Next Article in Journal
Vision-Based Moving-Target Geolocation Using Dual Unmanned Aerial Vehicles
Next Article in Special Issue
Mangrove Biodiversity Assessment Using UAV Lidar and Hyperspectral Data in China’s Pinglu Canal Estuary
Previous Article in Journal
Quantitative Evaluation of Dust and Black Carbon Column Concentration in the MERRA-2 Reanalysis Dataset Using Satellite-Based Component Retrievals
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 15
Issue 2
10.3390/rs15020387
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Prediction of Plant Diversity Using Multi-Seasonal Remotely Sensed and Geodiversity Data in a Mountainous Area

by
Soroor Rahmanian
1,
Vahid Nasiri
2,
Atiyeh Amindin
3,
Sahar Karami
4,
Sedigheh Maleki
5,6,
Soheila Pouyan
3 and
Stelian Alexandru Borz
1,*
1
Department of Forest Engineering, Forest Management Planning and Terrestrial Measurements, Faculty of Silviculture and Forest Engineering, Transilvania University of Brasov, 500123 Brasov, Romania
2
Faculty of Civil Engineering, Transilvania University of Brasov, 900152 Brasov, Romania
3
Department of Natural Resources and Environmental Engineering, College of Agriculture, Shiraz University, Shiraz 71348-14336, Iran
4
Quantitative Plant Ecology and Biodiversity Research Lab, Department of Biology, Faculty of Science, Ferdowsi University of Mashhad, Mashhad 91779-48974, Iran
5
Department of Plant Production, Faculty of Agriculture, University of Torbat Heydarieh, Torbat Heydarieh 95161-68595, Iran
6
Saffron Institute, University of Torbat Heydarieh, Torbat Heydarieh 95161-68595, Iran
*
Author to whom correspondence should be addressed.
Remote Sens. 2023, 15(2), 387; https://doi.org/10.3390/rs15020387
Submission received: 17 November 2022 / Revised: 29 December 2022 / Accepted: 6 January 2023 / Published: 8 January 2023
(This article belongs to the Special Issue Unmanned Aerial Vehicle (UAV) and Satellite Synergy for Assessment and Monitoring Biodiversity)
Browse Figures
Versions Notes

Abstract

:
Plant diversity measurement and monitoring are required for reversing biodiversity loss and ensuring sustainable management. Traditional methods have been using in situ measurements to build multivariate models connecting environmental factors to species diversity. Developments in remotely sensed datasets, processing techniques, and machine learning models provide new opportunities for assessing relevant environmental parameters and estimating species diversity. In this study, geodiversity variables containing the topographic and soil variables and multi-seasonal remote-sensing-based features were used to estimate plant diversity in a rangeland from southwest Iran. Shannon’s and Simpson’s indices, species richness, and vegetation cover were used to measure plant diversity and attributes in 96 plots. A random forest model was implemented to predict and map diversity indices, richness, and vegetation cover using 32 remotely sensed and 21 geodiversity variables. Additionally, the linear regression and Spearman’s correlation coefficient were used to assess the relationship between the spectral diversity, expressed as the coefficient of variation in vegetation indices, and species diversity metrics. The results indicated that the synergistic use of geodiversity and multi-seasonal remotely sensed features provide the highest accuracy for Shannon, Simpson, species richness, and vegetation cover indices (R2 up to 0.57), as compared to a single model for each date (February, April, and July). Furthermore, the strongest relationship between species diversity and the coefficient of variation in vegetation indices was based on the remotely-sensed data of April. The approach of multi-model evaluations using the full geodiversity and remotely sensed variables could be a useful method for biodiversity monitoring.

