1. Introduction
Mangrove has become a sensitive area in response to global climate change due to its coastal distribution, geographical location, and special environment. Extracting vegetation leaf area index from remote sensing observation has always been a difficult and hot issue in the field of quantitative remote sensing [
1]. The existence of mangroves also plays a key role in the balance of the global carbon cycle [
2,
3]. The evaluation of mangrove information indicators at different spatial scales is realized based on multi-source and multi-scale data such as satellite hyperspectral, ground hyperspectral, and chemical analysis, providing technical support for the establishment of a space–earth integrated monitoring platform that can be shared, compatible, and sustainable [
4]. The information of mangrove and non-mangrove areas, different population densities, main species composition, and the impact of human activities can meet the basic requirements of government departments for mangrove mapping and inventory with the combination of multiple remote sensing sensors [
5]. The introduction of remote sensing technology into the monitoring and detection of mangroves focuses on mangrove identification, mangrove health assessment, and mangrove physiological and biochemical information extraction [
4,
6,
7,
8].
The first research direction is the accurate extraction of mangrove distribution assisted by remote sensing technology. The traditional method is based on the idea of threshold; a mangrove index is established using Sentinel-2 image to achieve accurate extraction of submerged mangroves in view of the fact that mangroves will be flooded by tidal water, rainy climate, and other adverse factors [
9,
10]. Additionally, the maximum likelihood classification technology of remote sensing images is introduced to accurately distinguish mangroves [
11,
12,
13]. With the improvement in data resolution, the satellite data with high spatial resolution can accurately extract the boundary information of mangrove from the texture under the action of a neural network algorithm [
14,
15]. Additionally, then, a quantitative classification model of mangrove ecosystem degradation was developed using satellite observation data [
16,
17,
18]. The further application direction is that the carbon storage of mangroves can be roughly estimated using the spatial distribution information calculated by Landsat8, Worldview-2, and ASTER data [
19,
20,
21]. Then, a mangrove extraction model was established to effectively capture the spatial distribution of mangroves from sparse to dense, with different forms based on the spectral index analysis [
22,
23]. Remote sensing technology can collect, process, and image the reflected or radiated electromagnetic wave information to detect and identify various objects on the ground, thus accurately extracting the distribution of mangroves. Optical remote sensing images taken in different bands can help to distinguish the distribution range of mangroves, and radar remote sensing technology can also obtain data with different reflectivity because of the high water content of mangroves, which makes them easier to detect.
Secondly, mangrove health assessment based on multi-source remote sensing data has also been extensively explored. Remote sensing methods have been proven to be effective in mapping mangrove species, estimating their biomass, carbon storage, and assessing the range change [
24]. Combined with satellite data and ground survey data, it is proven that mangroves have good economic and social value [
25]. The precise mangrove map generated by using Sentinel-1 and Sentinel-2 images, combined with Google Earth Engine (GEE), provides a new technology for the evaluation of mangrove ecosystem functions [
26,
27,
28,
29]. The time series data of satellite sensors have become a necessary means to quantify the changes of mangrove cover at the regional and global levels in order to understand the changes of mangrove growth with time [
30,
31]. The remote sensing monitoring data in the past 40 years show that the mangrove area changes significantly from year to year in the Guangdong province. Natural factors such as temperature, sea level rise, extreme weather events, and coastline length have a macro impact on the distribution of mangroves [
32,
33]. The whole process of mangrove death caused by road construction was studied based on the analysis of optics, synthetic aperture radar, UAV image, and topographic survey data [
34,
35]. The spatial and spectral information obtained through remote sensing technology can reflect the growth status and species composition of mangroves, as well as the impact of habitat changes and other factors on the health of mangroves. Using remote sensing parameters to analyze vegetation cover, relative productivity, leaf area index, and other parameters of mangroves, the health status of the forest can be inferred. Meanwhile, combining multiple sources of information such as ground surveys and meteorological observations can improve the accuracy of mangrove health assessment.
