The Qinghai-Tibet Plateau (QTP) harbors significant amounts of soil organic carbon (SOC) in the permafrost regions, which are at risk of release as carbon dioxide or methane under global warming, amplifying the greenhouse effect. Despite this, long-term investigations into the spatiotemporal dynamics of SOC in the QTP's permafrost regions remain scarce. Furthermore, spatial scale mismatches between SOC maps and thermokarst landscape maps hinder a comprehensive understanding of carbon cycling mechanisms in these landscapes. Hyperspectral data, with its superior spectral richness, offers the potential to more precisely capture soil spectral characteristics, enhancing the accuracy of SOC estimations. However, the limited availability of long-term hyperspectral datasets for the QTP presents a major challenge to leveraging this technology for SOC estimation.
In this study, we developed a physically constrained hyperspectral generative model that integrated spectral response functions and diffusion models, utilizing satellite data from Landsat 5 TM, Landsat 7 ETM+, Landsat 8 OLI, and EO-1 Hyperion imagery. This method generated high-accuracy hyperspectral data (MSSIM = 0.96, PSNR = 38.65) for the permafrost regions of the QTP from 2000 to 2020, with a spatial resolution of 30 m and a spectral resolution of 10 nm. Leveraging these generated hyperspectral data, we constructed spectral indices and incorporated climate, topography, and soil characteristics into a dual-input convolutional neural network model. This model enabled the mapping of the spatiotemporal distribution of SOC in the 0-3 m layer across the QTP’s permafrost regions from 2000 to 2020 with resolution of 30 m. Compared to existing approaches, our model achieved a 22.9% improvement in the accuracy of SOC estimation in permafrost regions, highlighting its potential for advancing carbon estimation.
How to cite: Ni, C. and Zhu, B.: Digital mapping of soil organic carbon in permafrost regions over the Qinghai-Tibet Plateau based on deep learning and hyperspectral imaging, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7977, https://doi.org/10.5194/egusphere-egu25-7977, 2025.
Soil is fundamental to ecosystem services, agriculture, and climate regulation, serving as a medium for water and nutrient absorption, a habitat for biodiversity, and a major reservoir for organic carbon. Yet, soil faces increasing threats from degradation caused by intensive land use, deforestation, and climate change, which jeopardize food security and environmental sustainability.To address these challenges, this work harnesses recent advances in remote sensing and data analytics to create a comprehensive global soil dataset spanning from 1985 to 2023. This dataset covers five key properties: pH, salinity, nitrogen, phosphorus, and organic carbon content. By leveraging Google Earth Engine and machine learning algorithms, we generated global, high-resolution maps that enable researchers to monitor changes in soil health over time and predict future trends.The primary objective of this dataset is to support decision-making for sustainable land management, agriculture, and environmental conservation. It offers a critical tool for combating soil degradation and mitigating its impacts, empowering stakeholders with actionable insights to preserve and restore soil health on a global scale.
How to cite: Dalle Vaglie, M., Martellozzo, F., Chirici, G., and Francini, S.: Advancing Soil Health Monitoring: A Machine Learning Approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3300, https://doi.org/10.5194/egusphere-egu25-3300, 2025.
Leaf Area Index (LAI) is a key trait related to several agronomic issues such as soil cover, plant health, crop productivity, biomass and yield estimation. Availability of high-resolution LAI information at large scale is crucial for monitoring and managing agricultural landscapes effectively [1], as it can help monitor growth conditions and adapt practices. However, its satellite-based assessment is confounded by several factors such as soil background, vegetation type and noise. Today, the retrieval of LAI through the inversion of a radiative transfer model (RTM) is state-of-the-art. Still, research investigating the performance of crop-type specific models compared to across-biome models such as the ESA’s Sentinel Application Platform (SNAP) and in-situ data is rare.
In this research we propose to improve the combined leaf and canopy PROSAIL [2] RTM crop-specific reflectance simulations by integrating soil spectra into this model. We specifically sample Sentinel-2 spectra from fields over which we perform LAI retrieval. A neural network is trained to invert the RTM. To scale this strategy to larger areas (i.e. country scale) we exploit Sentinel-2 observations of bare soil and use clustering methods to generate a condensed soil dataset representing varying background conditions across space.
