Assessing GEDI-NASA system for forest fuels classification using machine learning techniques

Hoffrén, Raúl; Lamelas, María Teresa; Fernández, Juan Ramón de la Riva; Domingo, Darío; Montealegre, Antonio Luis; García-Martín, Alberto; Revilla, Sergio

doi:10.1016/j.jag.2022.103175

Cited by 8 publications

(5 citation statements)

References 49 publications

(80 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Knowing the fuel type in a forest plot is relevant when this information acts as the ground truth of classification models to accurately predict fire behavior over larger forest areas [20]. In this regard, previous studies have noticed common classification discrepancies between the field data (i.e., the fuel type acting as the dependent variable) and the results of predictive models, for instance, between the shrub and tree fuel types [21][22][23], but more commonly between the types of the same dominant stratum, such as between shrub types [16,24] and between tree types [25][26][27]. In a previous work carried out by Hoffrén et al (2023) [26], in the same study area, predictive classification models based on machine learning techniques were performed to classify Prometheus fuel types using the data obtained from a photogrammetric unmanned aerial vehicle.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands

Hoffrén,

Lamelas,

de la Riva

2024

Fire

View full text Add to dashboard Cite

The exposure of Mediterranean forests to large wildfires requires mechanisms to prevent and mitigate their negative effects on the territory and ecosystems. Fuel models synthesize the complexity and heterogeneity of forest fuels and allow for the understanding and modeling of fire behavior. However, it is sometimes challenging to define the fuel type in a structurally heterogeneous forest stand due to the mixture of characteristics from the different types and limitations of qualitative field observations and passive and active airborne remote sensing. This can impact the performance of classification models that rely on the in situ identification of fuel types as the ground truth, which can lead to a mistaken prediction of fuel types over larger areas in fire prediction models. In this study, a handheld mobile laser scanner (HMLS) system was used to assess its capability to define Prometheus fuel types in 43 forest plots in Aragón (NE Spain). The HMLS system captured the vertical and horizontal distribution of fuel at an extremely high resolution to derive high-density three-dimensional point clouds (average: 63,148 points/m2), which were discretized into voxels of 0.05 m3. The total number of voxels in each 5 cm height stratum was calculated to quantify the fuel volume in each stratum, providing the vertical distribution of fuels (m3/m2) for each plot at a centimetric scale. Additionally, the fuel volume was computed for each Prometheus height stratum (0.60, 2, and 4 m) in each plot. The Prometheus fuel types were satisfactorily identified in each plot and were compared with the fuel types estimated in the field. This led to the modification of the ground truth in 10 out of the 43 plots, resulting in errors being found in the field estimation between types FT2–FT3, FT5–FT6, and FT6–FT7. These results demonstrate the ability of the HMLS systems to capture fuel heterogeneity at centimetric scales for the definition of fuel types in the field in Mediterranean forests, making them powerful tools for fuel mapping, fire modeling, and ultimately for improving wildfire prevention and forest management.

show abstract

Section: Introductionmentioning

confidence: 99%

Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands

Hoffrén,

Lamelas,

de la Riva

2024

Fire

View full text Add to dashboard Cite

show abstract

“…Despite using fewer predictor variables (no disturbance history or other spectral indices) than other similar studies [20,21,70] we achieved superior agreement between predicted and observed values. This may be largely attributed to the successful integration of vertical forest complexity measurements from large footprint full-waveform systems with Landsat imagery and terrain attributes, as seen with other forest attributes [69,[73][74][75][76]. Moreover, preliminary analysis in areas without recent disturbances, our models have consistently maintained the relationships between predictor variables.…”

Section: Vertical Forest Complexitymentioning

confidence: 67%

Bridging the Gap: Comprehensive Boreal Forest Complexity Mapping through LVIS Full-Waveform LiDAR, Single-Year and Time Series Landsat Imagery

