Large-scale forest inventory at the individual tree level is critical for natural resource management decision making. Terrestrial Laser Scanning (TLS) has been used for individual tree level inventory at plot level. However, due to the inflexibility of TLS and the complex scene of large-scale natural forests, it is still challenging to localize and measure every tree. In this paper, we present a framework to conduct large-scale natural forest inventory at the individual tree level by taking advantage of deep learning and Mobile Laser Scanning (MLS). First, a deep learning method, ForestSPG, was developed to perform large-scale semantic segmentation on MLS LiDAR data in natural forests. Then, the forest segmentation results were used for individual stem mapping. Finally, the Diameter at Breast Height (DBH) was measured for each individual stem. Two natural forests mapped with Backpack and Unmanned Aerial Vehicle (UAV) LiDAR were tested. The results showed that the proposed ForestSPG is able to segment large-scale forest LiDAR data into multiple ecologically meaningful classes. The proposed framework was able to localize and measure all 5838 stems at individual tree level in 20 ha of a natural forest in less than 20 min using UAV LiDAR. DBH measurement results on trees’ DBH larger than 38.1 cm (15 in) showed that Backpack LiDAR was able to achieve 1.82 cm of Root Mean Square Error (RMSE) and UAV LiDAR was able to achieve 3.13 cm of RMSE. The proposed framework can not only segment complex forest components in LiDAR data from different platforms but also demonstrate good performance on stem mapping and DBH measurement. Our research provides and automatic and scalable solution for large-scale natural forest inventory at individual tree level, which can be the basis for the large-scale estimation of wood volume and biomass.

The widely adaptable capabilities of artificial intelligence, in particular deep learning and computer vision have led to significant research output regarding flame and smoke detection. The composition of flame and smoke, also described as fire behavior, can be considerably different depending on factors like weather, fuels, and the specific landscape fire is being observed on. The ability to detect definable classes of fire behavior using computer vision has not been explored and could be helpful given it often dictates how firefighters respond to fire situations. To test whether types of fire behavior could be reliably classified, we collected and labeled a unique unmanned aerial system (UAS) image dataset of fire behavior classifications to be trained and validated using You Only Look Once (YOLO) detection models. Our 960 labeled images were sourced from over 21 h of UAS video collected during prescribed fire operations covering a large region of Texas and Louisiana, United States. National Wildfire Coordinating Group (NWCG) fire behavior observations and descriptions served as a reference for determining fire behavior classes during labeling. YOLOv8 models were trained on NWCG Rank 1–3 fire behavior descriptions in grassland, shrubland, forested, and combined fire regimes within our study area. Models were first trained and validated on classifying isolated image objects of fire behavior, and then separately trained to locate and segment fire behavior classifications in UAS images. Models trained to classify isolated image objects of fire behavior consistently performed at a mAP of 0.808 or higher, with combined fire regimes producing the best results (mAP = 0.897). Most segmentation models performed relatively poorly, except for the forest regime model at a box (locate) and mask (segment) mAP of 0.59 and 0.611, respectively. Our results indicate that classifying fire behavior with computer vision is possible in different fire regimes and fuel models, whereas locating and segmenting fire behavior types around background information is relatively difficult. However, it may be a manageable task with enough data, and when models are developed for a specific fire regime. With an increasing number of destructive wildfires and new challenges confronting fire managers, identifying how new technologies can quickly assess wildfire situations can assist wildfire responder awareness. Our conclusion is that levels of abstraction deeper than just detection of smoke or flame are possible using computer vision and could make even more detailed aerial fire monitoring possible using a UAS.

Lakes provide societies and natural ecosystems with valuable services such as freshwater supply and flood control. Water level changes in lakes reflect their natural responses to climatic and anthropogenic stressors; however, their monitoring is costly due to installation and maintenance requirements. With its advanced hardware and computational capabilities, altimetry has become a popular alternative to conventional in-situ gauging, although subject to the temporal availability of altimetric observations. To further improve the temporal resolution of altimetric measurements, we here combine radar altimetry data with Differential Interferometric Synthetic Aperture Radar (D-InSAR), using ten lakes in Sweden as a testing platform. First, we use Sentinel-1A and Sentinel-1B SAR images to generate consecutive six-day baseline interferograms across 2019. Then, we accumulate the phase change of coherent pixels to construct the time series of InSAR-derived water level anomalies. Finally, we retrieve altimetric observations from Sentinel-3, estimate their mean and standard deviation, and apply them to the D-InSAR standardized anomalies. In this way, we build a water-level time series with more temporal observations. In general, we find a strong agreement between water level estimates from the combination of D-InSAR and Satellite Altimetry (DInSAlt) and in-situ observations in eight lakes (Concordance Correlation Coefficient - CCC >0.8) and moderate agreement in two lakes (CCC >0.57). The applicability of DInSAlt is limited to lakes with suitable conditions for double-bounce scattering, such as the presence of trees or marshes. The accuracy of the water level estimates depends on the quality of the altimetry observations and the lake's width. These findings are important considering the recently launched Surface Water and Ocean Topography (SWOT) satellite, whose capabilities could expand our methodology's geographical applicability and reduce its reliance on ground measurements.