1. Introduction

Assessing biodiversity over large geographical areas is a difficult task [1,2]. There are several traditional in situ approaches for biodiversity surveying [2,3,4]. In this regard, diversity indices, such as the richness, Shannon–Wiener, and Simpson indexes, including the number of individuals within plant communities, may be used to quantify taxonomic diversity [3,4]. Despite the fact that field-based botanical surveys can accurately estimate plant diversity and species composition, they are time consuming and costly, and they often lack a complete spatial coverage [5,6]. As a result, they cannot support checking for trends in biodiversity at large scales. Remote sensing can provide a wider view on the Earth’s surface across space and time [5,6,7] and, as a result, it has frequently been described as a promising solution to monitor biodiversity [8,9] by stepping up monitoring processes at various spatial and temporal scales. When remote sensing data are coupled with field surveys, it can lead to significant improvements and understanding of the natural world. In particular, remotely sensed imagery has become available as open data and significantly improved monitoring of the natural habitats, which is required for biodiversity conservation [10,11].
Recently, remotely sensed data have revolutionized scaled plant monitoring and has been confirmed as a useful method for mapping and monitoring biodiversity [12,13]. In contrast to other land cover types, such as forests or croplands, grasslands have not received much attention in the remote sensing literature [14]. In recent decades, the availability of satellite-based vegetation indices has increased significantly, easing the study of all parts of the world by removing the barriers brought by accessibility and time [15,16]. Spectral diversity, which is the variability in remotely sensed data, as reflected by plants, is the most important indirect observation approach, providing a broad range of possibilities for monitoring species diversity. Remotely sensed features, such as the vegetation indices, have been proved to be useful estimates of productivity and plant diversity, since they characterize vegetation attributes [16,17]; they can be used to estimate species counts or other biodiversity-related metrics from remote sensing data, and quantify the spatial non-uniformity that indicate the patterns in biodiversity [18,19,20,21]. This is based on the link between plant biodiversity and spectral variation [18,22], according to which, the variability in remotely sensed spectral patterns is connected to plant diversity [23]. Accordingly, when the plant species diversity of a specific region increases, a heterogeneous plant community is, therefore, expected to have increased spectral diversity and variability [24].
There have been many documented efforts to monitor biodiversity by remote sensing [25,26,27], but most of these studies have overlooked the impact of temporal dynamics in plant communities (i.e., phenology) on the remote sensing of plant diversity. Specifically, remote sensing in biodiversity studies has often been confined to data gathered at a single time [28,29,30], which is usually constrained by the availability of in situ data. The lack of focus on the temporal dimension was mainly caused by the difficulty of implementing repetitive field data collection campaigns, which are time consuming and costly [31].
Habitat heterogeneity, defined as the spatial and temporal variability in environmental parameters (e.g., rainfall and temperature), management processes (e.g., grazing, fire, and mowing), and other successional shifts, have been identified as fundamental ecosystem attributes affecting the patterns of biodiversity across spatial and temporal scales [32]. We presume that by providing multi-temporal data at fine spatial resolution and across large geographical extents, remote sensing can be useful in estimating environmental heterogeneity and thereby biodiversity. In this research, geodiversity and remotely sensed data were used to identify how biodiversity of grasslands changes over time. The following objectives were set to reach the goal of the study: (i) to evaluate the capability of remotely sensed features, environmental data (hereafter geodiversity), and their integration to estimate plant diversity, (ii) to identify the best time point of the year for estimating plant diversity for the study area, and (iii) to test the relationship between spectral and species diversity as a prerequisite to monitor species diversity. To do so, remotely sensed data, geodiversity features, and in situ measurements taken during the growing season in the southwest of Iran were integrated to develop machine learning algorithms for biodiversity modeling and prediction.

2. Materials and Methods

2.1. Study Area and Sampling Design

Dakal-kooh is a mountainous rangeland, which was used as the study area in this research; it covers 1756 ha, and is located in the north-west of Fars Province, Iran (51.30° to 55.57°E, and 33.32° to 33.37°N; Figure 1). The elevation of the area ranges from 920 to 1391 m (Figure 1). According to the de Martonne aridity index, the climate is semi-arid [33].
The 11-year average rainfall is 489 mm, and the average temperature is 21 °C [33]. A total of 96 (10 × 10 m, Figure 1) plots were chosen at random throughout the study area. Each plot was spatially characterized by its geographical coordinates. In these plots, abundance (cover percentage) of all plant species was recorded between April and May of 2020. The flora in the study region is mostly made of ephemeral and perennial forbs and grasses [33]. The most common species are Aegilops umbellulata Zhuk, Heteranthelium piliferum Hochst. ex Jaub., and Spach and Stipa capensis Thumb, which are accompanied by Amygdalus eburnea Spach, Amygdalus scoparia Spach., Astragalus baba-alliar Parsa, and Cerasus microcarpa Boiss [33].

2.2. Data Preparation

2.2.1. Remotely Sensed Variables

Expert knowledge of vegetation, edaphic, and hydrologic characteristics were used to select a multi-seasonal collection of remotely sensed features. In this regard, the Google Earth Engine (GEE) cloud computing platform was used to create multi-seasonal Sentinel-2 (S2) and Landsat-8 (L8) surface reflectance image collections. For image collection, three-time milestones were considered: mid-winter (February 2019), when some herbaceous plant species start their growth; mid-spring (April 2020) during the peak growing season; and mid-summer (July 2020) during the senescence of most herbaceous plants (Figure 2). S2 monthly image collections were used to extract the mean spectral indices, and L8 datasets were used to extract the monthly land surface temperature (LST). These processes resulted in 32 features extracted from remotely sensed data (Table 1).