Diaz-Kloch,

Murray

2023

Remote Sensing

View full text Add to dashboard Cite

The extrapolation of forest structural attributes from LiDAR has traditionally been restricted to local or regional scales, hindering a thorough assessment of single-year versus time series predictors across expansive spatial scales. We extrapolated the vertical complexity captured by the Land, Vegetation, and Ice Sensor (LVIS) full-wave form LiDAR of boreal forests in the Alaska–Yukon–Northwest Territories region, utilizing predictors from Landsat images from 1989 to 2019. This included both single-year and long-term estimates of vegetation indices, alongside constant factors like terrain slope and location. Random forest regression models comparing the single-year and 15-year and 30-year time series models were applied. Additionally, the potential of estimating horizontal forest complexity from vertical complexity was explored using a moving window approach in the Kluane Valley. While the extended time series marginally enhanced model accuracy, a fine-tuned single-year model proved superior (R2 = 0.84, relative RRMSE = 8.4%). In estimating the horizontal complexity, the variance in a 5 × 5 moving window displayed the most promising results, aligning with traditional horizontal structure measures. Single-year Landsat models could potentially surpass time series models in predicting forest vertical complexity, with the added capability to estimate horizontal complexity using variance in a moving window approach.

show abstract

“…Therefore, quality filtering for degrade and beam sensitivity was conducted according to best practices: According to general quality flags (L2A: "quality_flag"; L2B: "l2a_quality_flag", "l2b_quality_flag", L4A: "l2_quality_flag", "l4_quality_flag"), low quality shots (value = 0) could be removed. In addition, degraded shots (value > 0) could be filtered out using the "degrade_flag" and low sensitivity shots (value < 0.95) were removed based on the "sensitivity" attribute [59][60][61]. For each year of GEDI data available, a data set of canopy height, total canopy cover and AGBD was generated and temporally filtered to the months of maximum vegetation growth (June to incl.…”

Section: Data and Pre-processingmentioning

confidence: 99%

Forest Structure Characterization in Germany: Novel Products and Analysis Based on GEDI, Sentinel-1 and Sentinel-2 Data

et al. 2023

View full text Add to dashboard Cite

Monitoring forest conditions is an essential task in the context of global climate change to preserve biodiversity, protect carbon sinks and foster future forest resilience. Severe impacts of heatwaves and droughts triggering cascading effects such as insect infestation are challenging the semi-natural forests in Germany. As a consequence of repeated drought years since 2018, large-scale canopy cover loss has occurred calling for an improved disturbance monitoring and assessment of forest structure conditions. The present study demonstrates the potential of complementary remote sensing sensors to generate wall-to-wall products of forest structure for Germany. The combination of high spatial and temporal resolution imagery from Sentinel-1 (Synthetic Aperture Radar, SAR) and Sentinel-2 (multispectral) with novel samples on forest structure from the Global Ecosystem Dynamics Investigation (GEDI, LiDAR, Light detection and ranging) enables the analysis of forest structure dynamics. Modeling the three-dimensional structure of forests from GEDI samples in machine learning models reveals the recent changes in German forests due to disturbances (e.g., canopy cover degradation, salvage logging). This first consistent data set on forest structure for Germany from 2017 to 2022 provides information of forest canopy height, forest canopy cover and forest biomass and allows estimating recent forest conditions at 10 m spatial resolution. The wall-to-wall maps of the forest structure support a better understanding of post-disturbance forest structure and forest resilience.

show abstract

Assessing GEDI-NASA system for forest fuels classification using machine learning techniques

Cited by 8 publications

References 49 publications

Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands

Evaluation of Handheld Mobile Laser Scanner Systems for the Definition of Fuel Types in Structurally Complex Mediterranean Forest Stands

Bridging the Gap: Comprehensive Boreal Forest Complexity Mapping through LVIS Full-Waveform LiDAR, Single-Year and Time Series Landsat Imagery

Forest Structure Characterization in Germany: Novel Products and Analysis Based on GEDI, Sentinel-1 and Sentinel-2 Data

Contact Info

Product

Resources

About