How can spatial structural metrics improve the accuracy of forest disturbance and recovery detection using dense Landsat time series?

Meng, Yuanyuan; Liu, Xiangnan; Wang, Zheng; Ding, Chao; Zhu, Lihong

doi:10.1016/j.ecolind.2021.108336

Cited by 9 publications

(5 citation statements)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Textural information, which characterizes the spatial structure of a landscape, can enhance the predictive power of spectral information. Meng et al [23] used textural metrics derived from NBR to map the restoration of forests in China from orchards and other land cover types. NBR, like NDVI, is commonly used to monitor forest disturbance and recovery.…”

Section: Discussionmentioning

confidence: 99%

“…Time-series segmentation of spectral information yielded accuracies above 80% for the detection of large open-cast (coal, mineral) mines and soil moisture declines caused by mining [20][21][22]. Temporally segmented spectral metrics derived from SITS better capture changes in forest cover through time when textural metrics are included as covariates [23]. The integration of temporally segmented spectral and textural features could therefore enhance ASM detection, but such an evaluation has not been previously reported.…”

Section: Introductionmentioning

confidence: 97%

See 1 more Smart Citation

Enhanced Detection of Artisanal Small-Scale Mining with Spectral and Textural Segmentation of Landsat Time Series

Fonseca,

Marshall,

Salama

2024

Remote Sensing

View full text Add to dashboard Cite

Artisanal small-scale mines (ASMs) in the Amazon Rainforest are an important cause of deforestation, forest degradation, biodiversity loss, sedimentation in rivers, and mercury emissions. Satellite image data are widely used in environmental decision-making to monitor changes in the land surface, but ASMs are difficult to map from space. ASMs are small, irregularly shaped, unevenly distributed, and confused (spectrally) with other land clearance types. To address this issue, we developed a reliable and efficient ASM detection method for the Tapajós River Basin of Brazil—an important gold mining region of the Amazon Rainforest. We enhanced detection in three key ways. First, we used the time-series segmentation (LandTrendr) Google Earth Engine (GEE) Application Programming Interface to map the pixel-wise trajectory of natural vegetation disturbance and recovery on an annual basis with a 2000 to 2019 Landsat image time series. Second, we segmented 26 textural features in addition to 5 spectral features to account for the high spatial heterogeneity in ASM pixels. Third, we trained and tested a Random Forest model to detect ASMs after eliminating irrelevant and redundant features with the Variable Selection Using Random Forests “ensemble of ensembles” technique. The out-of-bag error and overall accuracy of the final Random Forest was 3.73 and 92.6%, which are comparable to studies mapping large industrial mines with the normalized difference vegetation index (NDVI) and LandTrendr. The most important feature in our study was NDVI, followed by textural features in the near and shortwave infrared. Our work paves the way for future ASM regulation through large area monitoring from space with free and open-source GEE and operational satellites. Studies with sufficient computational resources can improve ASM monitoring with advanced sensors consisting of spectral narrow bands (Sentinel-2, Environmental Mapping and Analysis Program, PRecursore IperSpettrale della Missione Applicativa) and deep learning.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 97%

Enhanced Detection of Artisanal Small-Scale Mining with Spectral and Textural Segmentation of Landsat Time Series

Fonseca,

Marshall,

Salama

2024

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Our results showed that multi-source data and the recovery value of the hybrid method were able to provide reliable refined temporal and spatial information on the recovery of forest ecology in the pixel and regional scales. In general, scholars proposed considering both the multispectral metrics of the post-disturbance and spatial structural characteristics for improving the forest recovery detection accuracy [24,58,59]. The hybrid method was constructed based on the structure-function-habitat three aspects which reflected the ecological meaning and overcame the disadvantage of the fast recovery rate of the spectral index.…”

Section: The Hybrid Model Accuracymentioning

confidence: 99%

“…Each algorithm has its strengths and limitations [23]. Adjusting different parameters for the various locations, aims, and methods will improve the accuracy of the approach [24]. However, the recovery of forest ecosystems detected by optical remote sensing images is the recovery of spectral information and lacks ecological understanding.…”

