Developing a new disturbance index for tracking gradual change of forest ecosystems in the hilly red soil region of southern China using dense Landsat time series
Introduction
Gradual changes are “within-state” changes that occur within the forest landscape and do not involve changes in land cover types (Vogelmann et al., 2012; Vogelmann et al., 2016). These changes mainly relate to diseases and insect pests, vegetation growth and succession, soil erosion, wind, pollution, and climate. Gradual change is not related to the normal phenological cycle; however, if the phenological cycle changes over time, this is also considered a gradual change (Vogelmann et al., 2016). Gradual change generates long-term effects for the landscape surface which can effectively characterize the change trajectory of ground features. Understanding gradual change of the forest landscape is beneficial for the management of forest ecosystems, helping to improve their quality and stability; further, it also aids in the analysis of environmental degradation caused by forest water and soil loss.
Red soil is mainly distributed in the low mountains and hills to the south of the Yangtze River in China (Liang et al., 2010). Due to the interference of natural and human factors, this red soil region has become one of the two major areas of soil erosion and its severity is second only to that of the Loess Plateau (Zhao, 1995). Soil erosion affects forest ecosystems by making them sensitive to disturbance. As disturbance is the main driver of forest ecosystem changes (Cohen et al., 2016); the forest landscape within the red soil region has undergone significant change, and due to long-term changes of various degrees, the forest ecosystem in the hilly red soil region has become vulnerable (Li et al., 2011).
From the perspective of mapping and monitoring the landscape changes, scientists and forest managers have classified the forest landscape changes into three types: abrupt change, seasonal change, and gradual change (Verbesselt et al., 2010; Vogelmann et al., 2012; Vogelmann et al., 2016). Abrupt changes are associated with changes in land cover types, typically caused by deforestation, urbanization, fire, agricultural expansion, and flooding (Hansen and Loveland, 2012; Vogelmann et al., 2016). These changes are relatively simple to detect using current remote sensing technology. Seasonal changes are related to the phenological period of vegetation, which includes the response to the sun angle and climate characteristics (Vogelmann et al., 2016), generally showing periodical change. Gradual change occurs within vegetation communities, and commonly occurs throughout the forest landscape in the same manner as seasonal change; however, gradual change is typically ignored (Vogelmann et al., 2016). This is problematic because, similar to abrupt change events and seasonal change events, the cumulative effects of forest gradual change have substantial impacts on ecosystem processes (Lovett et al., 2006; Vogelmann et al., 2012). At present, most research focuses on abrupt changes rather than gradual change (Beck et al., 2007; Kennedy et al., 2007; Kennedy et al., 2010; Mcdonald et al., 2007; Meddens et al., 2013; Schwantes et al., 2016; Senf et al., 2015; Thessler et al., 2005; Vogelmann et al., 2009; Vogelmann et al., 2012; Vogelmann et al., 2016; Zhu et al., 2016); therefore, more attention should be paid to the gradual change of forests in red soil areas.
Choosing a suitable algorithm is a critical step for gradual change detection. In recent decades, many Landsat time-series change detection algorithms have been proposed and applied for forest gradual change detection (Lucorresponding et al., 2004). For example, Trajectory-Based Change Detection (TBCD) algorithm can be applied to capture gradual forest change trends as well as forest recovery rates (Kennedy et al., 2007), and a method of analyzing the linear regression of trends is used to describe the characteristics of gradual forest changes (Vogelmann et al., 2009; Vogelmann et al., 2012). Continuous Change Detection and Classification (CCDC) algorithm is used to detect wetland changes, drought, and pine beetle outbreaks (Vogelmann et al., 2016; Zhu et al., 2012), while Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr) algorithm has been used to detect drought-, pests-induced forest change, and forest biomass gradual change (Kennedy et al., 2010; Meigs et al., 2011; Pflugmacher et al., 2014; Schwantes et al., 2016). All these change detection algorithms have certain advantages and shortcomings regarding their application in forest gradual change detection. As the study area was located in southern China, sufficient aerial imagery could not be obtained due to cloud cover. Therefore, the LandTrendr algorithm was selected for this study, which can detect forest changes in areas affected by cloud cover with annual imagery, and has been applied using Google Earth Engine (GEE) (Kennedy et al., 2018). The GEE platform can completely access Landsat files and features parallel processing which can enhance the calculation speed of LandTrendr (Gorelick et al., 2017).
