Mapping forest canopy fuel parameters at European scale using spaceborne LiDAR and satellite data

Remote Sensing of Environment, Год журнала: 2024, Номер 303, С. 114005 - 114005

Опубликована: Янв. 30, 2024

Spatially explicit data on forest canopy fuel parameters provide critical information for wildfire propagation modelling, emission estimations and risk assessment. LiDAR observations enable accurate retrieval of the vertical structure vegetation, which makes them an excellent alternative characterising structures. In most cases, parameterisation has been based Airborne Laser Scanning (ALS) observations, are costly best suited local research. Spaceborne acquisitions overcome limited spatiotemporal coverage airborne systems, as they can cover much wider geographical areas. However, do not continuous data, requiring spatial interpolation methods to obtain wall-to-wall information. We developed a two-step, easily replicable methodology estimate entire European territory, from Global Ecosystem Dynamics Investigation (GEDI) sensor, onboard International Space Station (ISS). First, we simulated GEDI pseudo-waveforms discrete ALS about plots. then used metrics derived mean height (Hm), (CC) base (CBH), national inventory reference. The RH80 metric had strongest correlation with Hm all types (r = 0.96–0.97, Bias −0.16-0.30 m, RMSE 1.53–2.52 rRMSE 13.23–19.75%). A strong was also observed between ALS-CC GEDI-CC 0.94, −0.02, 0.09, 16.26%), whereas weaker correlations were obtained CBH 0.46, 0 0.89 39.80%). second stage generate maps continent Europe at resolution 1 km using GEDI-based estimates within-fuel polygons covered by footprints. available some (mainly Northern latitudes, above 51.6°N). these estimated random regression models multispectral SAR imagery biophysical variables. Errors higher than direct retrievals, but still within range previous results 0.72–0.82, −0.18-0.29 3.63–4.18 m 28.43–30.66% Hm; r 0.82–0.91, 0, 0.07–0.09 10.65–14.42% CC; 0.62–0.75, 0.01–0.02 0.60–0.74 19.16–22.93% CBH). Uncertainty provided grid level, purpose considered individual errors each step in methodology. final outputs, publicly (https://doi.org/10.21950/KTALA8), estimation three modelling crown fire potential demonstrate capacity improve characterisation models.

Язык: Английский

---

Evaluating the Potential of Sentinel-2 Time Series Imagery and Machine Learning for Tree Species Classification in a Mountainous Forest

Remote Sensing, Год журнала: 2024, Номер 16(2), С. 293 - 293

Опубликована: Янв. 11, 2024

Accurate and reliable information on tree species composition distribution is crucial in operational sustainable forest management. Developing a high-precision map based time series satellite data an effective cost-efficient approach. However, we do not quantitatively know how the scale of acquisitions contributes to complex mapping. This study aimed produce detailed typical zone Changbai Mountains by incorporating Sentinel-2 images, topography data, machine learning algorithms. We focused exploring effects three-year within monthly, seasonal, yearly scales classification ten dominant species. A random (RF) support vector (SVM) were compared employed continuous The results showed that with monthly datasets (overall accuracy (OA): 83.38–87.45%) outperformed seasonal (OA:72.38–85.91%), RF (OA: 81.70–87.45%) was better than SVM 72.38–83.38%) at processing same datasets. Short-wave infrared, normalized vegetation index, elevation most important variables for classification. highest 87.45% achieved combining RF, datasets, information. In terms single species’ accuracy, F1 scores ranged from 62.99% (Manchurian ash) 97.04% (Mongolian Oak), eight them obtained high greater 87%. confirmed algorithms have great potential accurate mapping mountainous regions.

Язык: Английский

---

Soil organic carbon stocks in native forest of Argentina: a useful surrogate for mitigation and conservation planning under climate variability

Ecological Processes, Год журнала: 2024, Номер 13(1)

