Evaluating remote sensing datasets and machine learning algorithms for mapping plantations and successional forests in Phnom Kulen National Park of Cambodia

Spatial-temporal variations in deforestation hotspots in Sumatra and Kalimantan from 2001-2018

Tropical deforestation varies temporally and spatially which can inhibit the ability of existing protected areas to stem forest loss. Identifying the spatial-temporal distribution of deforestation and its concentration can help decision makers decide conservation priorities and leverage limited resources. This study assessed how topographic and anthropogenic variables affect deforestation patterns within and outside protected areas on the islands of Sumatra and Kalimantan in Indonesia. Emerging hotspot analysis (EHA) was used to evaluate spatial and temporal trends of forest loss on the Hansen annual forest loss data for these islands from 2001-2018. For the two islands, most hotspots were detected outside protected areas; those within protected areas were mainly concentrated at boundaries, where lower elevation/slope and high human pressure could be observed. New hotspots were identified within the three PAs in Sumatra, while three kinds of hotspots (consecutive, oscillating, and sporadic) were found in the two PAs of Kalimantan (Kutai and Teluk Kelumpang). Areas with high human pressure (average human footprint higher than 12) were covered by a high density of hotspots. The results identify specific areas where forest loss has emerged recently, which could indicate a conservation priority. It is suggested that new protected areas be established in locations showing intensifying and persistent hotspots-those where deforestation has occurred for ≥16 of 18 years of the study period.

Assessment of k-Nearest Neighbor and Random Forest Classifiers for Mapping Forest Fire Areas in Central Portugal Using Landsat-8, Sentinel-2, and Terra Imagery

Forest fires threaten the population’s health, biomass, and biodiversity, intensifying the desertification processes and causing temporary damage to conservation areas. Remote sensing has been used to detect, map, and monitor areas that are affected by forest fires due to the fact that the different areas burned by a fire have similar spectral characteristics. This study analyzes the performance of the k-Nearest Neighbor (kNN) and Random Forest (RF) classifiers for the classification of an area that is affected by fires in central Portugal. For that, image data from Landsat-8, Sentinel-2, and Terra satellites and the peculiarities of each of these platforms with the support of Jeffries–Matusita (JM) separability statistics were analyzed. The event under study was a 93.40 km2 fire that occurred on 20 July 2019 and was located in the districts of Santarém and Castelo Branco. The results showed that the problems of spectral mixing, registration date, and those associated with the spatial resolution of the sensors were the main factors that led to commission errors with variation between 1% and 15.7% and omission errors between 8.8% and 20%. The classifiers, which performed well, were assessed using the receiver operating characteristic (ROC) curve method, generating maps that were compared based on the areas under the curves (AUC). All of the AUC were greater than 0.88 and the Overall Accuracy (OA) ranged from 89 to 93%. The classification methods that were based on the kNN and RF algorithms showed satisfactory results.

SAR and LIDAR Datasets for Building Damage Evaluation Based on Support Vector Machine and Random Forest Algorithms—A Case Study of Kumamoto Earthquake, Japan

The evaluation of buildings damage following disasters from natural hazards is a crucial step in determining the extent of the damage and measuring renovation needs. In this study, a combination of the synthetic aperture radar (SAR) and light detection and ranging (LIDAR) data before and after the earthquake were used to assess the damage to buildings caused by the Kumamoto earthquake. For damage assessment, three variables including elevation difference (ELD) and texture difference (TD) in pre- and post-event LIDAR images and coherence difference (CD) in SAR images before and after the event were considered and their results were extracted. Machine learning algorithms including random forest (RDF) and the support vector machine (SVM) were used to classify and predict the rate of damage. The results showed that ELD parameter played a key role in identifying the damaged buildings. The SVM algorithm using the ELD parameter and considering three damage rates, including D0 and D1 (Negligible to slight damages), D2, D3 and D4 (Moderate to Heavy damages) and D5 and D6 (Collapsed buildings) provided an overall accuracy of about 87.1%. In addition, for four damage rates, the overall accuracy was about 78.1%.

Mapping the Natural Distribution of Bamboo and Related Carbon Stocks in the Tropics Using Google Earth Engine, Phenological Behavior, Landsat 8, and Sentinel-2

Although vegetation phenology thresholds have been developed for a wide range of mapping applications, their use for assessing the distribution of natural bamboo and the related carbon stocks is still limited, especially in Southeast Asia. Here, we used Google Earth Engine (GEE) to collect time-series of Landsat 8 Operational Land Imager (OLI) and Sentinel-2 images and employed a phenology-based threshold classification method (PBTC) to map the natural bamboo distribution and estimate carbon stocks in Siem Reap Province, Cambodia. We processed 337 collections of Landsat 8 OLI for phenological assessment and generated 121 phenological profiles of the average vegetation index for three vegetation land cover categories from 2015 to 2018. After determining the minimum and maximum threshold values for bamboo during the leaf-shedding phenology stage, the PBTC method was applied to produce a seasonal composite enhanced vegetation index (EVI) for Landsat collections and assess the bamboo distributions in 2015 and 2018. Bamboo distributions in 2019 were then mapped by applying the EVI phenological threshold values for 10 m resolution Sentinel-2 satellite imagery by accessing 442 tiles. The overall Landsat 8 OLI bamboo maps for 2015 and 2018 had user’s accuracies (UAs) of 86.6% and 87.9% and producer’s accuracies (PAs) of 95.7% and 97.8%, respectively, and a UA of 86.5% and PA of 91.7% were obtained from Sentinel-2 imagery for 2019. Accordingly, carbon stocks of natural bamboo by district in Siem Reap at the province level were estimated. Emission reductions from the protection of natural bamboo can be used to offset 6% of the carbon emissions from tourists who visit this tourism-destination province. It is concluded that a combination of GEE and PBTC and the increasing availability of remote sensing data make it possible to map the natural distribution of bamboo and carbon stocks.