2.2.2. Geodiversity Variables

Geodiversity variables used in this study were a combination of topographic, climatic, and edaphic variables (Figure 2). Twelve topographic factors were considered, namely the altitude (ALT), slope degree (SLP), aspect (ASP), profile (PRC) and plan curvatures (PLC), stream power index (SPI), Multiresolution Index of the Ridge Top Flatness (MRRTF), topographic position index (TPI), topographic wetness index (TWI), Topographic Ruggedness Index (TRI), multiresolution index of valley bottom flatness (MRVBF), and Mass Balance Index (MBI). The freely available ALOS PALSAR—Radiometric Terrain Correction DEM (https://vertex.daac.asf.alaska.edu/ accessed on 16 November 2022), with a resolution of 12.5 m, was used to generate the topographic layers.
Long-term mean annual rainfall (Rain), calculated based on the datasets provided by the Meteorological Office of Fars province (http://www.farsmet.ir accessed on 16 November 2022), was used as a climatic variable. Moreover, soil acidity (pH), electrical conductivity (EC), sand, clay, and silt percentages, organic carbon (OC) content, organic matter content (OM), and nitrogen content (N) were used to specify the edaphic features. In this regard, 50 samples were collected in the upper layer (0 to 30 cm in depth) of the soils from the study area during the study of Rahmanian et al. [2]. These were used herein to evaluate the relevant soil parameters across the study area (Figure 1). Procedures used to process the soil samples and to estimate the relevant properties are given in Rahmanian et al. [2]. Then, the edaphic values were interpolated using the inverse distance weighting (IDW) interpolation method to generate raster layers. In IDW interpolation, the closest measured values place more weight on determining the anticipated value than those further away [57]. IDW was performed by using the ArcGIS geoprocessing tool (version 10.8).

2.3. Statistical Analysis

2.3.1. Statistical Workflow

This study considered the available geodiversity and time series of remotely sensed variables that can predict the plant diversity metrics in the study area. To find the most important variables that affect plant diversity, a multicollinearity analysis was used. To characterize plant diversity, in addition to the species richness and vegetation cover that were recorded during field sampling in each plot, the most popular diversity metrics, such as the Simpson (sensitive-to-dominant species) and Shannon (sensitive-to-rare species) indices, were estimated for each plot. To map and predict species diversity (Shannon’s, Simpson’s indexes), and species richness and vegetation cover, several random forest (RF) models were developed based on the set of remotely sensed and geodiversity variables, as well as based on a combination of them in three seasons (winter, spring, and summer). To find out which variables placed the most importance on predicting plant diversity and vegetation attributes, a variable importance analysis was implemented. In addition, spectral diversity is characterized hereafter by the coefficient of variation (CV), which reflects the variation in remote sensing data spectrum based on the patterns in vegetation characteristics of the plant species. Since CVs of vegetation indices are related to traditional measures of plant diversity [58], the relationships between CVs of vegetation indices, species diversity, and vegetation cover were tested to estimate plant diversity.

2.3.2. Preparation of Variables

By multicollinearity testing of independent variables, only the most influential variables were kept among the associated pairs using a stepwise process. The strongly associated variables were excluded to avoid potential overfitting issues. The “vifstep” function of “usdm” R package was used to examine the collinearity among the 53 variables used in this study. The variance inflation factor (VIF) and tolerance (T) were used as metrics in multicollinearity testing. The presence of significant collinearity was defined for the condition in which the VIF index was larger than 5.0 and T was less than 0.1 [59]. By locating and eliminating strongly associated variables, the size of the datasets was reduced (Table 2).
Diversity metrics, namely species richness, Shannon and Simpson indices, and also vegetation cover, were measured in this study. Species richness stands for the number of species for which ground sampling makes it easier to obtain an accurate value. Percentage cover is an abundance measurement metric that indicates how much space a species occupies in a plot, and it was recorded during the field sampling. R software (version 3.5.1) was used to calculate Shannon’s H and Simpson’s D diversity indices by implementing the package “vegan” (R Core Team, R Foundation for Statistical Computing, Vienna, Austria). These two indices take species uniformity into account and were calculated using the following equations [60,61]:
Shannon ’ s   Index   ( H ) = − ∑ [ p i × ln ( p i ) ]
Simpson ’ s   Index   ( D ) = 1 / ∑ ( p i ) 2  
where pi is the ratio (n/N) of the number of individuals of one species (n) to the total number of individuals (N). The number of species present in an environment is described as the species richness [62].
To monitor spectral diversity, which refers to the variation in remote sensing measurements, the vegetation indices (GNDVI, GRVI, MASVI, Ndre1, NDVIre2, SATVI, TCIre1, VSDI, V, TVI, EVI, MSAVI, and CRSI) were used as input for estimating the coefficient of variation (CV) between all the pixels inside the study area. The CV was calculated across the entire study area as the standard deviation divided by the mean value of all pixels located in a reference moving window of 3 × 3 pixels [63,64]. Then, the pixel values were extracted for the 96 studied plots as these pixel values relate to conventional metrics of biodiversity [63].
The linear regression and Spearman’s correlation coefficient were used to find out what kind of relationship exists between CV of vegetation indices and plant diversity metrics (Shannon, Simpson, richness, and vegetation cover). By this approach, the aim was to find if there was a strong relationship, and which relationship was the strongest between the coefficient of variation in vegetation indices and diversity metrics.