While Leaf Area Index (LAI) is critical for understanding forest canopy, photosynthesis and forest growth, traditional field-based LAI measurements are laborious and costly. Remote sensing offers a practical alternative for extensive assessments. Satellite imagery provides broad-scale, long-term monitoring; however, may lack detail needed to guide specific forest management actions. Conversely, Airborne Laser Scanning (ALS) provides accurate LAI estimates at fine spatial detail but is limited by cost and temporal monitoring constraints. Combining ALS data with satellite observations could enhance plantation management decisions by balancing extensive coverage with detailed observations. This study explores the integration of ALS and satellite remote sensing as a comprehensive alternative for assessing LAI and stand volume growth rate (m3/ha/year) in operational Pinus radiata plantations in central-south Chile. Our approach comprised four major steps. First, we applied the Beer-Lambert law using ALS vertical profiles to estimate LAI across a forest plantation (LAIALS). We found that ALS accurately estimated LAI across 121 plots (R2 = 0.82 and RMSE = 0.51). Second, we built a simple linear regression to link LAIALS with the Normalized Difference Moisture Index (NDMI) derived from surface reflectance information from the Landsat/Sentinel-2 satellites, resulting in an R2 of 0.53 and an RMSE of 1.17. This step showed a higher correlation with satellite data compared to using only ground-based LAI estimates (R2 = 0.38; RMSE = 1.18). Third, we transformed biweekly NDMI time series to LAI, then derived peak annual LAI as an indicator of mean annual increment (MAI) (R2 = 0.51; RMSE = 5.27 m³/ha/year). This allowed us to characterize stand growth and LAI on a yearly wall-to-wall basis. Throughout the modelling steps, we incorporated error propagation, allowing final estimates to be error bounded. This integrated approach serves as a tool for identifying and visualizing growth irregularities, guiding adaptive management strategies to maintain or enhance stand productivity over time.

Canopy cover (CC) quantifies the proportion of canopy materials projected vertically onto the ground surface. CC is a crucial canopy structural variable and is commonly used in many ecological and climatic models. The vertical CC profile product is currently available from the Global Ecosystem Dynamics Investigation (GEDI). However, detailed information about the accuracy and uncertainty of the GEDI vertical CC profile product remains limited. The objective of this study is to validate the GEDI CC product over selected forest sites using reference values derived from digital hemispherical photography (DHP), airborne laser scanning (ALS) point clouds, and simulated waveforms. The accuracy of CC was quantified and analyzed regarding GEDI observation conditions, waveform processing, and estimation methods. The results show that the GEDI total CC correlates well with those estimated from DHP, ALS, and simulated waveform data (r2 = 0.65, 0.71, and 0.71, respectively) but is systematically underestimated (bias = −0.05, −0.11, and −0.07, respectively) based on reference data. Compared with the ALS-estimated CC, needleleaf forest shows the highest correlation for vertical CC (r2 ≥ 0.65) and shrubland shows the lowest bias for total CC (bias = −0.13). The mean absolute error (MAE) of the GEDI CC decreases from 0.15 to 0.09 as the estimation height increases from ground to 35 m. The GEDI total CCs derived from the waveform interpretation algorithms A2 and A6 display the highest r2 (≥ 0.6) and smallest RMSE (≤ 0.23) compared to those of the other algorithms. The CC accuracy increases with beam sensitivity and decreases with increasing canopy cover. The GEDI CC was improved at moderate CC values using a canopy-to-ground backscattering coefficient ratio (ρv/ρg) determined with the regression method. The partial difference between GEDI CC and ALS CC is attributed to definitional discrepancies. Further improvement of the CC algorithm can be made by using vegetation-specific waveform processing algorithms and realistic ρv/ρg values.

