1. Introduction
Mangroves are salt tolerant trees distributed along intertidal coasts of tropical and subtropical regions [
1,
2]. These forests provide a variety of environmental functions such as a nursery habitat for both terrestrial and marine fauna; they act as a natural barrier against tropical storms and hurricanes; and are an important resource for local communities [
3]. Mangrove forests are also extremely important contributors to global organic carbon dynamics [
4,
5,
6], to primary productivity [
7,
8], for nutrient recycling [
5,
9], and can help to mitigate climate change impacts [
10]. Despite their ecological relevance, mangrove forests are under considerable degradation due to anthropogenic perturbations including aquaculture expansion [
11,
12]. In fact, Duke
et al. [
13] have suggested that the services offered by mangrove ecosystems could become ecologically insignificant within the next 100 years if the current deforestation rates are maintained. Consequently, many techniques, including remote sensing, have been investigated in order to properly classify and monitor these forested wetlands (e.g., [
14,
15,
16,
17,
18]).
A considerable number of the remote sensing assessments of mangroves have used optical spaceborne data (e.g., [
19,
20,
21,
22,
23]). Although the use of these multispectral data can provide basic mangrove forest mapping (e.g., Landsat Thematic Mapper (TM) and Satellite Pour l’Observation de la Terre (SPOT)), the coarse spectral resolution of these platforms limits the extent of biophysical data that can be extracted from these forests. Thus, there has been an increasing interest in the use of spectroscopy data, currently using
in situ instruments, as an alternative to multispectral data for accurate quantification of mangrove biophysical variables such as leaf pigment contents and nitrogen. For example, Vaiphasa
et al. [
24] used reflectance of 16 mangrove species in order to test species separability. Their results indicated that 16 mangrove species were statistically different at most spectral locations. Wang and Sousa [
25] conducted a laboratory study measuring the reflectance of leaves collected from a forest canopy dominated by
Avicennia germinans (
A. germinans),
Laguncularia racemosa (
L. racemosa), and
Rhizophora mangle (
R. mangle) located along the Caribbean coast of Panama. The results from this study demonstrated that wavebands at 780, 790, 800, 1480, 1530, and 1550 nm were identified as the most useful bands for mangrove species classification. Panigrahy
et al. [
26] examined the leaf reflectance characteristics of four tropical mangrove species (
Avicennia alba,
Avicennia marina,
Rhizophora mucronata, and
Sonneratia caseolaris) common to India showing unique spectral signatures four all species using information collected from the red, near infra-red, and middle infra-red regions of the spectrum. Zhang
et al. [
27] performed a laboratory study assessing relationships between pigment content (chlorophyll-a, chlorophyll-b, and total carotenoids) and leaf reflectance (350–2500 nm) from two subtropical mangrove species (
A. germinans and
R. mangle). The results from this work indicated that traditional vegetation indices do not necessarily improve the ability to predict pigment content. Contrary, wavebands at the red-edge position were found to be the best predictors of the leaf pigment contents. Zhang
et al. [
28] applied the use of spectral response curves for estimating nitrogen leaf concentrations in two mangrove species (
A. germinans and
R. mangle). Their results confirmed that artificial neural networks could be used to assess mangrove health based on the amount of leaf nitrogen content. Flores-de-Santiago
et al. [
29] studied the influence of seasonality in estimating chlorophyll-a content from 35 vegetation indices using a pooled sample of three mangrove species (
A. germinans,
L. racemosa, and
R. mangle). Their results indicated that vegetation indices using information from the red-edge wavebands are best at predicting leaf chlorophyll-a content. Zhang
et al. [
30] conducted a laboratory study in order to select optimal wavebands for species discrimination based on three mangrove species (
A. germinans,
L. racemosa, and
R. mangle). Their results indicated that wavebands at 520, 560, 650, 710, 760, 2100, and 2230 are the most appropriate for mangrove classification.