2.4. Modeling Species Diversity by Random Forest

Several random forest (RF) machine learning models were used to map and predict species diversity (Shannon’s, Simpson’s, and richness indexes) and vegetation cover, taking the remotely sensed, topographic, climatic, and edaphic variables as inputs. A training sample accounting for 60% of the plots was used. RF is a machine learning algorithm that consists of an ensemble of bootstrapped trees, which are used to reach a vote-based decision, making it useful and performant in classification and regression problems [65]. RF generates accurate forecasts that are simple to comprehend, and it is able to effectively handle large datasets. Compared to the decision tree algorithm, RF is more accurate in forecasting outcomes [66,67].
In total, 48 models were developed to predict four plant diversity indices (Shannon, Simpson, richness and vegetation cover) by considering four time milestones (February, April and July, and their aggregate expressed as multi-seasonal) and three groups of variables (geodiversity, remotely sensed, and a combination of the two). In detail, a subset of sixteen RF models were based on the remotely sensed data and were developed for each of the plant diversity indices (Shannon and Simpson), species richness, and vegetation cover; of these, three models were developed to predict plant diversity for each time milestone (February, April, and July), with remotely sensed variables (Table A1, Appendix A), and one model was used to predict plant diversity for all time milestones (multi-seasonal) considered together (Table A1, Appendix A). Moreover, a subset of sixteen RF models were developed taking the geodiversity variables as inputs (Table A1, Appendix A) to predict diversity indices (Shannon and Simpson), richness and vegetation cover for each time milestone, and one model using geodiversity variables was developed to predict plant diversity for all time milestones considered together (Table A1, Appendix A). Finally, twelve models were developed using the final set of remotely sensed and geodiversity variables (Table A1, Appendix A) for each time milestone (February, April, and July), and one model was developed for all the time milestones (multi-seasonal), with a combination of all geodiversity and remotely sensed variables (Table A1, Appendix A). Then, the maps of diversity metrics were prepared based on the final set of remotely sensed and geodiversity variables.
To understand the importance of geodiversity and remotely sensed variables in predicting plant diversity metrics of each time milestone (February, April, and July) by RF models, variable importance measure was estimated [65,68] in “Caret” package of R [69]. Ten of the most important variables were selected based on their scores, which were scaled from 0 to 100 [70].
RF models were developed in “randomForest” R package, and two main hyperparameters related to this model were tuned to enhance the modelling performance in terms of learning and generalization: the number of random variables used in each tree (mtry) and the number of regression trees (ntree), which was optimized based on an out-of-bag estimate error [65]. To infer the optimal values of these hyperparameters, a ten-fold cross-validation resampling with 50 iterations was applied to the training samples. For more details about the optimal values of hyperparameters, see Table 3.

Evaluation of the Models’ Performance

Spatial distribution of the Shannon’s, Simpson’s, richness’, and vegetation cover’s indexes in the three studied seasons was predicted by RF (Figure 3). Based on the species presence in the sampling plots, the values of predicted diversity, vegetation cover, and richness were extracted from the maps obtained for the considered seasons. Models’ performance was evaluated by using the conventional methods of comparing the predicted values against those from the testing samples (40% of plots Section 2.1). The used metrics were the root mean square error (RMSE), mean absolute error (MAE), and the coefficient of determination (R2) [71,72].