Section: Introductionmentioning

confidence: 99%

An Approach Integrating Multi-Source Data with LandTrendr Algorithm for Refining Forest Recovery Detection

Zuo

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Disturbances to forests are getting worse with climate change and urbanization. Assessing the functionality of forest ecosystems is challenging because it requires not only a large amount of input data but also comprehensive estimation indicator methods. The object of the evaluation index of forest ecosystem restoration relies on the ecosystem function instead of the area. To develop the appropriate index with ecological implications, we built the hybrid assessment approach including ecosystem structure-function-habitat representatives. It was based on the Normalized Burn Ratio (NBR) spectral indicator and combined with the local forest management inventory (LFMI), Landsat, Light Detection and Ranging (LiDAR) data. The results of the visual interpretation of Google Earth’s historical imagery showed that the total accuracy of the hybrid approach was 0.94. The output of the hybrid model increased as the biodiversity index value increased. Furthermore, to solve the multi-source data availability problem, the random forest model (R2 = 0.78, RMSE = 0.14) with 0.77 total accuracy was built to generate an annual recovery index. A random forest model based on tree age is provided to simplify the hybrid approach while extending the results on time series. The recovery index obtained by the random forest model could facilitate monitoring the forest recovery rate of cold spots. The regional ecological recovery time could be predicted. These two results could provide a scientific basis for forest managers to make more effective forest restoration plans. From the perspective of space, it could ensure that the areas with slow recovery would be allocated enough restoration resources. From the perspective of time, the implementation period of the closed forest policy could also be estimated.

show abstract

“…Additionally, even with an ideal image obtained annually, forest disturbances cannot be captured at the sub-annual scale. To utilize as many useful data as possible and avoid interannual delays,Meng et al tested the LandTrendr algorithm on all available Landsat data in subtropical regions [31]. This method involves the following steps: First, all Landsat data are acquired and seasonally composited into four seasonal time series data (January to March, April to June, July to September, and October to December), following which the LandTrendr algorithm is applied to each of the four seasonal time series separately.…”

Section: Introductionmentioning

confidence: 99%

Sub-Annual Scale LandTrendr: Sub-Annual Scale Deforestation Detection Algorithm Using Multi-Source Time Series Data

Yang,

Wu,

et al. 2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

In cloudy and rainy regions, frequent cloud cover limits clear data obtained using a single optical sensor, posing a substantial challenge for detecting deforestation events on subannual scale. In this study, a sub-annual scale deforestation detection algorithm, namely, the sub-annual scale LandTrendr (SSLT) change detection algorithm, was developed using synergies from multiple data sources. First, a combined time series was constructed by combining Landsat and Sentinel-2 data. Second, a sliding window was applied to spatially normalize the normalized burn ratioand eliminate the effects of forest phenological changes and sensor differences. Finally, an integrated time series was created to fit the SSLT trajectory, and the root mean square error (RMSE) of the fitted trajectory was calculated to determine the segmentation threshold. Pixels with a magnitude of change greater than the RMSE for three consecutive times were marked as deforestation pixels. Application of the algorithm to a subtropical forest with low density of clear observations resulted in spatial and temporal accuracies of 88% and 92.8%, respectively. Conclusively, this method provides accurate and timely identification of deforestation events on a sub-annual scale.

show abstract

How can spatial structural metrics improve the accuracy of forest disturbance and recovery detection using dense Landsat time series?

Cited by 9 publications

References 72 publications

Enhanced Detection of Artisanal Small-Scale Mining with Spectral and Textural Segmentation of Landsat Time Series

Enhanced Detection of Artisanal Small-Scale Mining with Spectral and Textural Segmentation of Landsat Time Series

An Approach Integrating Multi-Source Data with LandTrendr Algorithm for Refining Forest Recovery Detection

Sub-Annual Scale LandTrendr: Sub-Annual Scale Deforestation Detection Algorithm Using Multi-Source Time Series Data

Contact Info

Product

Resources

About