A new method for grassland degradation monitoring by vegetation species composition using hyperspectral remote sensing
Introduction
As an important part of the terrestrial ecosystem, the grassland ecosystem not only provides a material basis for animal husbandry, but also plays a critical role in wind erosion protection, sand fixation, and water conservation. With the rapidly growing global demand for livestock products and the emphasis on the sustainable development of the ecological environment, grassland degradation is an important problem confronting countries around the world (R, 2010, Ayantunde et al., 2011, Boval and Dixon, 2012). Accounting for about 13% of the world’s grasslands, China’s natural grasslands are an important part of the Eurasian grasslands. According to relevant studies, nearly 90% of the grasslands in northern China are degrading to varying degrees (Nan, 2005, Zhou et al., 2017), intensifying the situation of grassland degradation in China. The evaluation of grassland degradation can help solve the problem of continuously intensifying grassland degradation (He et al., 2015), and it provides important information that can further reveal the operation mechanism of grassland ecosystems (Gang et al., 2014, Liu et al., 2018). As an important ecological barrier in northern China, many studies have been conducted on the grassland degradation in the Inner Mongolia Grassland (Li et al., 2017, Wen et al., 2018). For example, Sun et al. discussed grassland degradation and its driving forces based on long time-series remote sensing data in Xilin Gol League of Inner Mongolia (Sun et al., 2017).
The development and application of remote sensing technology makes it possible to monitor grassland degradation on a large scale. Many studies have been carried out worldwide using remote sensing technology (Fassnacht et al., 2015, Kong et al., 2019, Shen et al., 2019). The prerequisite for performing remote sensing monitoring is to establish grassland degradation evaluation indexes, among which the currently extensively-used indexes include the net primary production (NPP), vegetation coverage (VC), and above ground biomass (AGB) (Sun et al., 2017, Zhang et al., 2019). Monitoring grassland degradation using remote sensing is mainly achieved by establishing a relationship between the vegetation indexes and grassland degradation evaluation indexes (Li and Liu, 2007, Xu et al., 2020). When constructing the relationship between vegetation indexes and evaluation indexes based on multispectral remote sensing images, it is impossible to obtain information reflecting the structure of grassland ecosystems due to the low spectral resolution of multispectral images, and the vegetation indexes are susceptible to time and space scales. Hence, multispectral remote sensing has certain limitations in the remote sensing monitoring of grassland degradation.
The development of hyperspectral remote sensing has effectively promoted grassland degradation monitoring (Wiesmair et al., 2016, Cao et al., 2018, Obermeier et al., 2019). Hyperspectral remote sensing offers a high spectral resolution and a large amount of data, so hyperspectral data can identify grassland vegetation types and monitor grassland vegetation composition and growth status in a refined way (Cochrane, 2000, Wang et al., 2010, Li, 2019). Many foreign and domestic studies have been carried out on how to associate spectral characteristics with species composition and vegetation status (Susan, 2010). In the Kazbegi district of the Republic of Georgia, Magiera et al. tested the feasibility of identifying the vegetation species composition based on the canopy reflectance (Magiera et al., 2013). Using the vegetation species composition, Han et al. assessed the status and characteristics of grassland degradation in the northeastern region of Inner Mongolia in northern China (Han et al., 2018). Liu et al. found the grassland structural characteristics can reflect the growth status of vegetation; there are differences between the spectral characteristics of different vegetation components, so the grassland degradation status can be monitored for the spectral characteristics data of the vegetation (Liu et al., 2015). However, hyperspectral images cover limited areas and have many limitations in their application, so they are often used with multispectral data, or the sample areas are set by the image window method (Blanco et al., 2014, Yao, 2015).
The current remote sensing monitoring of grassland degradation is mostly based on the vegetation indexes or NPP. However, since the process of grassland degradation is complicated and has diverse manifestations, this biomass-based monitoring method has specific limitations, making it necessary to analyze grassland degradation from the perspective of the structural characteristics of the vegetation.
In this study, a methodological framework was proposed for monitoring grassland degradation using hyperspectral data from the perspective of the grassland vegetation species composition. First, the characteristic parameters of the measured spectral data were extracted, and PCA was used to screen the representative characteristic parameters as indexes for the spectral dimensionality reduction and un-mixing. The MLPNN was applied to compare the classification accuracy of the mixed spectra of grassland vegetation, and the spectra of the typical plants. Through spectral un-mixing, the typical vegetation species were identified, and then combined with the grassland degradation monitoring standards established by the measured spectra, so as to evaluate the degree of grassland degradation from the angle of vegetation composition. The study shows that this method can enhance the monitoring accuracy of grassland degradation by remote sensing, and provide new ideas and scientific reference for the related research.
