is based on using reconstructed global complete consistent bimonthly Landsat archive (400TB of data), combined with 1 km MOD11A1 temperature data and 1° CERES Photosynthetically Active Radiation (PAR). First, the LUE model was implemented by taking the biome-specific productivity factor (maximum LUE parameter) as a global constant, producing a global bimonthly (uncalibrated) productivity data for the complete land mask. Second, the GPP 30 m bimonthly maps were derived for the global grassland annual predictions and calibrating the values based on the maximum LUE factor of 0.86 gCm−2d−1MJ−1. The results of validation of the produced GPP estimates based on 527 eddy covariance flux towers show an R-square between 0.48–0.71 and root mean square error (RMSE) below ~2.3 gCm−2d−1 for all land cover classes. Using a total of 92 flux towers located in grasslands, the validation of the GPP product calibrated for the grassland biome revealed an R-square between 0.51–0.70 and an RMSE smaller than ~2 gCm−2d−1. The final time-series of maps (uncalibrated and grassland GPP) are available as bimonthly (daily estimates in units of gCm−2d−1) and annual (daily average accumulated by 365 days in units of gCm−2yr−1) in Cloud-Optimized GeoTIFF (~23TB in size) as open data (CC-BY license). The recommended uses of data include: trend analysis e.g., to determine where are the largest losses in GPP and which could be an indicator of potential land degradation, crop yield mapping and for modeling GHG fluxes at finer spatial resolution. Produced maps are available via SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine.
Open AccessLight use efficiency (LUE) based bimonthly gross primary productivity (GPP) for global grasslands at 30 m spatial resolution (2000–2022)
The article describes production of a high spatial resolution (30 m) bimonthly light use efficiency (LUE) based gross primary productivity (GPP) data set representing grasslands for the period 2000 to 2022. The data set is based on using reconstructed global complete consistent bimonthly Landsat archive (400TB of data), combined with 1 km MOD11A1 temperature data and 1° CERES Photosynthetically Active Radiation (PAR). First, the LUE model was implemented by taking the biome-specific productivity factor (maximum LUE parameter) as a global constant, producing a global bimonthly (uncalibrated) productivity data for the complete land mask. Second, the GPP 30 m bimonthly maps were derived for the global grassland annual predictions and calibrating the values based on the maximum LUE factor of 0.86 gCm−2d−1MJ−1. The results of validation of the produced GPP estimates based on 527 eddy covariance flux towers show an R-square between 0.48–0.71 and root mean square error (RMSE) below ~2.3 gCm−2d−1 for all land cover classes. Using a total of 92 flux towers located in grasslands, the validation of the GPP product calibrated for the grassland biome revealed an R-square between 0.51–0.70 and an RMSE smaller than ~2 gCm−2d−1. The final time-series of maps (uncalibrated and grassland GPP) are available as bimonthly (daily estimates in units of gCm−2d−1) and annual (daily average accumulated by 365 days in units of gCm−2yr−1) in Cloud-Optimized GeoTIFF (~23TB in size) as open data (CC-BY license). The recommended uses of data include: trend analysis e.g., to determine where are the largest losses in GPP and which could be an indicator of potential land degradation, crop yield mapping and for modeling GHG fluxes at finer spatial resolution. Produced maps are available via SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine.
