•Sub-annual tropical forest disturbance is detected from HLS data.•BFAST Monitor and random forest were used to monitor forest disturbance.•Combining Landsat 8 and Sentinel-2 improves the spatial and ...temporal accuracy.
Accurate sub-annual detection of forest disturbance provides timely baseline information for understanding forest change and dynamics to support the development of sustainable forest management strategies. Traditionally, Landsat imagery was widely used to monitor forest disturbance, but the low temporal density of Landsat observations limits the timely detection of forest disturbance. Recently a new harmonized dataset of Landsat and Sentinel-2 imagery (HLS) has been created to increase the density of observations and provide more frequent images, but the added-value of this dataset for sub-annual tropical forest disturbance monitoring has not been investigated yet. Here, we used all available HLS images acquired from 2016 to 2019 to detect forest disturbance at two tropical forest sites in Tanzania and Brazil. Based on HLS data, forest disturbance was detected by combining normalized difference moisture index (NDMI) and normalized difference vegetation index (NDVI) time series using BFAST monitor and random forest algorithms. To assess the added-value of the HLS time series, we also detected forest disturbance from (i) Landsat-8/OLI time series only and (ii) Sentinel-2 time series only data. The spatial accuracy assessment of forest disturbance detection at the Tanzania site shows that the combined Landsat-8/OLI and Sentinel-2 data achieved the highest overall accuracy (84.5%), more than 3.5% higher than the accuracy of using only Landsat-8/OLI or Sentinel-2. Similarly, for the Brazil site, the overall accuracy of using the combined Landsat-8/OLI and Sentinel-2 data was 95.5%, at least 2% higher than others. In terms of temporal accuracy, the mean time lag of 2.0 months, was achieved from the combined data and Sentinel-2 only at the Tanzania site. This mean time lag is at least one month shorter than that of using Landsat-8/OLI only (3.3 months). At the Brazil site, the mean time lag of forest disturbance detection based on the combined data was 0.22 months, shorter by 0.50 and 0.15 months when compared to using Landsat-8/OLI only (0.72 months) or Sentinel-2 only (0.37 months), respectively. Our results indicate that HLS data is promising for accurate and timely forest disturbance detection particularly in the moist forest where cloud cover is often high.
Remote sensing data play an important role in the monitoring of forest changes. Methods are needed to provide objective estimates of forest loss to support monitoring efforts at various scales, and ...with increasing public availability of remote sensing data, accurate deforestation measurements at high temporal resolution are becoming more realistic. While several time series based methods have recently been described in the literature, there are few studies focusing on tropical forest areas, where low data availability and complex change processes present challenges to forest disturbance monitoring. Here, we present a robust data-driven method to track tropical deforestation and degradation based on Landsat time series data. Based on the previously reported Breaks For Additive Season and Trend Monitor (BFAST Monitor) method (Verbesselt etal., 2012), we show that BFAST Monitor, when applied to Landsat NDVI time series data using sequentially defined monitoring periods, can be used to track small-scale forest disturbances annually in an Afromontane forest system in southern Ethiopia. Using an ordinal logistic regression (OLR) approach, change magnitude, calculated based on differences between observed and expected values in a monitoring period, was found to be an essential predictor variable for disturbances. After applying a NDVI change magnitude threshold of −0.065, overall accuracy was estimated to be 78%, and both producer's and user's accuracy of the disturbance class were estimated to be 73%. The method and results presented here are relevant to tropical countries engaged in REDD+ for whom data availability and complex forest change dynamics limit the ability to reliably track forest disturbances over time.
•Robust data-driven method to track complex forest change processes•Small-scale forest disturbances detected using NDVI time series•Demonstration of BFAST Monitor algorithm on irregular time series data•First forest change study in Ethiopian montane forests at high temporal resolution•Potential of detecting degradation using change magnitude is shown.
•BFAST Monitor implemented with bi-temporal spectral indices for burned area monitoring.•The proposed approach does not require a change magnitude threshold.•Bi-temporal indices increase accuracy and ...reduce post-processing needs.•RBR usage results in higher accuracy and lower time lag errors compared to NBR.