Of particular interest in remote sensing is leaf reflectance, which can provide unique information regarding the actual physiological state of plants [
31,
32]. Specifically, spectroscopy assessments of pigments, such as chlorophyll-a (chl-a), chlorophyll-b (chl-b), and total carotenoids (tcar), may provide more detailed, accurate, and quick results for environmental monitoring of mangroves using spaceborne platforms. It is well known that the aforementioned pigments present different absorption patterns with maximum reflectance at unique wavelengths [
33,
34]. Thus, potential spectroscopy discrimination could be achieved for separating mangrove conditions (
i.e., degraded) according to pigment contents. Moreover, prior to any application of multispectral or hyperspectral satellite imagery in mapping mangroves, it is necessary to test the feasibility of applying laboratory-based spectroscopy data for identifying mangrove species under various health conditions (e.g., stressed, healthy).
The vast majority of spectroscopy studies have focused on the association between spectral reflectance and leaf chl a or chl b contents (e.g., [
35,
36,
37,
38,
39,
40]) but some have also assessed tcar content (e.g., [
41,
42,
43,
44]), and a few have considered the chlorophyll a/b ratio (chl a/b) (e.g., [
45,
46,
47]). Regarding mangroves, the results of previous studies indicate that each mangrove species presents a unique characteristic with regards to their reflectance [
25,
29,
30]. Therefore, the purpose of this investigation was to examine relationships between spectroscopy variability (450–1000 nm) and leaf pigment contents (chl-a, chl-b, tcar, and chl a/b) in the three most common species found in the Americas (
A. germinans,
L. racemosa, and
R. mangle). Given that a previous work [
29] has shown a shift in leaf pigment contents for these species, we further consider the potential impact of two contrasting seasons on these relationships.
4. Discussion
There is a constant need to improve and develop new techniques to assess biochemical mangrove characteristics for environmental purposes. This study assessed seasonal relationships between leaf reflectance and pigment contents, among the three species and two conditions (stress and healthy) for a semi-arid region of Mexico. The results from our investigation indicated that the unique spectral signatures of the six mangrove classes could, to some degree, be explained in terms of leaf pigment variability. Moreover, the results would indicate that fresh water availability and proximity to the tidal channel greatly influences the mangrove leaf chlorophyll content. In fact, mangroves are generally considered a very tolerant group of trees because they can accommodate very stressful conditions encountered along tropical and subtropical coastlines. However, mangroves located in subtropical regions (
i.e., arid or semi-arid environment) are very sensitive to seasonal variability in precipitation, hydroperiod (
i.e., duration of flooding), and light irradiance levels [
4,
48]. Hence, it has been suggested that higher solar irradiance and hypersaline conditions could affect mangrove growth in subtropical regions by affecting metabolic processes such as the reduction in the stomatal conductance [
10]. Given the sub-optimal hydrological conditions (salinity of 35–80 psu) present for even the healthier mangroves of this semi-arid subtropical forest it is not surprising to find much lower overall chlorophyll content as compared to mangroves located in more tropical latitudes [
57] where higher freshwater inputs and less severe seasonality are present.
The importance of the green and red-edge wavelengths is not a novel contribution to the reflectance assessments of tropical leaf pigment contents. For instance, previous studies have found high sensitivity to pigment content using reflectance at wavelengths around 550 nm and 705 nm in a variety of plants [
58,
59,
60,
61,
62,
63]. However, the differences among the three species regarding pigment contents linked to the high variability in the spectroscopy responses among our mangroves, suggest that the three species respond differently due to seasonal and tree health condition. Thus, in our study chl-a, chl-b, and tcar presented the highest correlations at similar wavelengths, even though the three pigments are chemically quantified using different absorbance equations [
54].