- Open Access
- 24 June 2025
- Published by Copernicus GmbH in EGUsphere
- Vol. 2025, 1-66
OpenLandMap-soildb: global soil information at 30 m spatial resolution for 2000–2022+ based on spatiotemporal Machine Learning and harmonized legacy soil samples and observations
There is increasing interest in global dynamic soil information with changes in soil properties mapped over time and at high spatial resolution. Thanks to long-term, multi-temporal, and fine- and medium-resolution satellite missions such as Landsat, MODIS, Copernicus Sentinel and similar, it is possible to produce globally consistent predictions of key soil variables that match other 10–30 m spatial resolution global data sets. This paper describes data preparation, modeling, and production of OpenLandMap-soildb: global dynamic predictions of soil organic carbon content, soil organic carbon density, bulk density, soil pH in H2O, soil texture fractions (clay, sand and slit) and USDA subgroup soil types (USDA soil taxonomy subgroups) at 30 m spatial resolution based on spatiotemporal Machine Learning (Quantile Regression Random Forest with output predictions showing the mean plus the lower and upper prediction intervals of 68 % probability). To train the models, a large compilation of soil samples imported from legacy soil projects was used: 216,000 soil samples with soil carbon density (kg m-3), 408,000 soil samples with soil carbon content (g kg-1), 272,000 samples with soil pH in H2O, 363,000 samples with clay, silt, and sand (%), and 134,000 samples with bulk density oven dry (t m-3). Soil carbon and soil pH were mapped with 5 year time-intervals; soil texture fractions, bulk density, and soil types were mapped for recent years only. The cross-validation results indicate RMSE of 17.7 (kg m-3; 0.486 in log-scale) and CCC of 0.88 for SOC density, RMSE of 51.3 (g kg-1; 0.574 in log-scale) and CCC of 0.87 for SOC content, RMSE of 0.15 (t m-3) and CCC of 0.92 for bulk density of fine-earth, RMSE of 0.51 and CCC of 0.91 for soil pH, RMSE of 8.4 % and CCC of 0.87 for soil clay content, and RMSE of 12.6 % and CCC of 0.84 for soil sand content respectively. The most important variables for predicting soil organic carbon density (kg m-3) were: soil depth, Landsat-based uncalibrated Gross Primary Productivity (GPP), Normalized Difference Vegetation Index (NDVI) and CHELSA bioclimatic indices. The global distribution of soil pH can be primarily explained by the CHELSA Aridity Index (long-term), annual precipitation, and salinity grade. The global stocks for 2020–2022+ period for 0–30 cm depth interval are estimated at 461 Pg (Peta grams); the results further indicate that, in the last 25 years, the world has lost at least 11 Pg of SOC in the top soil. Suggestions are made on how to set up global permanent monitoring stations to accurately track land degradation and enable land restoration projects. The training dataset is available at https://doi.org/10.5281/zenodo.4748499 (Hengl and Gupta, 2025), while the resulting data products can be accessed at https://doi.org/10.5281/zenodo.15470431 (Consoli et al., 2025). Both datasets are released under a CC-BY license.
- 24 June 2025
- Published by Copernicus GmbH in EGUsphere
- Vol. 2025, 1-66
Open AccessOpenLandMap-soildb: global soil information at 30 m spatial resolution for 2000–2022+ based on spatiotemporal Machine Learning and harmonized legacy soil samples and observations
There is increasing interest in global dynamic soil information with changes in soil properties mapped over time and at high spatial resolution. Thanks to long-term, multi-temporal, and fine- and medium-resolution satellite missions such as Landsat, MODIS, Copernicus Sentinel and similar, it is possible to produce globally consistent predictions of key soil variables that match other 10–30 m spatial resolution global data sets. This paper describes data preparation, modeling, and production of OpenLandMap-soildb: global dynamic predictions of soil organic carbon content, soil organic carbon density, bulk density, soil pH in H2O, soil texture fractions (clay, sand and slit) and USDA subgroup soil types (USDA soil taxonomy subgroups) at 30 m spatial resolution based on