Savannas cover more than 20% of the Earth and account for the third largest stock of global aboveground biomass yet estimates of their above ground biomass density (AGBD) are very inaccurate. The Global Ecosystem Dynamic Investigation (GEDI) sensor provides near-global full-waveform LiDAR data with 25 m footprints, from which various structural metrics are derived that are used to predict footprint level AGBD. The current GEDI L4A AGBD product uses a comprehensive Forest Structure and Biomass Database (FSBD) to develop models for specific plant functional types and geographic regions, but southern African savannas have been underrepresented in the reference data. The objectives of this study were to (i) validate GEDI L4A AGBD in South African savannas using field measurements and ALS datasets and (ii) develop and evaluate local GEDI footprint-level AGBD estimates from multiple L2A and L2B metrics. The local GEDI AGBD models outperformed GEDI L4A AGBD (R2 = 0.42, RMSE = 12 Mg/ha, %RMSE = 79.5%) with higher R2 and smaller error measures. The local GEDI AGBD using a random forest model (RF) had the highest R2 of 0.71 and lowest %RMSE of 53.3%, while the generalized linear model (GLM) results provided the lowest Relative Mean Systematic Deviation (RMSD) of 9.2%, which was half that of RF model. L4A significantly underestimated AGBD with an RMSD up to −37%. This highlights the importance and benefits of local calibration of biomass models to unlock the full potential of GEDI metrics for estimating AGBD. The field and ALS data have subsequently been contributed to the GEDI FSBD and should be used in calibration of future versions of GEDI L4A AGBD product. This research paves the way for the integration of the local GEDI AGBD estimates with other sensors, notable the eminent NISAR mission, to derive regional to global gridded AGBD products that will enable the monitoring of savanna carbon stocks.

The increase in Earth observations from space in recent years supports improved quantification of carbon storage by terrestrial vegetation and fosters studies that relate satellite measurements to biomass retrieval algorithms. However, satellite observations are only indirectly related to the carbon stored by vegetation. While ground surveys provide biomass stock measurements to act as reference for training the models, they are sparsely distributed. Here, we addressed this problem by designing an algorithm that harnesses the interplay of satellite observations, modeling frameworks and field measurements, and generated global estimates of above-ground biomass (AGB) density that meet the requirements of the scientific community in terms of accuracy, spatial and temporal resolution. The design was adapted to the amount, type and spatial distribution of satellite data available around the year 2020. The retrieval algorithm estimated AGB annually by merging estimates derived from C- and L-band synthetic aperture radar (SAR) backscatter observations with a Water Cloud type of model and does not rely on AGB reference data at the same spatial scale as the SAR data. This model is integrated with functions relating to forest structural variables that were trained on spaceborne LiDAR observations and sub-national AGB statistics. The yearly estimates of AGB were successively harmonized using a cost function that minimizes spurious fluctuations arising from the moderate-to-weak sensitivity of the SAR backscatter to AGB. The spatial distribution of the AGB estimates was correctly reproduced when the retrieval model was correctly set. Over-predictions occasionally occurred in the low AGB range (< 50 Mg ha-1) and under-predictions in the high AGB range (> 300 Mg ha-1). These errors were a consequence of sometimes too strong generalizations made within the modeling framework to allow reliable retrieval worldwide at the expense of accuracy. The precision of the estimates was mostly between 30% and 80% relative to the estimated value. While the framework is well founded, it could be improved by incorporating additional satellite observations that capture structural properties of vegetation (e.g., from SAR interferometry, low-frequency SAR, or high-resolution observations), a dense network of regularly monitored high-quality forest biomass reference sites, and spatially more detailed characterization of all model parameters estimates to better reflect regional differences.

