Critical Analysis of Forest Degradation in the Southern Eastern Ghats of India: Comparison of Satellite Imagery and Soil Quality Index

India has one of the largest assemblages of tropical biodiversity, with its unique floristic composition of endemic species. However, current forest cover assessment is performed via satellite-based forest surveys, which have many limitations. The present study, which was performed in the Eastern Ghats, analysed the satellite-based inventory provided by forest surveys and inferred from the results that this process no longer provides adequate information for quantifying forest degradation in an empirical manner. The study analysed 21 soil properties and generated a forest soil quality index of the Eastern Ghats, using principal component analysis. Using matrix modules and geospatial technology, we compared the forest degradation status calculated from satellite-based forest surveys with the degradation status calculated from the forest soil quality index. The Forest Survey of India classified about 1.8% of the Eastern Ghats’ total area as degraded forests and the remainder (98.2%) as open, dense, and very dense forests, whereas the soil quality index results found that about 42.4% of the total area is degraded, with the remainder (57.6%) being non-degraded. Our ground truth verification analyses indicate that the forest soil quality index along with the forest cover density data from the Forest Survey of India are ideal tools for evaluating forest degradation.


Introduction
India, a country with rich biodiversity, has different forest types in different climatic zones, each with a unique floristic composition [1,2]. The forests are mainly distributed in the Himalayas, Western Ghats, Eastern Ghats (EG), and Vindhya ranges. Indian forests have historically experienced a wide variety of management practices because of the differing policies of diverse peoples and governances [2,3]. In particular, the protected forest reserves of the southern EG, an area in Tamil Nadu [4,5], had experienced considerable need-based forestry practices (e.g. felling of timber for construction, fuel, or the creation of grazing lands and real estate) for more than 150 years until 1980 [6][7][8][9][10]. As a result, the floristic composition of the forests in the EG underwent massive changes [10,11], including the emergence of secondary and tertiary vegetation. Furthermore, severe, negative effects on soil health were also reported [8]. Therefore, effective monitoring and inventory strategies are crucial for the continued conservation of these forests.
Currently, the Forest Survey of India (FSI) periodically monitors forest status based on canopy density, which is obtained via satellite data with ground truth verification. Canopy density is classified into four categories: (i) very dense forest (VDF) with >70% canopy density, (ii) moderately dense forest with 40%-70%, canopy density, (iii) open forest with 10%-40% canopy density, and (iv) degraded forest with <10% canopy density [12]. This classification is based on a normalised difference vegetation index (NDVI), which does not distinguish between crown density and ground density [13]. However, research has shown that the NDVI is not detailed enough to provide reliable data and may underestimate the degradation status [14][15][16][17]. Menon and Bawa [18], who estimated the deforestation rates in India and identified large disparities in the forest definitions, also highlighted the need for improved remote sensing and ground truth verification data to discriminate native and plantation forests. Moreover Aziz et al. [19] clearly inferred that the NDVI could not capture the difference between natural forests and rubber plantations in Malaysian forests.
Soil is a vital component of forest ecosystems and is responsible for processes that support biomass production and carbon sequestration [20]. Forest soil health is sensitive to numerous natural and anthropogenic factors, and a deterioration of soil quality typically leads to a deterioration of site quality [21]. Thus, it is important to establish appropriate soil quality indices for measuring the various properties of soil. A soil quality index is defined as the minimum set of parameters that, when interrelated, provides numerical data on the capacity of a soil to carry out one or more functions. A soil quality indicator is a measurable property that influences the capacity of a soil to carry out a given function [22]. A soil quality index, to some extent, should be use-dependent, so that it can be applicable on a larger scale. SQI will help us to select a minimum set of indicators that can address the overall soil quality of a particular region [23]. Foresters have always relied on their knowledge of the chemical, physical, and biological properties of soils to assess the capacity of a site to support productive forests [24]. Recently, the need to assess soil properties has expanded because of growing public interest in determining the consequences of management practices on the quality of soil relative to the sustainability of forest ecosystem functions in addition to plant productivity [25]. The concept of soil quality includes the assessment of soil properties and processes as they relate to the soil's ability to function effectively as a component of a healthy ecosystem [26]. The specific functions and subsequent values provided by forest ecosystems are variable and rely on numerous soil properties and processes [27]. Thus, an understanding of soil properties and soil biota is imperative to the accurate classification of the forest degradation level [8].
Forest soil quality assessment is well established in several developed countries, such as the United States [28], New Zealand [29], and numerous European nations [30], including the United Kingdom [31]. In India, because the majority of soil quality studies are performed in agricultural and horticultural sectors [32,33], very little data are available on forest soil quality and biota [8]. Therefore, the aim of this study was to identify the degradation status of forests in the EG by performing a detailed soil quality assessment, and then using geospatial technology to compare the soil quality results with those obtained from FSI forest density classes.