spatiotemporal Machine Learning (Quantile Regression Random Forest with output predictions showing the mean plus the lower and upper prediction intervals of 68 % probability). To train the models, a large compilation of soil samples imported from legacy soil projects was used: 216,000 soil samples with soil carbon density (kg m-3), 408,000 soil samples with soil carbon content (g kg-1), 272,000 samples with soil pH in H2O, 363,000 samples with clay, silt, and sand (%), and 134,000 samples with bulk density oven dry (t m-3). Soil carbon and soil pH were mapped with 5 year time-intervals; soil texture fractions, bulk density, and soil types were mapped for recent years only. The cross-validation results indicate RMSE of 17.7 (kg m-3; 0.486 in log-scale) and CCC of 0.88 for SOC density, RMSE of 51.3 (g kg-1; 0.574 in log-scale) and CCC of 0.87 for SOC content, RMSE of 0.15 (t m-3) and CCC of 0.92 for bulk density of fine-earth, RMSE of 0.51 and CCC of 0.91 for soil pH, RMSE of 8.4 % and CCC of 0.87 for soil clay content, and RMSE of 12.6 % and CCC of 0.84 for soil sand content respectively. The most important variables for predicting soil organic carbon density (kg m-3) were: soil depth, Landsat-based uncalibrated Gross Primary Productivity (GPP), Normalized Difference Vegetation Index (NDVI) and CHELSA bioclimatic indices. The global distribution of soil pH can be primarily explained by the CHELSA Aridity Index (long-term), annual precipitation, and salinity grade. The global stocks for 2020–2022+ period for 0–30 cm depth interval are estimated at 461 Pg (Peta grams); the results further indicate that, in the last 25 years, the world has lost at least 11 Pg of SOC in the top soil. Suggestions are made on how to set up global permanent monitoring stations to accurately track land degradation and enable land restoration projects. The training dataset is available at https://doi.org/10.5281/zenodo.4748499 (Hengl and Gupta, 2025), while the resulting data products can be accessed at https://doi.org/10.5281/zenodo.15470431 (Consoli et al., 2025). Both datasets are released under a CC-BY license.
- Open Access
- 27 May 2025
- Published by Springer Nature in Research Square
Light Use Efficiency (LUE) based bimonthly Gross Primary Productivity (GPP) for global grasslands at 30 m spatial resolution (2000-2022)
The paper describes production of a high spatial resolution (30 m) bimonthly Light Use Efficiency (LUE) based Gross Primary Productivity (GPP) data set representing grasslands for the period 2000 to 2022. The data set is based on using reconstructed global complete consistent bimonthly Landsat archive (400TB of data), combined with 1 km MOD11A1 temperature data and 1o CERES Photosynthetically Active Radiation (PAR). First, the LUE model was implemented by taking the biome-specific productivity factor (maximum LUE parameter) as a global constant, producing a global bimonthly (uncalibrated) productivity data for the complete land mask. Second, the GPP 30 m bimonthly maps were derived for the global grassland annual predictions and calibrating the values based on the maximum LUE factor of 0.86 gCm-2d-1MJ-1. The results of validation of the produced GPP estimates based on more than 500 eddy covariance flux towers show an R-square between 0.48-0.71 and RMSE bellow ~2.3 gCm-2d-1 for all land cover classes. Using only the flux towers located in grasslands, the validation of the GPP product calibrated for the grassland biome revealed an R-square between 0.51-0.70 and an RMSE smaller than 2 gCm-2d-1. The final time-series of maps (uncalibrated and grassland GPP) are available as bimonthly (daily estimates in units of gCm-2d-1) and annual (daily average accumulated by 365 days in units of gCm-2yr-1 ) in Cloud-Optimized GeoTIFF (~23TB in size) as open data (CC-BY license). The recommended uses of data include: trend analysis e.g. to determine where are the largest losses in GPP and which could be an indicator of potential land degradation, crop yield mapping and for modeling GHG fluxes at finer spatial resolution. Produced maps will be made available via the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters), which provides data in exchange for feedback.