Extreme wildfires are a major agent of disturbance in present-day forest ecosystems and, have serious impacts on society in terms of human casualties and economic losses. Prioritizing and implementing relief and recovery measures after fires require explicit spatial information on the disaster extent. Information for detecting and monitoring ecosystem disturbance due to wildfires is obtainable using time series analysis of remote sensing data. However, the accuracy of this information depends on many factors, such as the input spectral features and the scene characteristics. In this study, we appraise the use of three spectral indices for continuous burned area monitoring in mainland Greece using the Breaks For Additive Seasonal and Trend (BFAST) Monitor unsupervised algorithm. In addition to the commonly employed Normalized Burn Ratio (NBR) we evaluated the use of differenced Normalized Burn Ratio (dNBR) and Relativized Burn Ratio (RBR) generated from a 16-day Landsat-8 time series. The spatiotemporal Overall Accuracy (OA) is assessed using Copernicus Emergency Mapping Service (EMS) grading products and an independent visual interpretation procedure. The latter approach suggests that the using the dNBR (OA = 87.0 %) and RBR (OA = 86.9 %) indices is more efficient than using the NBR time series (OA = 66.4 %). Copernicus EMS validation data show that the RBR-based approach attains a higher OA (80.5 %) -higher than either dNBR (OA = 75.0 %) or NBR (OA = 72.1 %). Evaluating the temporal agreement with Copernicus EMS products, bi-temporal indices achieved mean time lags of 0.023 (dNBR), 0.019 (RBR) and −0.015 (NBR) years. Overall, RBR usage provides the most reliable input for consistent continuous burned area monitoring.
Access to temporally dense time series such as data from the Landsat and Sentinel-2 missions has lead to an increase in methods which aim to monitor land cover change on a per-acquisition rather than ...a yearly basis. Evaluating the accuracy and limitations of these methods can be difficult because validation data are limited and often rely on human interpretation. Simulated time series offer an objective method for evaluating and comparing between change detection algorithms. A set of simulated time series was used to evaluate four change detection methods: (1) Breaks for Additive and Seasonal Trend (BFAST); (2) BFAST Monitor; (3) Continuous Change Detection and Classification (CCDC); and (4) Exponentially Weighted Moving Average Change Detection (EWMACD). In total, 151,200 simulations were generated to represent a range of abrupt, gradual, and seasonal changes. EWMACD was found to give the best performance overall, correctly identifying the true date of change in 76.6% of cases. CCDC performed worst (51.8%). BFAST performed well overall but correctly identified less than 10% of seasonal changes (changes in amplitude, length of season, or number of seasons). All methods showed some decrease in performance with increased noise and missing data, apart from BFAST Monitor which improved when data were removed. The following recommendations are made as a starting point for future studies: EWMACD should be used for detection of lower magnitude changes and changes in seasonality; CCDC should be used for robust detection of complete land cover class changes; EWMACD and BFAST are suitable for noisy datasets, depending on the application; and CCDC should be used where there are high quantities of missing data. The simulated datasets have been made freely available online as a foundation for future work.
Forest cover and vegetation degradation was monitored across the Kavango-Zambezi Transfrontier Conservation Area (KAZA) in southern Africa and the performance of three different methods in detecting ...degradation was assessed using reference data. Breaks for Additive Season and Trend (BFAST) Monitor was used to identify potential forest cover and vegetation degradation using Landsat Normalized Difference Moisture Index (NDMI) time series data. Parametric probability-based magnitude thresholds, non-parametric random forest in conjunction with Soil-Adjusted Vegetation Index (SAVI) time series, and the combination of both methods were evaluated for their suitability to detect degradation for six land cover classes ranging from closed canopy forest to open grassland. The performance of degradation detection was largely dependent on tree cover and vegetation density. Satisfactory accuracies were obtained for closed woodland (user’s accuracy 87%, producer’s accuracy 71%) and closed forest (user’s accuracy 92%, producer’s accuracy 90%), with lower accuracies for open canopies. The performance of the three methods was more similar for closed canopies and differed for land cover classes with open canopies. Highest user’s accuracy was achieved when methods were combined, and the best performance for producer’s accuracy was obtained when random forest was used.
The Colombian Andes foothills have seen an expansion of forest disturbance since the 1950s. While understanding the drivers of disturbance is important for quantifying the implications of land use ...change on regional biodiversity, methods for attributing disturbance to specific drivers of change at a high temporal and spatial resolution are still lacking in the Andes region, in part due to persistent cloud cover. Using 20 years of Landsat images (1996-2015) covering Picachos National Park in the Colombian Andes, we detected sub-annual forest cover disturbances using the Breaks For Additive Season and Trend (BFAST) Monitor algorithm; characterized different types of disturbance using spectral, spatial, and topographic indicators; and attributed causes of forest disturbance such as conversion to pasture, conversion to agriculture, and non-stand replacing disturbance (i.e., thinning) using a Random Forest (RF) classifier. Conversion to pasture has been the main driver of forest disturbance in Picachos, responsible for 11,395 ± 72 ha (17%) of forest cover loss, followed by non-stand replacing disturbance and conversion to agriculture. Disturbance detection had 96% overall agreement with validation data, although we had a high omission error of 21% primarily associated with forest to agriculture conversion. Other change drivers had a much more reliable attribution with forest to pasture conversion or non-stand-replacing disturbance, showing only 1-5% commission and 2-14% omission errors. Our results provide spatially-explicit information on sub-annual disturbances and associated drivers of change that are necessary for evaluating and improving domestic conservation efforts and establishing systematic ecological observations, which is currently absent from Colombia. While effective at revealing forest change dynamics in a geographically remote and socio-politically complex region like Picachos, our approach is highly automated and it can be easily extended to the rest of Andes-Amazon transition belt where low availability of remote sensing data and high cloud cover impede efforts at consistent monitoring of forest cover change dynamics and drivers.