Thirdly, remote sensing image analysis, synthetic aperture radar interferometry, and machine learning algorithms have proven the effectiveness of extracting mangrove species, leaf area, crown height, and stand biomass in the research of quantitative information extraction of mangrove physiology and biochemistry [
36]. The biophysical parameters of mangrove can be extracted in a large area, including height, LAI, stem density, and basal area with the help of ground laboratory data [
37,
38]. The change in the wetland vegetation community at different times can be obtained by spectral analysis of satellite images of remote sensors with different resolutions [
39]. Sentinel series data have reasonable correlation with leaf area index, vegetation coverage, and canopy height. These data can be combined with a machine learning model to predict canopy height [
40,
41]. Remote sensing data have a significant correlation with canopy height, canopy shape, and height changes [
42]. A productivity model based on remote sensing was designed to estimate the light use efficiency (LUE) and primary gross product (GPP) of mangroves in China [
43]. Recursive feature is used to select spectral and texture feature variables of vegetation, and random forest and support vector machine algorithm are used as classifiers. The research shows that the combined use of data and methods is helpful for the estimation of mangrove biomass [
44]. The relative amounts of morphology, forest age, canopy coverage, aboveground biomass, and wood debris were extracted from the time series data of spaceborne optical radar and interferometric radar data [
45,
46]. The best method to interpret the change in mangrove carbon storage using remote sensing data was found through image processing [
47,
48]. Using remote sensing technology to extract quantitative information on the physiological and biochemical characteristics of mangrove forests is a practical and effective method. It can reflect indicators such as the photosynthetic activity, leaf area index, and relative productivity of mangrove forests, and thus infer their productivity and growth rate. It can also demonstrate the spatial distribution of water content in forests, revealing their water use efficiency. By combining remote sensing data with ground observation data, such as meteorological station data and soil moisture monitoring data, more accurate physiological and biochemical parameters of mangrove forests can be obtained, such as net photosynthetic rate and water use efficiency. This approach can largely overcome the limitations of traditional quantitative methods in terms of spatial observation range and time scale, enabling a multi-angle understanding of the spatial distribution and dynamic variations of physiological and biochemical information in forests.
Compared with other forest ecosystems, the characteristics of mangrove ecosystems in some aspects are still weak [
40]. It will be more difficult to implement effective policies and actions for sustainable protection of mangroves in the context of climate change mitigation and adaptation without effective quantitative methods to monitor the biophysical parameters of mangroves [
24]. In this regard, remote sensing is an important tool for monitoring mangroves and determining species and other attributes, and the accurate measurement of species leaf area is crucial for assessing forest growth and health [
4,
48].
However, there is some uncertainty in the application of quantitative inversion of LAI due to the wide and discontinuous band of multispectral remote sensing data [
1]. Hyperspectral data can provide rich and detailed continuous spectral band information [
37,
46]. With the continuous development of hyperspectral remote sensing technology, a large number of researchers began to retrieve LAI, chlorophyll, and other plant physiological parameters based on hyperspectral remote sensing methods [
40,
47]. Therefore, the estimation of canopy species abundance based on hyperspectral data and LAI remote sensing retrieval for mangrove communities are of great significance in future forest ecosystem monitoring or research [
26,
32].
To explore the response mechanism of mangrove leaf area index and main nutrient content in the overall framework, firstly, multi-source data acquisition and pre-processing are carried out, in combination with monitoring indicators such as leaf area index, mangrove canopy leaf water content, chlorophyll a, chlorophyll b, total nitrogen, total phosphorus, total potassium, etc., to analyze the mangrove canopy spectral response characteristics of different species composition and abundance in the study area, and to construct the extraction method of mangrove canopy species end elements; secondly, the characteristic bands of mangrove evaluation indicators are screened, and the hyperspectral inversion of mangrove monitoring indicators is carried out by using various linear and nonlinear methods, and the best model is selected; finally, the mapping of hyperspectral retrieval results of mangrove canopy indicators was completed. The research results provide basic data and technical support for mangrove ecological remote sensing monitoring [
12,
28,
47].
4. Discussion
Mangrove is a special vegetation type that grows in the upper part of the intertidal zone of tropical and subtropical coasts, and is an ecological key area in the transition between land and sea, with unique hydrological characteristics, biogeochemical, and ecological functions [
4,
9,
56]. Hainan Province is one of the areas with the widest distribution of mangroves and the richest biodiversity in China [
17,
41]. LAI represents the density of vegetation leaves, and is the key factor affecting the photosynthetic effective radiation capacity of the canopy in the carbon cycle. Its level directly affects the strength of the photosynthetic capacity, and has an important impact on the global carbon cycle and vegetation growth and development [
37,
46,
47]. Accurately grasping the mangrove LAI represents the basic work undertaken to evaluate the vegetation growth status in the land and sea transition zone. The relevant variables of mangrove LAI can be used as the health indicators of the forest ecosystem. The ability of LAI to characterize the canopy structure is crucial to understand the LAI in assessing the health status, predicting future growth, and mangrove production.