- 27 May 2025
- Published by Springer Nature in Research Square
Open AccessLight Use Efficiency (LUE) based bimonthly Gross Primary Productivity (GPP) for global grasslands at 30 m spatial resolution (2000-2022)
The paper describes production of a high spatial resolution (30 m) bimonthly Light Use Efficiency (LUE) based Gross Primary Productivity (GPP) data set representing grasslands for the period 2000 to 2022. The data set is based on using reconstructed global complete consistent bimonthly Landsat archive (400TB of data), combined with 1 km MOD11A1 temperature data and 1o CERES Photosynthetically Active Radiation (PAR). First, the LUE model was implemented by taking the biome-specific productivity factor (maximum LUE parameter) as a global constant, producing a global bimonthly (uncalibrated) productivity data for the complete land mask. Second, the GPP 30 m bimonthly maps were derived for the global grassland annual predictions and calibrating the values based on the maximum LUE factor of 0.86 gCm-2d-1MJ-1. The results of validation of the produced GPP estimates based on more than 500 eddy covariance flux towers show an R-square between 0.48-0.71 and RMSE bellow ~2.3 gCm-2d-1 for all land cover classes. Using only the flux towers located in grasslands, the validation of the GPP product calibrated for the grassland biome revealed an R-square between 0.51-0.70 and an RMSE smaller than 2 gCm-2d-1. The final time-series of maps (uncalibrated and grassland GPP) are available as bimonthly (daily estimates in units of gCm-2d-1) and annual (daily average accumulated by 365 days in units of gCm-2yr-1 ) in Cloud-Optimized GeoTIFF (~23TB in size) as open data (CC-BY license). The recommended uses of data include: trend analysis e.g. to determine where are the largest losses in GPP and which could be an indicator of potential land degradation, crop yield mapping and for modeling GHG fluxes at finer spatial resolution. Produced maps will be made available via the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters), which provides data in exchange for feedback.
- Open Access
- 2 May 2025
- Published by Springer Nature in Research Square
Global 30-m annual median vegetation height maps (2000-2022) based on ICESat-2 data and Machine Learning
Accurately measuring vegetation height is essential for understanding ecosystem structure, carbon storage, and biodiversity, yet global height models have overwhelmingly focused on forests, excluding ecosystems with shorter herbaceous vegetation or shrubs. To address this gap in vegetation structure data, we developed the first global model to estimate median vegetation height annually from 2000-2022 at 30 m resolution, using ICESat-2 satellite Lidar, Landsat cloud free composites, and other Earth Observation raster data. Thirty (30) million ICESat-2 20 m segments were used within 10 independent draws to build ensemble Gradient Boosted Tree (GBT) models and estimate 90% prediction intervals. Our model achieves a root mean square error (RMSE) of 2.35 m, R2 values of 0.515 and a D2 regression score of 0.62 estimated via hold-out validation. Comparisons with existing global height products show that our approach increases detail and heterogeneity of height in short vegetation ecosystems. The modeling pipeline is open source and available on Github with reference samples, trained models and output maps publicly available under CC-BY license. Produced maps will be made available via the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters), which provides data in exchange for feedback.
- 2 May 2025
- Published by Springer Nature in Research Square
Open AccessGlobal 30-m annual median vegetation height maps (2000-2022) based on ICESat-2 data and Machine Learning
Accurately measuring vegetation height is essential for understanding ecosystem structure, carbon storage, and biodiversity, yet global height models have overwhelmingly focused on forests, excluding ecosystems with shorter herbaceous vegetation or shrubs. To address this gap in vegetation structure data, we developed the first global model to estimate median vegetation height annually from 2000-2022 at 30 m resolution, using ICESat-2 satellite Lidar, Landsat cloud free composites, and other Earth Observation raster data. Thirty (30) million ICESat-2 20 m segments were used within 10 independent draws to build ensemble Gradient Boosted Tree (GBT) models and estimate 90% prediction intervals. Our model achieves a root mean square error (RMSE) of 2.35 m, R2 values of 0.515 and a D2 regression score of 0.62 estimated via hold-out validation. Comparisons with existing global height products show that our approach increases detail and heterogeneity of height in short vegetation ecosystems. The modeling pipeline is open source and available on Github with reference samples, trained models and output maps publicly available under CC-BY license. Produced maps will be made available via the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters), which provides data in exchange for feedback.