Based on the correlograms from
Figure 5 and
Figure 6, the specific reflectance at the 550–570 nm and 705–715 nm could be suitable for remote sensing estimations of chl-a leaf contents in some species of mangroves (e.g., red and black). As a result, reflectance-based algorithms such as Vog1 and REIP for quantifying mangrove leaf chl-a and chl-b contents in semi-arid regions should use the wavebands between 550–570 nm (green channel) and 705–715 nm (red-edge) as optimal zones. Conversely, wavelengths at the 650–680 nm (red channel), and 732–1000 nm (near-infrared) ranges should be used as insensitive areas. Contrary to [
26], we did not find high correlation values within the near-infrared regions (800–1000 nm). However, they tested different species of mangroves (
i.e.,
Avicennia alba,
Avicennia marina, and
Rhizophora mucronata). Consequently, these results indicate that each mangrove species and condition should be considered when dealing with spectroscopy assessments of leaf pigment contents. Additionally, it is also apparent that seasonality can play a key role in the success of such endeavors as shown for these semi-arid mangroves.
While the methods used in our study may not provide the most accurate results of leaf pigment contents for all mangrove species, it does offer a convenient approach when considering seasonal spectroscopy variability for the three dominant mangrove species of the Americas. We believe the use of mangrove leaves from different conditions (i.e., fringe–basin) provides a more realistic assessment of the relationships between reflectance and mangrove leaf pigment contents as the samples were collected within close proximity from one another and yet shown such high variability.
Among the three mangrove species analyzed in our study,
L. racemosa was the most difficult to assess regarding the relationships between the reflectance and the leaf pigment contents. Compared to the red and black mangroves under stress condition, the white mangrove showed minimal correlations throughout the wavelengths with the leaf chlorophylls content (
i.e., chl-a and chl-b). This pattern was less evident in the healthy classes, but the white mangrove still showed the lowest correlations with both chlorophylls during the dry and rainy seasons. Unfortunately, there are just a few works related to mangrove reflectance at the leaf level (e.g., [
24,
25,
26,
27,
28,
29,
30,
32]), and at the canopy level (e.g., [
64,
65,
66,
67]) to derive general assumptions of these relationships which can be a challenge give the variety of mangrove species and conditions found around the world. Previous works, for example, used the same three mangrove species in the tropical coast of Panama for reflectance separability among such species [
25]. A more recent work carried out by [
26] assessed the separability at leaf level, but in tropical mangrove species in the Indian Ocean (
i.e.,
Avicennia alba,
Avicennia marina,
Rhizophora mucronata, and
Sonneratia caseolaris). The closest work in semi-arid mangroves using spectroscopy data and pigment content was performed by [
27] who used the same three mangrove species for their assessment. However, only the red and black mangroves were used for the leaf pigment contents determination, but similar results with maximum correlations and the green and the red-edge were found. As mentioned before, further investigation needs to be carried out regarding mangrove biochemistry at the leaf level for the accurate explanation of
L. racemosa response found in our study.
There may be many reasons why the observed discrepancies in relationships among the mangrove species and conditions were observed in this study. For example, the locations of the spectral responses of mangrove leaf pigments are between 400–700 nm [
68], and the organic compounds related to the absorption features in the visible and near-infrared regions are responses to inorganic C-H, N-H, and O-H bonds [
26]. Primary results indicate that the reflectance characteristics of each mangrove are related to the absorption properties of the organic compounds presented within the leaves. According to Snedaker
et al. [
69],
L. racemosa is the only mangrove species in the Americas that has shown to decrease stomatal conductance under high CO
2 levels. Moreover, they indicated that the water use efficiency of
L. racemosa was highly altered under the same conditions. Consequently, it is possible that
L. racemosa might have a competitive disadvantaged among other mangrove species, particularly in semi-arid regions [
2]. For example, it is well known that mangroves from semi-arid regions can grow under salt concentration as high as 80 psu [
2,
70] and that
L. racemosa is the most susceptible to salinity changes, which affects the stomatal conductance of their leaves through variations in salinity, humidity, soil moisture, and temperature. Consequently,
L. racemosa leaf structure becomes unstable by isotope fraction [
71], under adverse salinity levels affecting the chemical composition and biochemical responses.