We use Switzerland to test the approach, with in-situ measurements of winter wheat from 2022 and 2023 available for validation. We focus on Sentinel-2 imagery for its high temporal and spatial resoltuions. Preliminary results show that a model trained on a data generated with a Switzerland-wide soil dataset and constrained for winter wheat (CH-LAI-WW model) outperformed predictions (nRMSE: 0.180) obtained from a classic setup without the soil inclusion (nRMSE: 0.201). Furthermore, prediction errors were improved compared to the across-biome SNAP LAI processor (nRMSE: 0.268). The proposed methodology demonstrates a way to improve the crop- and biome-specific prediction of key traits and consequently to improve the reliability for agricultural monitoring and management applications.
[1] B. Brisco, R. Brown, T. Hirose, H. McNairn, and K. Staenz, “Precision agriculture and the roleof remote sensing: A review,” Canadian Journal of Remote Sensing, vol. 24:3, pp. 315–327, 1998
[2] S. Jacquemoud, W. Verhoef, F. Baret, C. Bacour, P. J. Zarco-Tejada, G. P. Asner, C. François, and S. L. Ustin, “PROSPECT+SAIL models: A review of use for vegetation characterization,” Remote Sensing of Environment, vol. 113, pp. S56–S66, Sept. 2009.
How to cite: Ledain, S., Gilgen, A., and Aasen, H.: Enhanced winter wheat LAI retrieval from Sentinel-2: asoil-informed radiative transfer based approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1583, https://doi.org/10.5194/egusphere-egu25-1583, 2025.
Understanding global patterns of tree water use is crucial for predicting forest resilience and ecosystem responses under climate change. Despite its importance, a high-resolution global assessment of tree water-use where observations are not available remains lacking. This study aims to create a global map of tree water use using machine learning approaches applied to sap flow measurements. Here we leverage the SAPFLUXNET database, a global repository of standardized sap velocity measurements, combined with remote sensing, meteorological, and tree characteristics data utilizing machine learning techniques to estimate global gridded sap velocity. We employ an ensemble learning approach with two distinct setups: site-level and plant-level setup. The site-level setup aggregates plant measurements at each location and incorporates site-level predictors, while the plant-level setup utilizes individual measurements with both site and plant-level variables. For each setup, we implement four machine learning algorithms: Support Vector Machine, Random Forest, XGBoost, Artificial Neural Networks and Long Short-Term Memory. To optimize predictor selection and prevent model complexity, we employ the Guided Hybrid Genetic Algorithm. The final ensemble estimate will be derived as the median of all predictions. This analysis yields two major outcomes. First, the ensemble learning approach produces a daily global dataset of sap velocity at 0.1 ° from 2000-2018 and reveal global patterns of tree water use, highlighting systematic variations across biomes and their relationship to environmental gradients. Second, our methodology identified the relative importance of predictors, and dominant climatic controls across different ecosystems. These findings will advance our understanding of forest ecosystem responses to environmental change and support more accurate predictions of forest resilience under future climate scenarios.
Key words: tree water-use, machine learning, SAPFLUXNET, climate change
How to cite: Luo, Y. and Gharun, M.: Predicting Global Patterns of Tree Water Use: An Ensemble Learning Approach Using SAPFLUXNET, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2473, https://doi.org/10.5194/egusphere-egu25-2473, 2025.
The above ground biomass (AGB) of woody vegetation is proportional to the amount of carbon stored primarily in the trunks and branches, with changes over time indicating sources or sinks of carbon. Accurate quantification of AGB is indispensable for climate studies and policy development, yet significant gaps persist due to limitations in current observational and modeling approaches. Satellite-based Earth Observation (EO) provides a promising avenue for global biomass estimation, particularly when a diversity of data sources and advanced algorithms are used.
Recent initiatives, such as the European Space Agency’s (ESA) Climate Change Initiative (CCI) Biomass and BiomAP projects, have pioneered methodologies for generating time series of global maps of woody AGB at varying spatial resolutions. These efforts utilize multiple predictors derived from active and passive microwave data sources, including Sentinel-1, ALOS-2, SMOS, SMAP and ASCAT as well as LiDAR-based vegetation structural metrics. However, the absence of globally and evenly distributed AGB measurements acting as reference constrains retrievals to use fully physical models. These models are then calibrated using spatially explicit datasets from other satellite data (e.g., optical imagery) and AGB statistics. Evaluations of these maps with independent reference measurements not used in the retrieval process highlight the critical balance between data precision and algorithm design. The complexity of accurately mapping biomass at global scales is compounded by uncertainties in LiDAR sampling, satellite data uncertainty, and the dependence on high-quality reference data. Additionally, biases arise from the simplistic assumptions often required for model fitting, which can affect the reliability of AGB estimates. Temporal assessments of biomass change face additional hurdles, including uncertainties in AGB trends and a scarcity of reference data for validation.