Deforestation is a threat to biodiversity and the world’s climate. As agriculture and mining areas grow, forest loss becomes unbearable for the environment. Consequently, monitoring deforestation is ...crucial for decision makers to create polices. The most reliable deforestation data about the Amazon forest is generated by the Brazil’s National Institute for Space Research (INPE) through its PRODES project. This effort is labor and time intensive because it depends on visual interpretation from experts. Additionally, frequent Amazon’s atmospheric phenomena, such as clouds, difficult image analysis which induces alternative approaches such as time series analysis. One way to increase the number of images of an area consists of using images from different satellites. NASA provides the Harmonized Landsat and Sentinel-2 (HLS) dataset solving spectral dissimilarities of satellite sensors. In this paper, the possibilities of HLS for forest monitoring are explored by applying two deforestation detection methods, Break Detection for Additive Season and Trend (BFAST) monitor and Random Forest, over four different vegetation indices, NDVI, EVI, GEMI and SAVI. The SAVI index used as input for BFAST monitor performed the best in this data setup with 95.23% for deforested pixel, 53.69% for non-deforested pixels. Although the HLS data is described as analysis ready, further pre-processing can enhance the outcome of the analysis. Especially, since the cloud and cirrus cover in the Amazon causes gaps in the dataset, a best pixel method is recommended to create patched images and thus a continuous time series as input for any land cover and land use classification.
Accurate tropic deforestation monitoring using time series requires methods which can capture gradual to abrupt changes and can account for site-specific properties of the environment and the ...available data. The generic time series algorithm BFAST Monitor was tested using Landsat time series at three tropical sites. We evaluated the importance of how specific effects of site and radiometric correction affected the accuracy of deforestation monitoring when using BFAST Monitor. Twelve sets of time series of normalized difference vegetation index (NDVI) Landsat data (2000-2013) were analyzed. Time series properties varied according to site (Brazil, Ethiopia, and Vietnam) and which correction scheme was applied: Atmospheric Correction and Haze Reduction 2 and 3 (ATCOR 2 and 3), Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS), or Dark Object Subtraction (DOS). Mapping accuracy was compared using 1200 reference points per site and consistent designs for sampling, analysis (overall accuracy, user's accuracy, and producer's accuracy), and response (ground truth and very-high-resolution data). With the exception of DOS, mapping accuracies across correction methods were found to be similar but varied greatly with site. Mapping errors were modeled using a set of error parameters that yielded information on data and site-specific environmental properties. Important parameters for characterizing mapping errors were found to be variance of the NDVI and soil signal as well as availability of time series data, and forest edge effects. Based upon the results, local fine-tuning of the algorithm is essential for some areas but for others default settings create satisfactory accuracies.
High-Andean wetlands from northern Chile are considered worldwide biodiversity hot spots, however, they are subdued to high anthropic pressure. The monitoring of state variables, such as vegetation, ...allows to know the ecosystem's global condition, which could be assessed by the analysis of spectral vegetation indices. The main goal of this paper was to detect changes in the high-Andean wetland vegetation, with remote sensing tools, to focalize surveillance efforts and the use of resources from environmental agencies. NDVI time series were constructed spanning from 1986 to 2019 based on Landsat data, which were analyzed based on the vegetation change detection using BFAST Monitor method. Detected changes were categorized to highlight certain types of changes that were considered more relevant. Wetlands were separated in two rankings (A and B) based on detected changes and territorial context. From 5,622 wetlands, 81 were categorized into group A and 510 into group B. One affected wetland was used as study case to assess the method's efficiency, being able to detect changes and assign a relative importance to the case. It is shown that the proposed method has the capacity to detect vegetation degradation processes in high-Andean wetlands and could improve in the efficiency and effectiveness of the environmental agencies control labors over these ecosystems.