Satellite, airborne, and ground remote sensing sensors are used to receive the reflected signals of ground objects. Different mangrove tree species have different absorption and reflection characteristics of electromagnetic waves of different wavelengths, forming the characteristic spectrum of mangrove reflectivity changing with wavelength [
19]. The mangrove spectrum has a fingerprint effect on mangrove classification and target recognition, which is a bridge connecting remote sensing theory and remote sensing application. Spectral data sets of spectra and characteristic parameters that can cover a variety of typical targets are formed by collecting, processing, and analyzing the measured spectra of typical mangroves [
20].
Mangroves are affected by the special growth environment, and the traditional survey technology is faced with many challenges [
10]. Hyperspectral technology can obtain the nutrient content of mangrove plants from both mechanism and statistics. Satellite data have wide coverage, strong timeliness, and high spectral resolution, but limited spatial resolution, so are very suitable for large-scale regional surveys, covering regional, national, and even global scales [
35]. The spatial and spectral resolution of airborne hyperspectral are high, but the data acquisition conditions and costs are high, so it is suitable for a large-scale survey in key areas [
10]. The UAV hyperspectral data acquisition method is very flexible, but the data acquisition efficiency is low, so it is suitable for a small-scale survey or field test [
50]. The ground hyperspectral has the highest spectral resolution, but there is no image information, and the data are in the form of scattered points, which are suitable for data modeling, ground experiments, and verification [
44]. Through the cooperation of satellite hyperspectral remote sensing and ground hyperspectral data, this research obtains the data of photoelectric detection data to achieve more accurate quantitative remote sensing [
38,
45,
53].
This paper focuses on two methods in order to realize the effective assimilation of satellite data and ground data [
46,
47]. The first one is to determine the quantitative relationship between mangrove monitoring indicators and characteristic bands to achieve the extraction of nutrient content. Through processing the spectral data, including itself and its 23 kinds of transformation data, three band combination algorithms of band difference, band ratio, and band difference and ratio are tested to form a controllable machine learning model package [
12,
26,
39]. They are used to conduct large-scale training and learning on mangrove component content and hyperspectral data, and extract information from statistical significance. Although this method cannot explain the basis of feature band selection from the mechanism, it is effective in a certain region and a certain period of time [
11,
44].
The research reveals that, firstly, different spectral transformation and band combination models are needed for the extraction of different components of mangrove [
10,
20,
22,
47]. The processed data including spectral differentiation and de-enveloping can significantly improve the regression accuracy of the model. This modeling method is very accurate since spectral data are measured on the ground and in situ data are obtained; secondly, the numerical comparison of the nature reserve in different regions shows that the growth status of mangroves in the three regions is consistent, which shows the local importance to the nature reserve [
23,
24]. According to the satellite monitoring results, it is not found that human activities have affected the growth of mangroves in different regions. Thirdly, the study confirmed that LAI, as a geometric index of leaves, has no decisive effect on the nutrient composition of mangroves. Through covariance and correlation analysis, although LAI has little relationship with the content of nutrients, there is a significant correlation between the six nutrients [
37,
40,
46,
47]. This conclusion can not only guide the scientific evaluation of mangrove growth quality, but can also control the potential risk sources to guide the official work.
Recently, hyperspectral sensors based on satellites, aircraft, unmanned aerial vehicles, and the ground have emerged endlessly, and have been increasingly introduced into mangrove ecological assessment [
12,
21,
42,
44,
47]. The traditional sampling and analysis technology has been unable to meet the needs of digital applications due to the special growth environment of mangroves, and the intervention of new technologies is urgently needed [
14]. In the future, the research focus of hyperspectral technology in this field should be to form the mangrove basic spectral database, realize the scientific modeling of hyperspectral data, and solve the current regional and temporal constraints. Spectral technology is one of the important development directions in this field with the accumulation of data. As this region has been designated as a nature reserve by China, all the mangroves here are in very good condition. Among the mangrove forests in China, the mangroves of the study area are in one of the largest contiguous areas, with the most diverse tree species, best forest quality, and richest biodiversity, and the area has been listed as an internationally important wetland site. This indicates that the overall health level of the mangroves in Dongzhai Port is relatively high. Our field investigations have also confirmed this, with all the mangroves being classified as being in normal condition.