Study area
The EG (between 10°00 0 00@-13°00 0 00@N and 77°50 0 00@-79°10 0 00@E) is a discontinuous mountain range running almost parallel to the east coast of India and constitutes the watersheds of many rivers. In Tamil Nadu, the EG start from the Jawadi Hills and extend up to the Alagar Hills (Fig 1), comprising 13 major hills in total (Jawadi, Elagiri, Shevaroy, Chitteri, Kalrayan, Bodamalai, Kolli, Pachaimalai, Semmalai, Aiyalur, Karandamalai, Sirumalai, and Alagar). The hills of the EG in Tamil Nadu range from 70 km 2 to 1860 km 2 in area, and the altitude in this region ranges from 180 m above mean sea level (MSL) to 1650 m above MSL. The mean minimum and maximum temperatures are 17°C and 33°C, respectively, and the average annual rainfall is 800-1600 mm. The major soil types are entisols, inceptisols, and alfisols. Geologically, the EG consist of charnockite with minor bands of pyroxin granulate and magnetite quartzite. The tributaries of many perennial rivers like the Pennaiyar, Palar, Vellar, Cauvery, and Vaigai originate from these hills. These hills are still rich in biodiversity, despite heavy exploitation, and they contain more than 960 species of angiosperm and gymnosperm. There are many endemic, endangered, and medicinally important species distributed within these hills [34,35].
The forests are mostly degraded due to the repeated felling of trees in the past; presently, three major types of forests exist, with different compositions and canopy densities, as explained in detail below.  [50] as a data reduction technique to identify the appropriate indicators without losing the vital information [51,52]. The principle components (PCs) are linear combinations of variables that account for the maximum variance in a data set and are more accurate descriptors of the data than the variables themselves. For this study, PCs with eigenvalues !1.0 (>10% of the total variance explained) was taken into consideration for indexing. Highly weighed variables (eigenvectors ! ± 0.7) within each PC were considered significant and retained for the FSQI [24,53].
Normalisation of the indicators. As the selected soil parameters were measured in heterogeneous units, they were standardised to unit-less values between 0 and 1 [32]. Depending on whether a higher value was constructive or detrimental to soil quality, indicators were arranged in ascending or descending order [32,33]. The resultant outputs were called Indicator Scores. These normalised scores were calculated using the following two formulas. Eq 1 was used for indicators that tended to increase soil quality as the values increased: Eq 2 was used for indicators that tended to decrease soil quality as the values increased: where Si max is the maximum value of the soil indicator variable, Si is the actual observed value, and Si min is the minimum value of the soil indicator variable. Available Nitrogen (Ava N) Alkaline permanganate method [44] Available Phosphorus (Ava P) Olsen's extractant (0.5 M NaHCO3) at pH 8.5 [45] Available Potassium (Ava K) Neutral N NH4OAc method [46] Microbial Biomass Carbon (MBC) Fumigation extraction method [47] Microbial Biomass Nitrogen (MBN) Fumigation extraction method [47] Total and Functional Microbes Pour plate method [48] doi:10.1371/journal.pone.0147541.t001 Calculating the PCA-FSQI. After normalisation, the MDS variable of each observed value was weighed using the PCA-derived outcome. The weighing factor was calculated via dividing the percentage of each selected MDS variable by the total percentage of variation in all PCs with eigenvectors >1. The scored indicator for each observation was calculated with the following equation: where Wi is the PCA-derived weighed factor and Si is the normalised scores of the indicator i.
Higher index scores indicate better soil quality.