Section snippets
Study area
Located in Xilinhot, the central part of the Inner Mongolia, China (43°02′–44°52′ N, 115°18′–117°06′ E), the study area (Fig. 1) covers a total area of 14, 785 km and has a resident population of 267,000. The study area has a semi-arid continental climate and is in the middle temperate zone, which is characterized by cold temperature, frequent drought and strong wind (Hao et al., 2017). The study area features a high terrain in the south and low terrain in the north, with an average altitude of
Spectral analyses and parameter extraction
Pre-processing, like the format conversion, smoothing, and denoising, was performed on the spectral curves of the plants. In addition, to reduce the differences in the background (such as the soil and vegetation environment), different wave bands were selected and the spectral discrimination of each kind of plant was increased to calculate the first-order differential (Fig. 3c) and envelope (Fig. 3d) of the spectral curve. According to the analysis in Fig. 3, the different types of ground
Theoretical basis analyses
Hyperspectral remote sensing obtains the spatial and spectral data of the ground features by detecting the electromagnetic waves reflected by objects with hyperspectral sensors, and it is based on spectrometry (Yang, 2013). Different plants have different spectral characteristics, which is the basic principle for remote sensing to interpret data for vegetation. The reflectivity of a single leaf is mainly affected by its chlorophyll content and structure. In vegetation remote sensing, the
Conclusion
In this study, a method for the remote sensing monitoring of grassland degradation using hyperspectral images was proposed from the perspective of the vegetation species composition. The grassland degradation monitoring standards were established in the study area based on the regional measured data and data measured in the grazing-controlled experimental plot. Characteristic parameters were extracted from the measured spectra, and the measured spectral endmembers were constructed based on the
CRediT authorship contribution statement
Xin Lyu: Conceptualization, Writing - original draft, Writing - review & editing, Methodology, Visualization, Data curation, Software, Validation, Formal analysis. Xiaobing Li: Conceptualization, Writing - original draft, Writing - review & editing, Project administration, Funding acquisition, Resources, Supervision. Dongliang Dang: Conceptualization, Investigation, Writing - original draft, Formal analysis. Huashun Dou: Investigation, Writing - original draft, Methodology. Xiaojing Xuan:
Acknowledgements
This study was funded by the National Key R&D Key projects (grant no. 2016YFC0500502), the National Natural Science Foundation of China (grant no. 31570451), the Program for Changjiang Scholars and Innovative Research Team in University (grant no. IRT_15R06). We thank anonymous reviewers for their constructive comments on the paper.
References (53)
- et al.
Remote sensing of native and invasive species in Hawaiian forests
Remote Sens. Environ.
(2008) - et al.
Challenges of assessing the sustainability of (agro)-pastoral systems
Livest. Sci.
(2011) - et al.
Ecological site classification of semiarid rangelands: synergistic use of Landsat and Hyperion imagery
J. Appl. Earth. Obs.
(2014) - et al.
The importance of grasslands for animal production and other functions: a review on management and methodological progress in the tropics
Animal
(2012) - et al.
Estimating the age and population structure of encroaching shrubs in arid/semiarid grasslands using high spatial resolution remote sensing imagery
Remote Sens. Environ.
(2018) - et al.
A spatially adaptive spectral mixture analysis for mapping subpixel urban impervious surface distribution
Remote Sens. Environ.
(2013) - et al.
Enhancing endmember selection in multiple endmember spectral mixture analysis (MESMA) for urban impervious surface area mapping using spectral angle and spectral distance parameters
Int. J. Appl. Earth. Obs.
(2014) - et al.
Mapping degraded grassland on the Eastern Tibetan Plateau with multi-temporal Landsat 8 data — where do the severely degraded areas occur?
J. Appl. Earth. Obs.
(2015) - et al.
Fine-scale assessment of hay meadow productivity and plant diversity in the European Alps using field spectrometric data
Agr. Ecosyst. Environ.
(2010) - et al.
On variable relations between vegetation patterns and canopy reflectance
Ecol. Inform.
(2011)