Search

Classical statistical methods prove advantageous for small datasets, whereas machine learning algorithms can excel with larger datasets. Our paper challenges this conventional wisdom by addressing a highly significant problem: the identification of burned areas through satellite imagery, that is a clear example of imbalanced data. The methods are illustrated in the North-Central Portugal and the North-West of Spain in October 2017 within a multi-temporal setting of satellite imagery. Daily satellite images are taken from Moderate Resolution Imaging Spectroradiometer (MODIS) products. Our analysis shows that a classical Logistic regression (LR) model competes on par, if not surpasses, a widely employed machine learning algorithm called the extreme gradient boosting algorithm (XGBoost) within this particular domain. [ABSTRACT FROM AUTHOR]

This paper presents empirical models developed through stepwise multiple linear regression to estimate the live fuel moisture content (LFMC) in a Mediterranean area. The models are based on LFMC data measured in 50 field plots, considering four groups with similar bioclimatic characteristics and vegetation types (trees and shrubs). We also applied a species-specific LFMC model for Rosmarinus officinalis in plots with this dominant species. Spectral indices extracted from Sentinel-2 images and their averages over the study time period in each plot with a spatial resolution of 10 m were used as predictors, together with interpolated meteorological, topographic, and seasonal variables. The models achieved adjusted R2 values ranging between 52.1% and 74.4%. Spatial and temporal variations of LFMC in shrub areas were represented on a map. The results highlight the feasibility of developing satellite-derived LFMC operational empirical models in areas with various vegetation types and taking into account bioclimatic zones. The adjustment of data through GAM (generalized additive models) is also addressed in this study. The different error metrics obtained reflect that these models provided a better fit (most adjusted R2 values ranged between 65% and 74.1%) than the linear models, due to GAMs being more versatile and suitable for addressing complex problems such as LFMC behavior. [ABSTRACT FROM AUTHOR]

Copyright of Fire Ecology is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

• NBR-based indices got the strongest correlation with field burn severity regardless of the sensor. • Sentinel-2 MSI only slightly improved the performance of NBR-based indices compared to Landsat 8. • Mono-temporal and relativized indices best predicted vegetation burn severity. • Deimos-1 showed a weake performance to estimate burn severity specially with bi-temporal indices. The development of improved spatial and spectral resolution sensors provides new opportunities to assess burn severity more accurately. This study evaluates the ability of remote sensing indices derived from three remote sensing sensors (i.e., Landsat 8 OLI/TIRS, Sentinel-2 MSI and Deimos-1 SLIM-6-22) to assess burn severity (site, vegetation and soil burn severity). As a case study, we used a megafire (9,939 ha) that occurred in a Mediterranean ecosystem in northwestern Spain. Remote sensing indices included seven reflective, two thermal and four mixed indices, which were derived from each satellite and were validated with field burn severity metrics obtained from CBI index. Correlation patterns of field burn severity and remote sensing indices were relatively consistent across the different sensors. Additionally, regardless of the sensor, indices that incorporated SWIR bands (i.e., NBR-based indices), exceed those using red and NIR bands, and thermal and mixed indices. High resolution Sentinel-2 imagery only slightly improved the performance of indices based on NBR compared to Landsat 8. The dNDVI index from Landsat 8 and Sentinel-2 images showed relatively similar correlation values to NBR-based indices for site and soil burn severity, but showed limitations using Deimos-1. In general, mono-temporal and relativized indices better correlated with vegetation burn severity in heterogeneous systems than differenced indices. This study showed good potential for Landsat 8 OLI/TIRS and Sentinel-2 MSI for burn severity assessment in fire-prone heterogeneous ecosystems, although we highlight the need for further evaluation of Deimos-1 SLIM-6-22 in different fire scenarios, especially using bi-temporal indices. [ABSTRACT FROM AUTHOR]

Burn severity is a key factor in post-fire assessment and its estimation is traditionally restricted to field work and empirical fitting from remotely sensed data. However, the first method is limited in terms of spatial coverage and cost effectiveness and the second is site-and data-specific. Since alternative approaches based on radiative transfer models (RTM) have been usefully applied in retrieving several biophysical plant parameters (leaf area index, water and dry matter content, chlorophyll), this paper has applied the inversion of a simulation model to estimate burn severity in terms of the Composite Burn Index (CBI). The performance of the model inversion method was compared to standard empirical techniques. The study area chosen was a large forest fire in central Spain which occurred in July 2005. The model inversion showed the most accurate estimation for high severity levels (for CBI>2.7, RMSE=0.30) and for unburned areas (CBI<0.5, RMSE=0). In both methodologies, the error associated to CBI from 0.5 to 2.7 was not acceptable (RMSE>0.7), because it is higher than 25% of the total range of the index. Finally, burn severity maps from both methods were compared. [Copyright &y& Elsevier]

Early and detailed information regarding the location and extension of areas affected by forest fires is a critical issue for assessing their effects at several scales. Remote sensing is a valuable tool for burned area mapping, providing spatially explicit information on the scorched areas, even for remote regions.Various sensors have been used for burned land mapping in recent years, covering local (10–30 m pixel size) and global scales (500–1000 m pixel size). Regional inventories require higher detail than global studies, but typically cover larger territories than local studies. For this purpose, the use of medium spatial resolution sensors (100–300 m pixel size) is highly advisable. Few studies have evaluated these sensors, and none have specifically assessed the performance of images acquired by the Argentinian Satellite for Scientific Applications-C/Multispectral Medium Resolution Scanner (SAC-C/MMRS) for mapping burned areas, which is the main objective in this paper.Since information on calibration parameters was not available, raw data was converted into ground reflectance values after a “cross-calibration” method, using Landsat Enhanced Thematic Mapper (ETM+) data as reference. Estimated radiance values showed very high correlation with those for the ETM+ (r2>0.9). Several spectral indices were generated and tested. The Burned Area Index (BAI), derived from red and near infrared (NIR) reflectances showed the greatest sensitivity to discriminate burned areas from other land cover types, although the Normalized Difference Infrared Index (NDII), based on the Red and Short Wave Infrared (SWIR) bands also provided good discrimination capability.Image segmentation was performed by using a multi-threshold approach, based on the BAI and the NDII since these two spectral indices have the highest sensitivity to (1) discriminate burned areas and (2) to avoid confusion with water and cloud shadows.Burned perimeters delineated using a seeded region growing algorithm showed good agreement with the perimeters digitised on Landsat ETM+ data, providing accurate estimation of the burned areas compared to statistics provided by the Spanish forest fire authorities. In the first study area (Almería, south-eastern Spain), the perimeters showed a 90% agreement. In the second study site (Madrid), two fires were studied. The first one (Cadalso, south–west of Madrid) showed an agreement of 64%, and for the second one (Patones, north–east of Madrid), the agreement was 84%. The lower accuracy of the former was caused by the great spatial discontinuity of that fire, which includes many patches of unaffected vegetation within the scar perimeter. [Copyright &y& Elsevier]