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Spatiotemporal modeling and mapping of soil organic carbon density with uncertainty quantification across Europe (2000–2022)
The paper describes a comprehensive framework for soil organic carbon density (SOCD) (kg/m3) modeling and mapping, based on spatiotemporal Random Forest (RF) and Quantile Regression Forests (QRF). A total of 45,616 SOCD measurements and various feature layers, particularly 30m Landsat-based spectral indices, were used to produce 30m SOCD maps for the EU at four-year intervals (2000--2022) and four soil depth intervals (0--20cm, 20--50cm, 50--100cm, and 100--200cm). Per-pixel 95% probability prediction intervals (PIs) and extrapolation probabilities are also provided. Model evaluation indicates consistent accuracy, with R2 between 0.53--0.67 and CCC 0.68--0.80 across cross-validations and independent tests. Prediction accuracy varies by land cover, depth interval and year of prediction with accuracy the worst for shrubland and deeper soils 100--200cm. PI validation confirmed effective uncertainty estimation, though with reduced accuracy for higher SOCD values. Shapley analysis identified soil depth as the most influential feature, followed by vegetation, long-term bioclimate, and topographic features. While pixel-level uncertainty is substantial, spatial aggregation reduces uncertainty by approximately 66%. Detecting SOCD changes remains challenging but offers a baseline for future improvements. Maps, based primarily on topsoil data from cropland, grassland, and woodland, are best suited for applications related to these land covers and depths. Users should interpret the maps with local knowledge and consider the uncertainty and extrapolation probability layers. All data and code are available under an open license at https://doi.org/10.5281/zenodo.13754343 and https://github.com/AI4SoilHealth/SoilHealthDataCube/.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessSpatiotemporal modeling and mapping of soil organic carbon density with uncertainty quantification across Europe (2000–2022)
The paper describes a comprehensive framework for soil organic carbon density (SOCD) (kg/m3) modeling and mapping, based on spatiotemporal Random Forest (RF) and Quantile Regression Forests (QRF). A total of 45,616 SOCD measurements and various feature layers, particularly 30m Landsat-based spectral indices, were used to produce 30m SOCD maps for the EU at four-year intervals (2000--2022) and four soil depth intervals (0--20cm, 20--50cm, 50--100cm, and 100--200cm). Per-pixel 95% probability prediction intervals (PIs) and extrapolation probabilities are also provided. Model evaluation indicates consistent accuracy, with R2 between 0.53--0.67 and CCC 0.68--0.80 across cross-validations and independent tests. Prediction accuracy varies by land cover, depth interval and year of prediction with accuracy the worst for shrubland and deeper soils 100--200cm. PI validation confirmed effective uncertainty estimation, though with reduced accuracy for higher SOCD values. Shapley analysis identified soil depth as the most influential feature, followed by vegetation, long-term bioclimate, and topographic features. While pixel-level uncertainty is substantial, spatial aggregation reduces uncertainty by approximately 66%. Detecting SOCD changes remains challenging but offers a baseline for future improvements. Maps, based primarily on topsoil data from cropland, grassland, and woodland, are best suited for applications related to these land covers and depths. Users should interpret the maps with local knowledge and consider the uncertainty and extrapolation probability layers. All data and code are available under an open license at https://doi.org/10.5281/zenodo.13754343 and https://github.com/AI4SoilHealth/SoilHealthDataCube/.