Despite these challenges, EO-driven biomass mapping continues to advance, supported by improvements in sensor technologies and retrieval algorithms. Long-term maintenance of satellite missions suitable for AGB mapping is however essential as is the promotion of space-based LiDAR observations. Enhanced understanding of satellite signal characteristics will enable more accurate AGB retrievals, fostering the development of sophisticated retrieval models that may identify complex interactions not described by the physical models currently in use. Crucially, this progress must be complemented by spatially dense and continuous AGB measurements from local ground-based or airborne surveys.
The scope of this presentation is to emphasize the transformative potential of satellite EO in quantifying and monitoring AGB and detail efforts at quantifying and reducing uncertainties in retrieval. By reviewing existing data products and illustrating strategies to address data gaps and methodological challenges, this work aims to inform and guide future global biomass estimation efforts from existing, recently launched (e.g., ALOS-4 PALSAR, Sentinel-1C), and forthcoming (NASA/ISRO NISAR and ESA BIOMASS) missions.
How to cite: Santoro, M., Cartus, O., Favrichon, S., Quegan, S., Kay, H., Lucas, R., Araza, A., Herold, M., Labrière, N., Chave, J., Rosenqvist, Å., Tadono, T., Kobayashi, K., Kellndorfer, J., and Seifert, F. M.: Advancements and challenges in estimating terrestrial vegetation biomass using satellite data, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18186, https://doi.org/10.5194/egusphere-egu25-18186, 2025.
The Arctic is warming faster than anywhere else on Earth, placing tundra ecosystems at the forefront of global climate change. Plant biomass is a fundamental ecosystem attribute that is sensitive to changes in climate, closely tied to ecological function, and crucial for constraining ecosystem carbon dynamics. However, the amount, functional composition, and distribution of plant biomass are only coarsely quantified across the Arctic. Therefore, we developed the first moderate resolution (30 m) maps of live aboveground plant biomass (g m-2) and woody plant dominance (%) for the Arctic tundra biome, including the mountainous Oro Arctic. We modeled biomass for the year 2020 using a new synthesis dataset of field biomass harvest measurements, Landsat satellite seasonal synthetic composites, ancillary geospatial data, and machine learning models. Additionally, we quantified pixel-wise uncertainty in biomass predictions using Monte Carlo simulations and validated the models using a robust, spatially blocked and nested cross-validation procedure. Observed plant and woody plant biomass values ranged from 0 to ~6,000 g m-2 (mean ≈ 350 g m-2), while predicted values ranged from 0 to ~4,000 g m-2 (mean ≈ 275 g m-2), resulting in model validation root-mean-squared-error (RMSE) ≈ 400 g m-2 and R2 ≈ 0.6. Our maps not only capture large-scale patterns of plant biomass and woody plant dominance across the Arctic that are linked to climatic variation (e.g., thawing degree days), but also illustrate how fine-scale patterns are shaped by local surface hydrology, topography, and past disturbance. By providing data on plant biomass across Arctic tundra ecosystems at the highest resolution to date, our maps can significantly advance research and inform decision-making on topics ranging from Arctic vegetation monitoring and wildlife conservation to carbon accounting and land surface modeling.
How to cite: Orndahl, K., Berner, L., Macander, M., and Goetz, S. and the The Arctic Plant Aboveground Biomass Mapping Team: Next generation Arctic vegetation maps: Aboveground plant biomass and woody dominance mapped at 30 m resolution across the tundra biome, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2597, https://doi.org/10.5194/egusphere-egu25-2597, 2025.
Land ecosystems play a crucial role in the global carbon cycle, absorbing large amounts of atmospheric CO2 through photosynthesis and releasing it back through decomposition and respiration. However, predicting the net carbon flux is a complex task, as meteorological variability affects these processes in different ways and at various timescales. Data-driven models, such as global upscaling methods based on local flux tower measurements, struggle especially to accurately predict the year-to-year fluctuations in the net terrestrial carbon uptake, known as inter-annual variability (IAV). These difficulties are often being attributed to a lack of observational data, however, do we need longer time series of terrestrial carbon flux observations or better spatial coverage to improve IAV predictions?
Here, we test the change in performance of an interpretable machine learning (ML) framework given growing training datasets of different properties created based on output from a global land surface model (JSBACH3.2). All training datasets have an initial setting comparable to the actual observational setting (FLUXNET sites). We scale the training datasets either by increasing the number of pixels in training (space model), or by extending the time series (time model), or both (space-time model), however, we keep the increment in additional training samples to each training dataset constant. The ML framework is trained on the training datasets of different sizes and characteristics and evaluated in predicting IAV on an independent test set. To take the various effective time scales into account, our ML framework* is based on a wavelet transform of the predictor variables and a convolutional neural network to jointly predict carbon and water fluxes.