There are several challenges for detecting and monitoring forest gradual change, including atmospheric impacts, change index construction, and temporal image gaps (Vogelmann et al., 2009; Vogelmann et al., 2016). It is especially difficult to select the appropriate indicator for detecting forest gradual changes. Therefore, it was necessary to develop an effective disturbance sensitive vegetation index (DSVI) for monitoring forest gradual change. Other vegetation indices have been applied to detect forest gradual change, such as the normalized difference vegetation index (NDVI) (Beck et al., 2007; Vogelmann et al., 2012), enhanced vegetation index (EVI) (Zhu et al., 2016), short wave infrared/near infrared index (SWIR/NIR) (Vogelmann et al., 2009; Vogelmann et al., 2012), and the normalized burn ratio (NBR) (Senf et al., 2015). However, NDVI values are easily saturated and have low sensitivity to subtle changes in high vegetation areas (Li et al., 2007), while EVI increases sensitivity to high biomass areas and does not easily reach saturation, but is not sensitive to subtle disturbances (Huete et al., 2002; Li et al., 2007; Wang et al., 2006). Furthermore, SWIR/NIR is better suited for forest damage monitoring with low leaf area index and is not sensitive to disturbance changes with high leaf area index (Wu et al., 1997), while NBR is only sensitive to changes in the forest before and after a fire (García and Caselles, 1991; Key and Benson, 2004). These indices may not be sensitive enough to detect subtle disturbance changes of forest gradual change. Therefore, our overall goal was to propose a new vegetation index (i.e., DSVI) that can sufficiently detect subtle disturbances in order to analyze the gradual change of the forest in red soil regions. Using the LandTrendr algorithm on the GEE platform and selected evaluation aspects, we present analysis of the long-term processes of forest gradual change by ArcMap 10.2, MATLAB and Microsoft Excel, and evaluate the quality and stability of forest ecosystems in the red soil region of southern China.
Section snippets
Study area
Hengyang (26°07′05″–27°28′24″ N, 110°32′16″–113°16′32″ E) is a typical red soil region located in the south-central part of Hunan Province, within the middle reaches of the Xiang River, and south of Mount Heng. The research area we selected is covered by two Landsat scenes (path/row 123/41 and 123/42) (Fig. 1). Hengyang features a typical basin landscape in the south-central region of the concave surface of the axial zone, and belongs to a subtropical monsoon climate. The winter is warm, while
Comparison of vegetation indices response to forest gradual change
DSVI is developed on the ratio of NDVI and SWIR/NIR (Eq. (3)). When the NDVI value is larger, the DSVI value becomes larger, indicating that the vegetation greenness is higher; similarly, when the SWIR/NIR value is larger, the DSVI value becomes smaller, showing that a higher degree of vegetation disaster. Therefore, high value of DSVI indicates high vegetation coverage, while low value shows high degree of forest disaster.
Fig. 3a shows the gradual change curves of the three indexes and Fig. 3
Discussion
In this study, we developed a new spectral index DSVI. DSVI is composed of red, near-infrared, and short-wave infrared bands. Compared with NDVI, DSVI has one additional short-wave infrared band, making it sensitive to the reflection of vegetation water content (Seelig et al., 2008). With the decrease in leaf moisture, plant short-wave infrared reflectance increases notably. Compared with SWIR/NIR, DSVI has one additional red band. Red has a strong absorption band, is inversely proportional to
Conclusion
In this study, a change detection index, DSVI, was established and the LandTrendr method was used to characterize forest gradual change. Based on the time series of DSVI changes, three evaluation aspects—velocity (V), frequency (F), and variance (S)—were used to analyze the gradual change process of forest landscapes in the red soil region of southern China from 1985 to 2019. The following conclusions can be drawn from the study results:
- 1)
Compared with the other indexes of NDVI and SWIR/NIR, DSVI
Funding
This research was financially supported by the National Natural Science Foundation of China (No. 41871223).
Declaration of Competing Interest
The authors declare no conflict of interest.
Acknowledgments
The authors would like to thank the anonymous reviewers and the editor for their constructive comments and suggestions for this paper.
References (70)
- et al.
Random forest in remote sensing: a review of applications and future directions
ISPRS J. Photogramm. Remote Sens.
(2016) - et al.
Forest disturbance across the conterminous United States from 1985–2012: the emerging dominance of forest decline
For. Ecol. Manag.
(2016) - et al.
Mapping paddy rice planting area in northeastern Asia with Landsat 8 images, phenology-based algorithm and Google Earth Engine
Remote Sens. Environ.
(2016) Status of land cover classification accuracy assessment
Remote Sens. Environ.
(2002)- et al.