- Open Access
- 15 March 2025
- Published by Copernicus GmbH
Global Grassland Productivity Over Two Decades: 30m Bimonthly and Annual Gross Primary Productivity through Light Use Efficiency Modeling
This study presents a high-resolution mapping framework for estimating GPP in grasslands over the period 2000-2022 at a spatial resolution of 30 meters. The GPP values are derived utilizing a Light Use Efficiency (LUE) model using 30-m Landsat reconstructed images coupled with 1-km MOD11A1 temperature data and 1-degree CERES Photosynthetically Active Radiation (PAR). To implement the LUE model, we used the biome-specific productivity factor (maximum LUE parameter) as a global constant. This resulted in a productivity map that did not require specific land cover maps as inputs, allowing data users to calibrate GPP values accordingly to specific biomes/regions of interest. We then derived GPP maps for global grassland ecosystems based on maps produced by the Global Pasture Watch research consortium and calibrated the GPP values using the maximum LUE factor of 0.86 gCm−2d−1 MJ-1. Nearly 500 eddy covariance flux towers were used for validating the GPP estimates, resulting in R2 between 0.48-0.71 and RMSE below 2.3 gCm−2d−1 considering all land cover classes. The final time-series of maps (uncalibrated and grassland GPP) will be available as bimonthly and annual periods in Cloud-Optimized GeoTIFF (23 TB in size) as open data (CC-BY license). Users will be able to access the maps using the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters). This dataset is the first global GPP time-series map with a spatial resolution of 30 m covering a 23 year period to our knowledge.
- 15 March 2025
- Published by Copernicus GmbH
Open AccessGlobal Grassland Productivity Over Two Decades: 30m Bimonthly and Annual Gross Primary Productivity through Light Use Efficiency Modeling
This study presents a high-resolution mapping framework for estimating GPP in grasslands over the period 2000-2022 at a spatial resolution of 30 meters. The GPP values are derived utilizing a Light Use Efficiency (LUE) model using 30-m Landsat reconstructed images coupled with 1-km MOD11A1 temperature data and 1-degree CERES Photosynthetically Active Radiation (PAR). To implement the LUE model, we used the biome-specific productivity factor (maximum LUE parameter) as a global constant. This resulted in a productivity map that did not require specific land cover maps as inputs, allowing data users to calibrate GPP values accordingly to specific biomes/regions of interest. We then derived GPP maps for global grassland ecosystems based on maps produced by the Global Pasture Watch research consortium and calibrated the GPP values using the maximum LUE factor of 0.86 gCm−2d−1 MJ-1. Nearly 500 eddy covariance flux towers were used for validating the GPP estimates, resulting in R2 between 0.48-0.71 and RMSE below 2.3 gCm−2d−1 considering all land cover classes. The final time-series of maps (uncalibrated and grassland GPP) will be available as bimonthly and annual periods in Cloud-Optimized GeoTIFF (23 TB in size) as open data (CC-BY license). Users will be able to access the maps using the SpatioTemporal Asset Catalog (http://stac.openlandmap.org) and Google Earth Engine upon publication. In the meantime, beta versions of the product can be accessed through the Global Pasture Watch Early Access data program (https://survey.alchemer.com/s3/7859804/Pasture-Early-Adopters). This dataset is the first global GPP time-series map with a spatial resolution of 30 m covering a 23 year period to our knowledge.
- Open Access
- 26 February 2025
- Published by Copernicus GmbH in Earth System Science Data
- Vol. 17 (2) , 741-772
Time series of Landsat-based bimonthly and annual spectral indices for continental Europe for 2000–2022
The production and evaluation of the analysis-ready and cloud-optimized (ARCO) data cube for continental Europe (including Ukraine, the UK, and Türkiye), derived from the Landsat analysis-ready dataset version 2 (ARD V2) produced by Global Land Analysis and Discovery (GLAD) team and covering the period from 2000 to 2022, is described. The data cube consists of 17 TB of data at a 30 m resolution and includes bimonthly, annual, and long-term spectral indices on various thematic topics, including surface reflectance bands, normalized difference vegetation index (NDVI), soil adjusted vegetation index (SAVI), fraction of absorbed photosynthetically active radiation (FAPAR), normalized difference snow index (NDSI), normalized difference water index (NDWI), normalized difference tillage index (NDTI), minimum normalized difference tillage index (minNDTI), bare soil fraction (BSF), number of seasons (NOS), and crop duration ratio (CDR). The data cube was developed with the intention to provide a comprehensive feature space for environmental