Our results confirm that increasing the sample size in the training dataset substantially enhances the performance in predicting global IAV. Further, we find that increasing the spatial coverage during training improves model performance in predicting IAV more (space model; ΔR2=0.83) than increasing the length of the time series (time model; ΔR2=0.60) compared to the initial setup. Overall, the model trained with the largest number of pixels (space-model) outperforms the other models for the same total number of training samples but fewer pixels. Using the interpretable ML technique based on the wavelet transform, we investigate the differences among the three models towards their sensitivity to different meteorological factors. We focus this analysis part on test pixels where the space and time models show the largest performance discrepancy.
In conclusion, our study demonstrates that a large spatial representation in the observational training data is more important than longer observational time series for predicting year-to-year fluctuations in global land carbon uptake.
*Reimers, C., Hafezi Rachti, D. , Liu, G., & Winkler, A. J. (2024). Comparing Data-Driven and Mechanistic Models for Predicting Phenology in Deciduous Broadleaf Forests. arXiv preprint arXiv:2401.03960.
How to cite: Hafezi Rachti, D., Reimers, C., and Winkler, A. J.: Space versus Time: Better Spatial Representation in Training Beats Longer Time Series for Predicting Global Land Carbon Uptake Variability, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12808, https://doi.org/10.5194/egusphere-egu25-12808, 2025.
How to cite: Chen, R., Liu, L., Liu, X., and Rascher, U.: CMLR: A Mechanistic Global GPP Dataset Derived from TROPOMIS SIF Observations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14472, https://doi.org/10.5194/egusphere-egu25-14472, 2025.
Species distribution models (SDMs) are vital tools for monitoring biodiversity. By relating environmental conditions to species occurrences, these statistical models enable the mapping of species distribution and provide insights into the key drivers influencing their patterns. In this context, the recently proposed MaskSDM approach presents a novel opportunity to highlight and interpret the influence of input variables and data modalities on the modeled outcomes. By leveraging masked data modeling during training, MaskSDM allows the flexible selection of any subset of input variables, based on their availability for the given location and their relevance to the target species. This flexibility enables the evaluation of both predictions and model performance dynamics across different subsets of variables.
In this study, we use MaskSDM to investigate the impact of using time series alongside traditional tabular data for modeling the distribution of plant species across Europe. The temporal dimension of ecological processes is crucial, with phenology playing a significant role in driving plant species behavior and distribution. Time series data effectively capture these dynamic processes, providing valuable insights to the model. For our analysis, we utilize the GeoPlant dataset, which comprises monthly climatic time series for temperature and precipitation, as well as satellite-derived time series spanning six spectral bands at a quarterly resolution. These satellite data capture local patterns, such as seasonal vegetation changes and the effect of extreme natural events like wildfires.
MaskSDM being based on a transformer model, we evaluate several approaches for tokenizing the time series data, and assess the individual contribution of each input. Our results show that the performance of different tokenization methods is comparable. We then examine the effect of incorporating various types of time series data on model performance and compare it to the use of tabular data only: adding satellite time series increases the AUC on the spatially separated test set by 2.4%. The addition of climatic time series yields a smaller improvement, likely because the tabular data already includes some aggregated form of statistics redundant to these time series. The best performance is achieved by combining all time series data with the tabular data, showing their complementary nature.
Finally, we produce species distribution maps that consider different data types. The impact of adding time series data to the tabular data is evident after the analysis of the maps, which become closer to the spatial distribution of the presence observations. These findings emphasize the importance of incorporating time series data into SDMs, particularly satellite data, as it captures temporal dynamics that are difficult to represent through tabular data alone.
How to cite: Zbinden, R., Charlet, J., Sümbül, G., and Tuia, D.: Evaluating the impact of multimodal climate and satellite time series in species distribution modeling using MaskSDM, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11949, https://doi.org/10.5194/egusphere-egu25-11949, 2025.