Опубликована: Янв. 5, 2024

Abstract Background The nationally determined contribution (NDC) presented by Argentina within the framework of Paris Agreement is aligned with decisions made in context United Nations Framework Convention on Climate Change (UNFCCC) reduction emissions derived from deforestation and forest degradation, as well carbon conservation (REDD+). In addition, climate change constitutes one greatest threats to biodiversity ecosystem services. However, soil organic (SOC) stocks native forests have not been incorporated into Forest Reference Emission Levels calculations for planning under variability due a lack information. objectives this study were: (i) model SOC 30 cm at national scale using climatic, topographic vegetation predictor variables, (ii) relate spatial–temporal remotely sensed indices determine concerns high inter-annual variability. Methods We used 1040 samples (0–30 cm) generate spatially explicit estimates spatial resolution approximately 200 m. selected 52 potential predictive environmental covariates, which represent key factors distribution SOC. All covariate maps were uploaded Google Earth Engine cloud-based computing platform subsequent modelling. To variability, we employed satellite-derived based Enhanced Vegetation Index (EVI) land surface temperature (LST) images Landsat imagery. Results depth) prediction accounted 69% variation property across whole coverage Argentina. Total mean stock reached 2.81 Pg C (2.71–2.84 probability 90%) total area 460,790 km 2 , where Chaco represented 58.4% stored, followed Andean Patagonian (16.7%) Espinal (10.0%). was fitted function regional climate, greatly influenced ecosystems, including precipitation (annual warmest quarter) (day temperature, seasonality, maximum month, month night monthly minimum temperature). Biodiversity levels regions. Conclusions Kyoto Protocol REDD+, information present work estimate can be annual National Inventory Report assist management proposals. It also gives insight how more resilient reduce impact loss.

Язык: Английский

---

Forest aboveground biomass estimation by GEDI and multi-source EO data fusion over Indian forest

International Journal of Remote Sensing, Год журнала: 2024, Номер 45(4), С. 1304 - 1338

Опубликована: Фев. 2, 2024

Monitoring changes in carbon stocks through forest biomass assessment is crucial for cycle studies. However, challenges obtaining timely and reliable ground measurements hinder creation of the spatially continuous maps aboveground density (AGBD). This study proposes an approach generating (AGBD) by combining Global Ecosystem Dynamics Investigation (GEDI) LiDAR-based data with open-access earth observation (EO) data. The key contribution lies systematic evaluation various model configurations to select optimal AGBD generation. considered configurations, including predictor sets, spatial resolution, beam selection, sensitivity thresholds. We used a Random Forest model, trained five-fold cross-validation on 80% total data, estimate Indian region. Model performance was assessed using 20% independent test dataset. Results, Sentinel-1 2 predictors, yielded R2 values 0.55 0.60 RMSE 48.5 56.3 Mg/ha. Incorporating agroclimatic zone attributes improved (R2: 0.59 0.69, RMSE: 42.2 53.3 Mg/ha). selection top 15 which favoured features from Sentinel-2, DEM, attributes, zones, GEDI >0.98, 0.64 46.59 results underscore significance incorporating like agro-climatic zones need considering types shot characteristics. top-performing validated Simdega, Jharkhand 0.74, 39.3 Mg/ha), demonstrating methodological potential this approach. Overall, emphasizes prospects integrating multi-source EO produce (AGB) fusion.

Язык: Английский

---

Assessment of Machine Learning Algorithms for Land Cover Classification in a Complex Mountainous Landscape

Journal of Geovisualization and Spatial Analysis, Год журнала: 2024, Номер 8(2)

Опубликована: Авг. 30, 2024

Abstract Mapping land cover (LC) in mountainous regions, such as the Gilgit-Baltistan (GB) area of Pakistan, presents significant challenges due to complex terrain, limited data availability, and accessibility constraints. This study addresses these by developing a robust, data-driven approach classify LC using high-resolution Sentinel-2 (S-2) satellite imagery from 2019 within Google Earth Engine (GEE). The research evaluated performance various machine learning (ML) algorithms, including classification regression tree (CART), maximum entropy (gmoMaxEnt), minimum distance (minDistance), support vector (SVM), random forest (RF), without extensive hyperparameter tuning. Additionally, ten different scenarios based on band combinations S-2 were used input for running ML models. was performed 2759 sample points, with 70% training 30% validation. results indicate that RF algorithm outperformed all other classifiers under scenario S1 (using 10 bands), achieving an overall accuracy (OA) 0.79 kappa coefficient 0.76. final RF-based mapping shows following percentage distribution: barren (46.7%), snow (22.9%), glacier (7.9%), grasses (7.2%), water (4.7%), wetland (2.9%), built-up (2.7%), agriculture (1.9%), (1.2%). It is suggested best identified classifier GEE environment should be advanced multi-source image tuning increase OA. it build capacity stakeholders GB better monitoring changes resource management geospatial big data.