Google earth engine: planetary-scale geospatial analysis for everyone
Remote Sens. Environ.
(2017) - et al.
A review of large area monitoring of land cover change using Landsat data
Remote Sens. Environ.
(2012) - et al.
Comparison of Tasseled Cap-based Landsat data structures for use in forest disturbance detection
Remote Sens. Environ.
(2005) - et al.
An integrated Landsat time series protocol for change detection and generation of annual gap-free surface reflectance composites
Remote Sens. Environ.
(2015) The use of time-integrated NOAA NDVI data and rainfall to assess landscape degradation in the arid shrubland of Western Australia
Remote Sens. Environ.
(2003)- et al.
An automated approach for reconstructing recent forest disturbance history using dense Landsat time series stacks
Remote Sens. Environ.
(2010)
Overview of the radiometric and biophysical performance of the MODIS vegetation indices
Remote Sens. Environ.
Trajectory-based change detection for automated characterization of forest disturbance dynamics
Remote Sens. Environ.
Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr — temporal segmentation algorithms
Remote Sens. Environ.
Spatial and temporal patterns of forest disturbance and regrowth within the area of the Northwest Forest Plan
Remote Sens. Environ.
Attribution of disturbance change agent from Landsat time-series in support of habitat monitoring in the Puget Sound region, USA
Remote Sens. Environ.
Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation
Remote Sens. Environ.
Comparison study on ways of ecological vulnerability assessment—a case study in the Hengyang Basin
Procedia Environ. Sci.
Evaluating methods to detect bark beetle-caused tree mortality using single-date and multi-date Landsat imagery
Remote Sens. Environ.
A Landsat time series approach to characterize bark beetle and defoliator impacts on tree mortality and surface fuels in conifer forests
Remote Sens. Environ.
Using the satellite-derived NDVI to assess ecological responses to environmental change
Trends Ecol. Evol.
Using Landsat-derived disturbance and recovery history and lidar to map forest biomass dynamics
Remote Sens. Environ.
Quantifying drought-induced tree mortality in the open canopy woodlands of central Texas
Remote Sens. Environ.
Characterizing spectral–temporal patterns of defoliator and bark beetle disturbances using Landsat time series
Remote Sens. Environ.
A comprehensive evaluation of disturbance agent classification approaches: strengths of ensemble classification, multiple indices, spatio-temporal variables, and direct prediction
ISPRS J. Photogramm. Remote Sens.
Detecting trend and seasonal changes in satellite image time series
Remote Sens. Environ.
Assessing forest damage in high-elevation coniferous forests in Vermont and new Hampshire using thematic mapper data
Remote Sens. Environ.
Use of thematic mapper data for the detection of forest damage caused by the pear thrips
Remote Sens. Environ.
Monitoring forest changes in the southwestern United States using multitemporal Landsat data
Remote Sens. Environ.
Monitoring gradual ecosystem change using Landsat time series analyses: case studies in selected forest and rangeland ecosystems
Remote Sens. Environ.
Perspectives on monitoring gradual change across the continuity of Landsat sensors using time-series data
Remote Sens. Environ.
Object-based cloud and cloud shadow detection in Landsat imagery
Remote Sens. Environ.
Continuous monitoring of forest disturbance using all available Landsat imagery
Remote Sens. Environ.
Including land cover change in analysis of greenness trends using all available Landsat 5, 7, and 8 images: A case study from Guangzhou, China (2000–2014)
Remote Sens. Environ.
A ground-validated NDVI dataset for monitoring vegetation dynamics and mapping phenology in Fennoscandia and the Kola peninsula
Int. J. Remote Sens.
Random forests
Mach. Learn.
Cited by (13)
Demystifying LandTrendr and CCDC temporal segmentation
2022, International Journal of Applied Earth Observation and GeoinformationVegetation recovery and recent degradation in different karst landforms of southwest China over the past two decades using GEE satellite archives
2022, Ecological InformaticsCitation Excerpt :Therefore, it is necessary to fully understand the current distribution, greening trend and vegetation recovery pattern of typical Karst vegetation in China. Google Earth Engine (GEE) is a cloud-based planetary geospatial analysis platform, which applies Google's massive computing power to remote sensing research of global change, including vegetable change (Diao et al., 2021; Hansen et al., 2013; Ye et al., 2021; Zhao et al., 2021), land surface water change (Pekel et al., 2016), land use change (Liu et al., 2018c), etc. GEE provides a massive amount of free remote sensing data and an interactive development environment based on Python or JavaScript language, and supports users to calculate worldwide, long time series and multi-source satellite archives.