This research has two primary aims: first, to assess the capability of machine learning (ML) techniques generally used for flood susceptibility mapping in replicating flood hazard maps derived from hydraulic modeling, and second, to apply the trained ML models to unseen catchments using a similar set of parameters. The study focuses on two catchments—Argens and Gapeau—located in South-Eastern France. Reference flood hazard maps were generated using the FLOODOS 2D hydraulic model at a 5-meter resolution, simulating water depths for a 1,000-year return period across 1,163.1 km of rivers. From these maps, a balanced dataset of flood and non-flood points was created and split into training and validation subsets (70:30) via random sampling. The analysis employed several geo-environmental factors as explanatory variables, including a 5-meter resolution Digital Terrain Model, Height Above Nearest Drainage, river slope, and river discharge data used in hydraulic modeling. Three advanced ML models—artificial neural networks, random forests, and extreme gradient boosting—were trained on this dataset. These trained models were then tested on the Gapeau region to evaluate their robustness and effectiveness in replicating flood hazards. Model performance was assessed using metrics such as the Area Under the Receiver Operating Characteristic Curve (AUROC) and the Critical Success Index, which measure prediction accuracy for flood extents. Results indicated that ML models effectively mapped flood hazards in complex geo-topographic regions like the Argens basin. Notably, two models achieved AUROC scores exceeding 0.9 when applied to the untrained Gapeau region, demonstrating good transferability and predictive accuracy.
How to cite: Arora, A., Payrastre, O., and Nicolle, P.: Evaluation of Machine Learning Approaches and their Extrapolation for Flood Hazard Mapping, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12250, https://doi.org/10.5194/egusphere-egu25-12250, 2025.
The Fluorescence Explorer (FLEX) mission aims to monitor vegetation sun-induced chlorophyll fluorescence (SIF), a proxy for ecosystem health. Studying photosynthesis and its relationship with SIF provides valuable insights into the physiological responses of ecosystems to environmental stress. Peatlands are among the most valuable ecosystems in the carbon cycle, acting as carbon storage and sinks in normal conditions but becoming carbon sources in drought conditions.
This study investigates the correlation between SIF and Gross Primary Productivity (GPP) in the temperate peatland in Poland. Measurements by the FloX system with a temporal resolution of ~30 seconds between June and October 2019 and CO2 fluxes measured by manual chambers were used for this purpose. GPP was calculated from the subsequent net ecosystem exchange and respiration measurements, and the gaps between measurements were modelled with a Michaelis-Menten rectangular hyperbola. SIF was retrieved using the Improved Fraunhofer Line Depth (iFLD) Spectral Fit Method (SFM) and Spectrum Fitting (SpecFit) algorithms in the O2-A and O2-B bands. The data were analysed at different time intervals (30 min, 1 h, 3 h, whole day, and the entire dataset).
Our results show that the strength of the SIF-GPP relationships changes significantly with time interval. Correlations tend to weaken or break (r2 <0.5) more frequently at shorter intervals, while stronger, more consistent relationships are observed over full-day periods or when the entire dataset is combined. This highlights the importance of temporal resolution when interpreting SIF-GPP relationships. Although exponential correlations have been observed at whole-day or dataset scales, these patterns may mask short-term physiological responses and stress dynamics under varying environmental conditions.
As the FLEX satellite will only provide one observation per day, our results emphasize the limitations of single daily measurements, which are influenced by transient weather conditions or plant stress. Therefore, continuous ground-based spectral data are essential to improve the reliability of SIF-based ecosystem monitoring.
This study emphasizes the importance of temporal resolution in SIF-GPP analyses and contributes to the validation efforts of the FLEX mission. Future research should validate these results across other ecosystems and integrate data from the ESA FLEXSense tandem campaigns (2018–2019) to improve global photosynthesis monitoring.
The National Science Centre, Poland, funded the 2020/39/O/ST10/00775 research.
How to cite: Abdelmajeed, A. Y. A., Cendrero-Mateo, M. P., Antala, M., Albert-Saiz, M., Stróżecki, M., Rastogi, A., Julitta, T., Burkart, A., Schuettemeyer, D., and Juszczak, R.: Assessing SIF-GPP Relationships in Peatlands: Temporal Insights for FLEX Mission Validation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18025, https://doi.org/10.5194/egusphere-egu25-18025, 2025.
Due to the effects of global climate change, Korea is experiencing intensified heavy rainfall, leading a rapid runoff along with large amount of non-point source pollution flow into water bodies, posing a significant threat to the aquatic ecosystem and water supply system. Especially, stagnant water bodies are highly susceptible to eutrophication due to these pollutants, and warm water temperatures further cause the occurrence of the algal blooms. As such, developing accurate spatio-temporal monitoring of water quality parameters (WQPs) over water body become essential. In Korea, algae-related WQPs such as chlorophyll-a (Chl-a) and microcystin are mainly monitored using fixed observation stations. Korea operates 76 automated monitoring stations, but these methods have limitations in capturing the spatial distribution of WQPs. Furthermore, manual stations collect data once a week, resulting in the lack of spatio-temporal continuity.