Язык: Английский

---

Calibration of GEDI footprint aboveground biomass models in Mediterranean forests with NFI plots: A comparison of approaches

Journal of Environmental Management, Год журнала: 2025, Номер 375, С. 124313 - 124313

Опубликована: Янв. 31, 2025

Observations from the NASA Global Ecosystem Dynamics Investigation (GEDI) provide global information on forest structure and biomass. Footprint-level predictions of aboveground biomass density (AGBD) in GEDI mission are based training data sourced sparsely distributed field plots coincident with airborne laser scanning surveys. National Forest Inventories (NFI) rarely used to calibrate footprint models because their sampling positional accuracy prevent accurate colocation or ALS. This omission can limit harmonization jurisdictional estimates NFI's GEDI; however, there methods available improve NFI footprints. Focusing Mediterranean forests Spain, we compared different approaches collocation data: (i) simulated waveforms ALS; (ii) nearest-neighbor on-orbit waveforms; (iii) imputed plot locations using a novel geostatistical method. These potential solutions local performance address systematic deviations between estimates. We assess advantages limitations these locally quantify impact geolocation errors reference data. The new each method were predict level AGBD, which then gridded for province North-West Spain. It was found that imputation approach is not sensitive common geolocation, but it outperform ALS-based simulation some cases, highlighting benefit multiple footprints proximate improving predictions. research provides users benchmark techniques locally-calibrate models.

Язык: Английский

---

Improving National Forest Mapping in Romania Using Machine Learning and Sentinel-2 Multispectral Imagery

Remote Sensing, Год журнала: 2025, Номер 17(4), С. 715 - 715

Опубликована: Фев. 19, 2025

Forest attributes, such as standing stock, diameter at breast height (DBH), tree height, and basal area, are critical for effective forest management; yet, traditional estimation methods remain labor-intensive often lack the spatial detail required contemporary decision-making. This study addresses these challenges by integrating machine learning algorithms with high-resolution remotely sensed data rigorously collected ground truth measurements to produce accurate, national-scale maps of attributes in Romania. To ensure reliability model predictions, extensive field campaigns were conducted across representative Romanian forests. During campaigns, detailed recorded every within selected plots. For each tree, DBH was measured directly, heights obtained either direct measurement—using hypsometers or clinometers—or, when not feasible, applying well-established DBH—height allometric relationships that have been calibrated local types. comprehensive approach collection, supplemented an independent dataset from Brasov County using same protocols, allowed robust training validation models. evaluates performance three algorithms—Random (RF), Classification Regression Trees (CART), Gradient Boosting Tree Algorithm (GBTA)—in predicting Sentinel-2 satellite imagery. While Random consistently delivered high R2 values low root mean square errors (RMSE) all GBTA showed particular strength CART excelled area but less reliable other attributes. A sensitivity analysis multiple resolutions revealed varied significantly changes resolution, emphasizing importance selecting appropriate scale accurate mapping. By focusing on both methodological advancements applications rigorous, empirical this provides a clear solution problem obtaining reliable, spatially attribute maps.

Язык: Английский

---

A multi-sensor approach allows confident mapping of forest canopy fuel load and canopy bulk density to assess wildfire risk at the European scale

Remote Sensing of Environment, Год журнала: 2025, Номер 318, С. 114578 - 114578

Опубликована: Янв. 27, 2025

Язык: Английский

---

High-Resolution Canopy Height Mapping: Integrating NASA’s Global Ecosystem Dynamics Investigation (GEDI) with Multi-Source Remote Sensing Data

Remote Sensing, Год журнала: 2024, Номер 16(7), С. 1281 - 1281

Опубликована: Апрель 5, 2024

Accurate structural information about forests, including canopy heights and diameters, is crucial for quantifying tree volume, biomass, carbon stocks, enabling effective forest ecosystem management, particularly in response to changing environmental conditions. Since late 2018, NASA’s Global Ecosystem Dynamics Investigation (GEDI) mission has monitored global structure using a satellite Light Detection Ranging (LiDAR) instrument. While GEDI collected billions of LiDAR shots across near-global range (between 51.6°N >51.6°S), their spatial distribution remains dispersed, posing challenges achieving complete coverage. This study proposes evaluates an approach that generates high-resolution height maps by integrating data with Sentinel-1, Sentinel-2, topographical ancillary through three machine learning (ML) algorithms: random forests (RF), gradient boost (GB), classification regression trees (CART). To achieve this, the secondary aims included following: (1) assess performance ML algorithms, RF, GB, CART, predicting heights, (2) evaluate our reference from models (CHMs), (3) compare other two existing maps. RF GB were top-performing best 13.32% 16% root mean squared error broadleaf coniferous respectively. Validation proposed revealed 100th 98th percentile, followed average 75th, 90th, 95th, percentiles (AVG), most accurate metrics real heights. Comparisons between predicted CHMs demonstrated predictions stands (R-squared = 0.45, RMSE 29.16%).

Язык: Английский

---

High-resolution mapping of forest structure and carbon stock using multi-source remote sensing data in Japan

Remote Sensing of Environment, Год журнала: 2024, Номер 312, С. 114322 - 114322

Опубликована: Июль 26, 2024

Язык: Английский

---