L. racemosa manages salinity by controlling stomatal aperture, which in turn maintains carbon gain with improved leaf water use efficiency, but at the same time brings a parallel decline in potential nitrogen use effectiveness [
72]. In this sense, under stressful conditions as shown in our study area,
L. racemosa trees could change the biochemical properties of their leaves, affecting the reflectance as potentially seen in our investigation.
The dissipation of the excess of energy as heat, and the presence of tcar is common in stressed plants [
33]. In our study, there was a clear pattern among the healthy mangroves during the rainy season, where the tcar content decreased for all three species. Although there were no seasonal differences in chl-a and chl-b content among the healthy classes of
R. mangle and
A. germinans, the tcar significantly decreased while the chl a/b ratio remained constant. This pattern could be explained using the suggestions by [
48], where fringe mangroves (
i.e., healthy) from semi-arid regions are constantly flooded by local tidal influence avoiding hypersaline conditions and, thus, variability in major pigments, such as chl-a, are not affected by seasonal precipitation. However, the decrease of tcar during the rainy season could be the result of less solar irradiance as compared to the dry season. Changes in pigment composition results in visible and near-infrared alterations and thus the structure and function of the photosynthetic apparatus is adjusted in leaves depending on the response to changes of irradiance [
33]. It has been widely accepted that photosynthetic pigments, mostly chl-a and chl-b, tend to increase with decreasing irradiance facilitating light harvesting in shade-tolerant species [
73]. Contrary, shade-intolerant species such as
L. racemosa [
2,
48] could increase the canopy cover and thus decrease the leaf chl-a content when optimal environmental conditions (
i.e., rainy season) are present such as shown in this investigation.
5. Conclusions
Mangrove forests are one of the most important coastal ecosystems within tropical and subtropical regions, and yet, these forested wetlands are under a constant degradation due to many anthropogenic perturbations. This study aimed to examine the seasonal relationships between reflectance (400–1000 nm) and leaf pigment contents (chl-a, chl-b, tcar, and chl a/b ratio) in three mangrove species (A. germinans, L. racemosa, and R. mangle) under various conditions (stress vs. healthy) and two seasons (dry vs. rainy). The results of this study indicate that stressed mangroves had much higher variability in chl-a content during the dry season as compared to the same trees during the rainy season. Consequently, it is likely that photosynthesis in the stressed mangroves classes is frequently light saturated. Any excess of light would be wasteful and could, in fact, give raise to photoinhibition and other harmful effects. Contrary to the stressed mangroves, the correlation coefficient for the healthy mangroves did not show an evident seasonal change for the chl-a content. The results also indicate that spectrally, mangrove species and health conditions are relatively distinct in various wavelength regions. For instance, the stressed mangroves presented highest correlation with chl-a in the green and red-edge regions (r = 0.8 and 0.9, respectively), while correlation with tcar content did not show a relationship with reflectance during the both seasons. Conversely, relations with accessory pigment such as chl-b were higher in the three mangrove species under healthy condition during the dry season.
Monitoring the condition of mangrove forests using detailed spectroscopy correlations at the leaf level is the first obvious step in obtaining accurate spaceborne monitoring of large mangrove areas at the canopy level using high spatial resolution imagery. However, further studies should include the examination of the potential influence of canopy structure characteristics, such as leaf orientation, leaf morphology, three height, canopy coverage, and background information on the spectral response.
We believe that the present study is the first to include white mangrove spectroscopy correlation with pigment contents for semi-arid regions. Consequently, the identification of locations of high correlation along the electromagnetic spectrum could provide reliable information for accurate quantification of pigments. This could be of utmost importance for primary productivity models, carbon balance, and ecological assessments.