In this study, we developed and validated a Chl-a estimation model based on Random Forest Regression over Daecheong Lake using surface reflectance data from the Geostationary Ocean Color Imager-II (GOCI-II) onboard the Geo-Kompsat-2B (GK-2B) satellite. GOCI-II provides surface reflectance eight times per day at a spatial resolution of 250 m. The point observation data consisted of hourly Chl-a concentrations obtained from the Korean Water Environment Information System. The study period spanned three years (January 2021 to December 2023). For model development, the dataset was randomly divided into a 7:3 ratio for training and testing. The model's input variables included the spectral bands of GK-2B GOCI-II, the normalized difference chlorophyll index, the normalized fluorescence height index, and the fluorescence line height. The dependent variable was the log-transformed Chl-a data. Furthermore, the study assessed the model's efficiency by sequentially removing input variables based on their feature importance rankings.
As a result, the statistically optimal combination of input variables included all seven variables. The model's performance showed bias, Root Mean Square Error, and Correlation Coefficient values of -0.0041 ppb, 0.1649 ppb, and 0.91, respectively. Despite these favorable statistical results, uncertainties were observed during periods of extremely low or high Chl-a concentrations. Finally, the spatial distribution of Chl-a was estimated using the developed model demonstrated a clear spatial pattern with seasonal variations. However, uncertainties were evident at the boundaries between water bodies and land surfaces. These uncertainties likely arose due to the limited spatial resolution of 250 m, which was insufficient for capturing narrow lake widths.
Future studies should address these limitations, focusing on spatial downscaling of surface reflectance to reduce boundary-related uncertainties, as well as minimizing the underestimation/overestimation of extreme Chl-a values through establishing seperate training based on the seasonal characteristics or temporal behavior of hydrometeorological variables.
Acknowledgement: This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT)(RS-2024-00416443).
How to cite: Park, K. and Park, J.: Development and validation of chlorophyll-a estimation model using GOCI-II land surface reflectance and machine learning at Daecheong Lake in South Korea, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14300, https://doi.org/10.5194/egusphere-egu25-14300, 2025.
The Integration of extensive user-generated data, for example, Points of Interest (POI), along with novel advances in machine learning that are applied in the analysis of satellite imagery, brings about a revolutionary approach toward urban land use classification bridging significant deficiencies in understanding the dynamics of urban life. POI data captures much of the various social and economic activities, whereas satellite imagery expresses spatial and physical context, but neither together captures the complexities of the urban setting. This paper proposes a new approach in which all these types of data are fused into a single text-based format through transformation of structured POI datasets and satellite images, thus enabling topic modeling for the classification and mapping of land use. The integration of these data formats addresses a variety of challenges, such as spatial heterogeneity, non-linear relationships, and extrapolation artifacts, to allow for a scalable and precise solution in urban data analysis. This methodology is in line with the challenges and opportunities that large-scale mapping presents, whereby machine learning algorithms can yield robust and spatially explicit predictions by connecting diverse datasets. This approach improves the accuracy and resolution of urban land use maps and thus offers insights into the interplay of human activities and physical spaces, which are very important for planning and policy-making in urban regions. The results demonstrate significant promise as the combination of POI data and satellite imagery enhances the understanding of complex urban systems and supports sustainable development with practical tools for designing more adaptable and resilient urban environments. Moreover, this contribution is aligned with the session focus on comprehensive mapping techniques addressing some of the biggest challenges in upscaling data; building representative measurement models; handling uncertainty; and ensuring robust validation. In so doing, it helps to further develop better policies for urban land use, besides contributing to overall goals of mapping environmental variables in diverse and dynamic environments. This methodology not only enhances the techniques of urban planning but also sets a benchmark for the incorporation of intricate datasets to improve comprehension and management of the difficulties faced by contemporary urban areas, thereby promoting a more profound relationship between human actions and the physical contexts in which they take place.
How to cite: Dabbanavar, R. S. and Biswas, A.: Integrating User-Generated POI Data and Satellite Imagery for Enhanced Urban Land Use Classification: A Topic Modeling Approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8380, https://doi.org/10.5194/egusphere-egu25-8380, 2025.