Since 2001, the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on board the Aqua and Terra platforms has made great strides in providing information on global burned areas (BA). However, the MODIS mission is nearing its end. The Visible Infrared Imaging Radiometer Suite (VIIRS) sensors, presented as the MODIS Aqua heritage, could be an excellent alternative to ensure the temporal continuity of this information at a moderate resolution. This paper describes and evaluates the effectiveness of our developed hybrid algorithm, which utilizes VIIRS reflectance and active fire products on the Google Earth Engine platform, in producing efficient information about BA. The study investigates the algorithm's performance in sub-Saharan Africa as the region of interest in 2019, using biweekly outputs and a spatial resolution of 250 m. The algorithm encompasses several steps, including pre-processing individual scenes, creating cloud-free composites, generating binary reference data for burned and non-burned areas, conducting a supervised classification using random forest, and performing region shaping. The VIIRS-BA final product, which includes three confidence levels (low, moderate, and high) known as the uncertainty layer, is compared to four other burned area products. The validation is conducted against 27 reference sampling units from the Sentinel-2 Burned Area Reference Database dataset, allowing for a comprehensive uncertainty assessment across five various biomes. The VIIRS-BA product identified 5.1 million km2 of BA, which was significantly larger than other global coarse resolution BA products such as FireCCI51, FireCCIS310, and MCD64A1 and close to the fine resolution FireCCISFD20 with a difference of 7.3%. The differences were less significant in biomes such as “Tropical Savannas” and “Temperate Grasslands” which are characterized by persistent biomass burning. Based on a stratified random sampling, the validation results demonstrate varying levels of accuracy for the VIIRS-BA product across different confidence levels. The commission error (CE) ranges from 7.8% to 23.4%, while the omission error (OE) falls between 29.4% and 58.8%. Notably, there is a significant reduction in OE (ranging from 40.7% to 50.5%) compared to global BA products like FireCCI51, FireCCIS310, and MCD64A1. When compared to VIIRS-BA, the FireCCISFD20 regional product has a 37% better OE performance. While VIIRS-BA shows great potential in detecting fires that global products miss, the VIIRS-BA with low confidence level tends to overestimate BA in regions with high fire activity. To address this, future versions of the algorithm will integrate the updated VIIRS reflectance data alongside VIIRS active fire from the National Oceanic and Atmospheric Administration to reduce CE and improve understanding spatial patterns.

The photosynthetic rate has a nonlinear relationship with PAR during the day. We previously developed an algorithm for estimating GPP capacity, which is defined GPP under low-stress condition, using light response curves (LRCs). In this study, we studied the characteristics of LRC parameters of the initial slope and the maximum gross photosynthesis rate (Pmax), and formulas to calculate Pmax from the relationship between the chlorophyll index of the green and near-infrared (NIR) bands (CIgreen) and the GPP capacity at PAR = 2000 μmol m–2 s–1 (GP2000) for nine vegetation types spanning tropical to subarctic climates on the Eurasian and North American continents using eddy covariance flux measurements and Moderate Resolution Imaging Spectrometer (MODIS) data. The slope of the relationship between CIgreen and GP2000 was highest for sites dominated by herbaceous plants such as open shrubland, savanna, and cropland (rice paddy); it was lower at sites dominated by woody plants. The yearly GPP/GPP capacity ratio was close to one in flux data. When the method was applied to satellite data, the daily GPP capacity exhibited a similar seasonal pattern to that of the Flux GPP and MODIS GPP products. Under high dryness conditions, Flux GPP showed the drop from the GPP capacity estimated from CIgreen and diurnal PAR data around noon, and they were nearly identical during the early morning and late afternoon. The instantaneous GPP capacity could be considered the baseline of the instantaneous GPP with stress-free conditions and important for quantifying midday depression at the sub-day scale.

Snow albedo is a key geophysical parameter that controls the energy exchanges between the atmosphere and Earth’s surfaces and has been widely utilized in climatic and environmental change studies. However, recent studies have demonstrated that current albedo satellite products still have large uncertainties in snow-covered areas. In this study, we estimated the blue-sky shortwave albedo of snow surfaces using the eXtreme Gradient Boosting (XGBoost) algorithm with Moderate Resolution Imaging Spectroradiometer (MODIS) top-of-atmosphere (TOA) reflectance values, ERA-5 land reanalysis snow parameters (e.g., snow cover, snow density and snow depth water equivalent) and in situ measurements. In the XGBoost model, the MODIS MCD43 albedo values were input as prior knowledge, and the random sample validation results showed that the R2 and root mean square error (RMSE) values of this model were approximately 0.953 and 0.044, respectively. The typical sites for independent validation were subjected to in situ measurements at the UPE_L, AWS5, and CA_ARB sites. Finally, the retrieved XGBoost albedo values were compared with the official NASA MODIS (MCD43, collection 6), the Global Land Surface Satellite (GLASS), and the National Oceanic and Atmospheric Administration (NOAA) Visible Infrared Imaging Radiometer Suite (VIIRS) SURFALB albedo products. The validation results indicated that the proposed approach achieved much greater accuracy (RMSE = 0.052, bias = 0.002) than did the corresponding official MODIS (RMSE = 0.087, bias = -0.033), GLASS (RMSE = 0.089, bias = -0.031) and VIIRS SURFALB albedo (RMSE = 0.100, bias = -0.032) products. The improved shortwave albedo captured the rapid temporal changes in surface snow conditions.