modeling and mapping. The quality of the produced time series was assessed by (1) assessing the accuracy of gap-filled bimonthly Landsat data with artificially created gaps; (2) visual examination for artifacts and inconsistencies; (3) plausibility checks with ground survey data; and (4) predictive modeling tests, examples with soil organic carbon (SOC) and land cover (LC) classification. The time series reconstruction demonstrates high accuracy, with a root mean squared error (RMSE) smaller than 0.05, and R2 higher than 0.6, across all bands. The visual examination indicates that the product is complete and consistent, except for winter periods in northern latitudes and high-altitude areas, where high cloud and snow density introduce significant gaps and hence many artifacts remain. The plausibility check further shows that the indices logically and statistically capture the processes. The BSF index showed a strong negative correlation (−0.73) with crop coverage data, while the minNDTI index had a moderate positive correlation (0.57) with the Eurostat tillage practice survey data. The detailed temporal resolution and long-term characteristics provided by different tiers of predictors in this data cube proved to be important for both soil organic carbon regression and LC classification experiments based on 60 723 LUCAS observations: long-term characteristics (tier 4) were particularly valuable for predictive mapping of SOC and LC, coming out on top of variable importance assessment. Crop-specific indices (NOS and CDR) provided limited value for the tested applications, possibly due to noise or insufficient quantification methods. The data cube is made available at https://doi.org/10.5281/zenodo.10776891 (Tian et al., 2024) under a CC-BY license and will be continuously updated.
- 26 February 2025
- Published by Copernicus GmbH in Earth System Science Data
- Vol. 17 (2) , 741-772
Open AccessTime series of Landsat-based bimonthly and annual spectral indices for continental Europe for 2000–2022
The production and evaluation of the analysis-ready and cloud-optimized (ARCO) data cube for continental Europe (including Ukraine, the UK, and Türkiye), derived from the Landsat analysis-ready dataset version 2 (ARD V2) produced by Global Land Analysis and Discovery (GLAD) team and covering the period from 2000 to 2022, is described. The data cube consists of 17 TB of data at a 30 m resolution and includes bimonthly, annual, and long-term spectral indices on various thematic topics, including surface reflectance bands, normalized difference vegetation index (NDVI), soil adjusted vegetation index (SAVI), fraction of absorbed photosynthetically active radiation (FAPAR), normalized difference snow index (NDSI), normalized difference water index (NDWI), normalized difference tillage index (NDTI), minimum normalized difference tillage index (minNDTI), bare soil fraction (BSF), number of seasons (NOS), and crop duration ratio (CDR). The data cube was developed with the intention to provide a comprehensive feature space for environmental modeling and mapping. The quality of the produced time series was assessed by (1) assessing the accuracy of gap-filled bimonthly Landsat data with artificially created gaps; (2) visual examination for artifacts and inconsistencies; (3) plausibility checks with ground survey data; and (4) predictive modeling tests, examples with soil organic carbon (SOC) and land cover (LC) classification. The time series reconstruction demonstrates high accuracy, with a root mean squared error (RMSE) smaller than 0.05, and R2 higher than 0.6, across all bands. The visual examination indicates that the product is complete and consistent, except for winter periods in northern latitudes and high-altitude areas, where high cloud and snow density introduce significant gaps and hence many artifacts remain. The plausibility check further shows that the indices logically and statistically capture the processes. The BSF index showed a strong negative correlation (−0.73) with crop coverage data, while the minNDTI index had a moderate positive correlation (0.57) with the Eurostat tillage practice survey data. The detailed temporal resolution and long-term characteristics provided by different tiers of predictors in this data cube proved to be important for both soil organic carbon regression and LC classification experiments based on 60 723 LUCAS observations: long-term characteristics (tier 4) were particularly valuable for predictive mapping of SOC and LC, coming out on top of variable importance assessment. Crop-specific indices (NOS and CDR) provided limited value for the tested applications, possibly due to noise or insufficient quantification methods. The data cube is made available at https://doi.org/10.5281/zenodo.10776891 (Tian et al., 2024) under a CC-BY license and will be continuously updated.