Canopy temperature is a vital indicator of plant-environment interactions, playing a key role in assessing drought impacts and water use efficiency and understanding the global carbon cycle. Despite existing efforts on ground measurements at FLUXNET and ICOS sites, satellite Land Surface Temperature (LST) has been widely used as a proxy for canopy temperature in studies at different scales. However, satellite-based LST represents a mixture of vegetation, snow and soil temperatures, leading to biases in estimating plant-related processes, particularly for temperature-limited regions. Up to now, there is still no available global canopy temperature product.
To bridge the research gap, this study first combines in-situ canopy temperature measurements (from outgoing longwave radiation and thermal camera), satellite LST into a machine learning model for global canopy temperature mapping. In the algorithm, vegetation cover, ERA5 soil and air temperatures as well as other auxiliary data were adopted for estimating canopy temperature, by removing the contributions of soil and air information to satellite LST. The primary validation results over more than global 130 sites, by separating training group (2/3) and evaluation group (1/3), indicated the retrieval of canopy temperature is encouraging, by achieving average RMSE (Root Mean Squared Error) of 2.80 K, Bias of 0.12 K and correlation coefficient (R) of 0.96, against ground measurements. In the next steps, more auxiliary data including net radiation, vegetation water, vapor pressure deficit, wind speed, soil moisture and vegetation parameters, will be included for the retrieving and final global product development. In the process, the interactions of these variables with the targeted canopy temperature will be also investigated. The final global daily canopy temperature for more than 20 years, with a spatial resolution of 0.25°, will be released and further used to evaluate the land surface version of the dynamic vegetation model, LPJ-GUESS, to assess the impacts of canopy instead of air temperature on influencing global terrestrial water-carbon cycles.
How to cite: Ma, H., Schurgers, G., and Tang, J.: Global canopy temperature estimation by integrating satellite observations and ground measurements, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-826, https://doi.org/10.5194/egusphere-egu25-826, 2025.
Abstract
This research proposes a comprehensive and sustainable framework for wetland monitoring by evaluating the wetland environmental superiority index (WESI) of inland wetlands using a long short-term memory (LSTM) model. The WESI estimation method aims to establish a long-term and periodic monitoring system for extensive regions based on remote sensing and reanalysis data. To achieve this, exemplary wetland sites representing high-quality and vulnerable wetlands were selected, with a label of 1 assigned to high-quality wetlands and 0 to vulnerable wetlands. These exemplary wetland sites provide the target variables for training the LSTM-based WESI estimation model, while remote sensing and reanalysis datasets closely associated with the environmental characteristics of inland wetlands are utilized as input variables. In this study, a comprehensive database comprising 13 types of hydrometeorological, vegetation, topographic, and carbon-related remote sensing and reanalysis datasets was established. Additionally, 30 exemplary high-quality and 30 vulnerable inland wetlands—identified based on the field survey conducted by the National Institute of Ecology (NIE) in South Korea—were used to evaluate the applicability of the proposed framework. The WESI estimation method is expected to contribute to the establishment of a long-term, continuous monitoring system for inland wetlands by leveraging the stable data production capabilities of remote sensing and reanalysis data.
Acknowledgement
This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (RS-2024-00334564) and in part by Korea Environmental Industry&Technology Institute (KEITI) through Wetland Ecosystem Value Evaluation and Carbon Absorption Value Promotion Technology Development Project, funded by Korea Ministry of Environment (MOE). (2022003640001)
How to cite: Hwang, S., Baik, J., Ku, S., Sung, J., and Jun, C.: An inland wetlands monitoring framework leveraging remote sensing and reanalysis-based datasets: a case study of 60 inland wetlands in south korea, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5160, https://doi.org/10.5194/egusphere-egu25-5160, 2025.
Large-scale, high-resolution landscape mapping with precise classification is essential for understanding, managing, and protecting Earth's ecosystems. It provides granular spatial and thematic insights into land cover and land-use dynamics, allowing for a better representation of complex landscapes with multiple classes. By preserving fine-scale heterogeneity, such mapping enables the identification of subtle yet ecologically significant patterns, including habitat fragmentation, biodiversity hotspots, and land-use transitions. Despite the availability of several high-resolution global land cover products, there is a significant lack of detailed class information in these datasets. The existing classes are often too general and fail to accurately represent the inherent heterogeneity of landscapes. However, this task remains challenging due to intricate ground features, diverse landforms, and the limited availability of accurate training labels across extensive geographic regions.