Spatial data analysis
Spatial autocorrelation was performed using a semi-variogram that considered all sampling points to test the spatial autocorrelation between points. The global Moran's I test was used to estimate the overall degree of spatial autocorrelation and the analysis returns a single value, which is applicable to the entire study area [54].

FSQI threshold and spatial mapping
To fix the FSQI threshold, the mean and 95% confidence interval (CI) of the 408 samples were calculated [55,56]: The FSQI value obtained for each sample was interpolated using the inverse distance weighted method in Arc GIS 9.3. The interpolated values were grouped into two categories: degraded (FSQI < threshold value) and non-degraded (FSQI > threshold value).

Matrix analysis and ground truth verification
Matrix analysis is a spatial process where two raster data can be verified pixel-by-pixel for its agreement. Each pixel in the resultant image will carry information from both raster. Through this method the FSI forest cover density map and the soil quality index maps were analyzed. Each pixel of resultant image showed the forest cover density and its corresponding SQI index value.
The matrix analysis between the FSQI and the FCD of FSI was performed using ERDAS IMAGINE 2011 software (Erdas Imagine, Hexgon geospatial, ink GA,USA) [57] Ground truth verification was also performed for both the FCD and FSQI to verify whether the classified areas conformed to the field data across all forest density classes.  6.1-6.7 6.1-6.9 6.1-6.7 6.5-6.71 6.3-6.8 6.1-6.7 6.1-6.5 6.0-6.4 6.2-6.7 6.2-6.7 6.3-6.9 6.0-6.4  In total there were 8 classes and for each class 25 points were randomly created and were verified on the ground using GPS. Of the total 200 quadrates, each location the vegetation data on trees and shrubs were enumerated through quadrate of 20X 20 m size covering 13 hills.   Fig 2). Additionally, most of the areas occupied by scrub forests were classified as dense or moderately dense, indicating that the FSI data may be misleading and may fail to provide an accurate description of forest degradation.

FSQI
Soil physico-chemical and biological analysis. The physico-chemical and biological properties of the 408 soil samples are provided in Table 3. While most of the measured properties exhibited high variation, the soil pH was uniformly low. Across the 13 hills, Aiyalur displayed the lowest values for soil organic carbon (SOC), total nitrogen (TN), microbial biomass carbon (MBC), microbial biomass nitrogen (MBN), total bacteria (TB), and functional microbes. In contrast, the highest values of the measured properties were found in the Chitteri samples, which also exhibited greater variation in biological properties. Our results support previous findings that indicate that the variability in these factors is due to changes in climate, season, geographical location, and anthropogenic interference [24,[58][59][60][61][62].