In this study, we employed an efficient weakly supervised deep learning architecture to enable large-scale, high-resolution land cover mapping with detailed class distinctions. This was achieved by utilizing widely accessible and publicly available satellite products and global land cover (GLC) data, with a focus on Brandenburg, a federal state of Germany. We used the CORINE Land Cover (CLC) 2018 dataset as a low-resolution land cover label, alongside nine bands from Sentinel-2 MSI data and two bands (VV and VH) from Sentinel-1 SAR data, all at a 10-meter spatial resolution, organized into 256x256 pixel patches. While the CORINE dataset offers rich class information with 44 thematic classes (28 for Brandenburg), its coarse resolution (100 m) limits its utility for large-scale analyses. To address this, we enhanced the resolution of the dataset to 10 meters by integrating satellite data from hybrid sources. Additionally, we incorporated high-resolution global land cover databases, such as Dynamic World V1, into the model’s loss function to guide the generation of high-resolution data products while maintaining the same number of classes as CORINE. This framework addressed label noise resulting from the resolution mismatch between images and labels by combining a resolution-preserving CNN branch, a Transformer branch, a weakly supervised module, and a self-supervised loss function, enabling the automatic refinement of high-resolution land cover results without manual annotations.
Our results, obtained after running 30 epochs in the Google Colab Pro Python environment with a limited A100 GPU (~40 GB), show promising outcomes, with a gradual decrease in loss. The predicted validation data, aggregated into broader class categories, were compared with the Dynamic World dataset, yielding a match of 68%. Specific classes, such as cropland, vegetation, and grassland, demonstrated strong performance, with accuracy scores of 84%, 66%, and 55%, respectively. This framework generates high-resolution, detailed landscape maps with rich class information from accessible global land cover products, all without the need for manual annotation. It can also be applied across Europe, as the CORINE data covers the entire continent. While these results are encouraging, we are confident that further analyses, including additional training with more epochs and data, will improve performance even further.
How to cite: Halder, K., Srivastava, A. K., Muduchuru, K., Han, L., Singh, M., Gaiser, T., and Ewert, F.: Transforming Low-Resolution CORINE Data into High-Resolution Landscape Maps with Semi-Supervised Deep Learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18335, https://doi.org/10.5194/egusphere-egu25-18335, 2025.
Posters virtual: Wed, 30 Apr, 14:00–15:45 | vPoster spot A
EGU25-2041 | ECS | Posters virtual | VPS4
Integrating Automated Lineament Extraction, Magnetic Data, and Machine Learning-Based Lithological Mapping in the Anti Atlas, MoroccoWed, 30 Apr, 14:00–15:45 (CEST) | vPA.31
Abstract
This study explores advanced remote sensing, geophysical, and geospatial methodologies applied to the geologically diverse Aït Semgane region in Morocco. A multi-disciplinary approach was adopted, combining (1) automated lineament extraction using Digital Elevation Models (DEMs) and various topographic indices, (2) lithological classification leveraging machine learning algorithms on multispectral data, and (3) the integration of magnetic data to enhance geological interpretation.
For lineament analysis, approaches such as the Topographic Position Index (TPI), Hillshade, and shading models were applied to datasets including SRTM, ALOS PALSAR, and Sentinel-1 InSAR. Results highlighted the TPI method’s high sensitivity in detecting tectonic features, especially in NE-SW and E-W orientations, aligning with established geological knowledge. Cartographic analysis revealed fault density concentrations in the NW and southern sectors, confirming the tectonic complexity of the region.
Lithological classification was conducted using Support Vector Machines (SVM), Random Trees (RT), and Artificial Neural Networks (ANN) applied to Landsat 9 and Sentinel-2 data. SVM, particularly with Minimum Noise Fraction (MNF) transformation, consistently outperformed other algorithms, achieving high classification accuracies and well-defined lithological boundaries. The integration of dimensionality reduction techniques like MNF proved crucial for enhancing classification quality, while PCA showed limited efficacy.
Magnetic data were incorporated to validate and refine the tectonic and lithological interpretations, offering additional insights into subsurface structures and enhancing the understanding of fault systems and mineralized zones.
This research demonstrates the synergy between automated lineament extraction, machine learning-based lithological mapping, and magnetic data for improving geological analysis. The methodologies applied here have practical implications for mineral exploration and tectonic studies, offering robust tools for mapping complex terrains. Future research will aim to refine dimensionality reduction techniques, explore hyperspectral datasets, and further integrate geophysical data to enhance geological mapping accuracy.
How to cite: El-Omairi, M. A. and El Garouani, A.: Integrating Automated Lineament Extraction, Magnetic Data, and Machine Learning-Based Lithological Mapping in the Anti Atlas, Morocco, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2041, https://doi.org/10.5194/egusphere-egu25-2041, 2025.