Spatial data analysis
A fitted semi-variogram of the FSQI in the study area had a nugget effect of 0.55 and a total sill of 0.74. The high nugget-sill ratio of 74% reveals that strong local scale variation prevails. The spatial autocorrelation that was performed based on the global Moran's I of the FSQI was 0.216, which was low but still statistically significant (p = 0.000014). This indicates that spatial clusters of the FSQI may be found across the study area with similar values (S1 Fig). MDS. The results of our PCA on the 22 soil parameters (Table 3) revealed that the first four PCs explained approximately 80.45% of the soil variability. Further, the first two PCs accounted for approximately 57.17% of the variance, strongly indicating that the highly weighed variables in these two components can be used to assess the soil quality of this area.
Under PC1, these variables were SOC, MBC, TB, and TN; under PC2, NFB was the highest weighed value. These five properties indicate the level of soil disturbance since biological properties react relatively quickly to small changes in the soil environment and it has been broadly reported that any alteration in soil management and land use is reflected through soil properties [63][64]. Thus, this study has considered these five biological properties as the MDS for identifying the soil quality status in all 13 hills (Table 4, As previously stated, low FSQI values indicate degraded soil and high values indicate nondegraded soil. Soil from Chitteri Hill exhibited the highest FSQI (0.69), followed by soil from Kalrayan Hill (Table 5). Aiyalur and Bodamalai possessed soils with the lowest FSQI values: 0.32 and 0.36, respectively (Fig 3). The mean FSQI of all EG sites in Tamil Nadu was only 0.44, very close to the threshold value. Based on the FSQI, approximately 42.4% of the study area consists of degraded forests and 57.6% consists of non-degraded forests.
However, ground truth verification revealed a ±13% deviation from the FSQI-classified status, due to variation in vegetation within a particular forest density class. For instance, the presence of riparian flora along streams in the open and degraded forests meant they were apparently classified as dense forests.
Based on the FSQI, the highest soil degradation (>96.1%) was found in Aiyalur forests, followed by Bodamalai and Karandamalai (81.9%) and Alagar (69%) forests (Table 5). Furthermore, the lowest soil degradation occurred in Chitteri, with 11.2% of the area classified under the degraded category. In addition, the soil quality of Sirumalai, Kalrayan, and Elagiri was found to be less affected by forest degradation, with 34.3%, 35.8%, and 37.5% of their total areas classified as degraded, respectively ( Table 5).
The matrix analysis of the FCD and FSQI (Table 6) revealed more details about the degradation status. Forests in all four FCD categories have both degraded and non-degraded soil. In the degraded forest category, 74.2% possesses degraded soil, and in the open forest cover category, approximately 52% possesses degraded soil. In the moderately dense and very dense categories, degraded soil occurs in 38.9% and 26.7% of the forest areas, respectively. Across all forests, only 154474.2 ha (40.9%) were classified as non-degraded by both the FCD and FSQI (Fig 4). Based on previous reports of the EG, soil quality degenerated wherever severe anthropogenic exploitation slowly modified primary forest vegetation to secondary vegetation [8,10]. Such forests tend to consist of patchily distributed, low scrub that degenerates into spare bushes with occasional, stunted tree growth predominantly occupied by L. camara bushes that suppress the natural regeneration of other species [3,13,[65][66][67][68].   Although there is some correlation between the forest density and the percentage of degradation estimated by the FSI, the present study clearly addresses the inconsistencies between the density classes of the FSI and FSQI (Table 6). Forests classified by the FSI as very dense forests, moderately dense forests, and open forests contain soils below the FSQI threshold, while high FSQI values occurred in forests classified as degraded by the FSI. Overall, an imprecise relationship exists between the FSI density classes and the FSQI. We conclude that the NDVI-based forest cover assessment cannot comprehensively describe the real degradation status of forests.
The species composition of various slope categories in the EG (Table 7) reveals two distinct patterns that correspond to the FSQI rather than the FCD. Wherever the FSQI value is below the threshold, thorny scrubs instead of trees generally dominate the floristic composition. Moreover, when the FSQI value is higher than the threshold, dense tree growth dominates, with a very limited occurrence of herbs and shrubs (Table 7). A large contributor to these patterns is the invasive species L. camara [65][66][67][68], which was extremely prominent in areas with low FSQI values and appeared to prevent further proliferation of native species. Other factors that aggravate soil degradation and hinder plant growth include anthropogenic activity, environmental factors such as erratic rainfall and extreme heat during summer, as well as geological features like steep slopes [10].
In summary, our study lends further support to previous findings that demonstrate the impreciseness of NDVI-based forest cover assessments, which fail to differentiate between natural forests and plantations [69]. Our results strongly suggest that satellite-based assessment of canopy cover gives an incomplete idea of forest status. However, we also found that the FSQI can be an effective tool for categorising forest degradation with greater accuracy.

Conclusions
This study pioneers the use of the FSQI for estimating forest degradation in India, which currently depends on satellite-based canopy cover data to assess forest status. Using the FSQI, degradation was found in all FSI forest cover density classifications. Combining the FSQI with the FSI density classes resulted in a more detailed picture of forest degradation. Hence, the FSQI is an ideal tool for studying the degradation status distributed across different forest types and density classes. The present study reveals that the FSQI categorised approximately 42.4% of the total forests examined as degraded, whereas the FCD only categorised approximately 1.8% of the forests as degraded. Therefore, relying on satellite data alone may no longer provide an accurate representation of forest status, given the increase in environments that may appear as undisturbed forest without a more sensitive investigation. We suggest that the FSQI can be used as supplemental data along with the FSI forest cover density to efficiently categorise forest degradation and to address this continuing environmental issue, especially in developing countries. Management of degraded forests involves meticulous reforestation work to restore the forests, which could be achieved only by prioritising the area based on the FSQI.