3. Results

3.1. Multicollinearity of Independent Variables

The multicollinearity test results showed that lithology, wind, SLP, ASP, Rain, PLC, PRC, TWI, NSDSI3, GNDVI, NDVI, SATVI, VDSI, GRVI, TVI, ClaI, GRVI, NDVIre1, Ndre1, BI, S3, SI, NMDI, Int1, Int2, and BI2 were strongly correlated with other variables for the February dataset; therefore, these variables were excluded. Wind, SLP, sand, NDSI3, GSI, SI, GRVI, OC, BI, CaI, NDVIre, NDre1, GSI, TVI, Int1, Int2, NDWI, SI, GNDVI, CalsI, CI, V, SMMI, SAVI, MASVI, VDSI, and MSI were significantly correlated with other variables in the April dataset, and, therefore, they were excluded. For similar reasons, wind, sand, SI, NSDSI3, OC, Int1, Int2, CI, LSWI, BI, BI2, MSAVI, GNDVI, GRVI, SMMI, CalcI, NDre1, and NDWI were deleted from the July dataset. The list of variables kept for further analysis is shown in Table A1.

3.2. Relationship between Species and Spectral Diversity

The results of the general linear model analysis demonstrated a significant positive relationship (p < 0.05) between the CV of vegetation indexes, diversity indices, and vegetation cover in all three seasons. However, the mid-spring had the highest relationship between species diversity and the CV of vegetation indices. Among the diversity metrics, Shannon’s and Simpson’s indices showed a stronger relationship with spectral diversity (expressed as the CV of vegetation indices), as opposed to the species richness and vegetation cover (Table 2). These indices were correlated with most of the CV of vegetation in mid-spring (see Table 2 for more details). Furthermore, for mid-winter and mid-summer there were few significant relationships between the diversity indices and the CV of vegetation indices (Table 2). In mid-winter, TVI, MASVI, and CRSI significantly correlated with the Shannon index, and MASVI, SATVI, NDVI, VSDI, V, EVI, and CRSI had a significant correlation with the Simpson index. Moreover, only GRVI had a correlation with species richness. In mid-summer, MASVI, Ndre1, NDVIre2, SATVI and NDVI, SATVI and MASVI, GRVI, and MASVI and SATVI had significant correlations with the Shannon, Simpson, and richness indexes, respectively.

3.3. Performance of Plant Diversity Prediction and Mapping Accuracy

A summary of the models developed by using the full set of geodiversity and remotely sensed variables and testing samples is presented in Table 3. The results indicated that predictions of the examined diversity indices and species richness were more accurate when using multi-seasonal models for Shannon’s (R2 = 0.28, RMSE = 0.29, MAE = 0.237), Simpson’s (R2 = 0.44, RMSE = 0.02, MAE = 0.018), and species’ richness indexes (R2 = 0.27, RMSE = 0.075, MAE = 0.58) as well as vegetation cover (R2 = 0.2, RMSE = 0.11, MAE = 0.092). Regarding the single date models, the best results were produced using the April combined data model; results for Shannon’s (R2 = 0.28, RMSE = 0.312, MAE = 0.239), Simpson’s (R2 = 0.43, RMSE = 0.023, MAE = 0.019), and species’ richness indexes (R2 = 0.26, RMSE = 0.08, MAE = 0.6), as well as vegetation cover (R2 = 0.17, RMSE = 0.112, MAE = 0.085), are shown in Table 3.
The comparison of the validation results for the best-performing diversity models based on all variables, geodiversity- or remotely sensed-only variables, indicated that using the first option returned more accurate predictions in multi-seasonal and single-date models, except for vegetation cover (R2 = 0.21; RMSE = 0.11; MAE = 0.85) in April, which was more accurately predicted by the use of remotely sensed variables, as well as Shannon’s (R2 = 0.3; RMSE = 0.3; MAE = 0.24) and Simpson’s indices (R2 = 53; RMSE = 0.023; MAE = 0.017), for February models, which were more accurately predicted by geodiversity variables. Moreover, the accuracy results obtained from all single-date and multi-seasonal models suggest that models predicting the Simpson’s diversity were the best performing ones (Table 3). As the most precise models using the entire geodiversity and remotely sensed datasets, three single dates are shown in Figure 3 to depict the mapping of the best plant diversity prediction based on the RF models.

3.4. The Most Predictive Independent Variables

As shown in Figure 4, Figure 5 and Figure 6, the results indicate the relevance of the best-selected models for species diversity, richness, and vegetation cover indicators. The findings on the variable importance analysis revealed that in mid-winter (February) and mid-summer (July), remotely sensed variables had more importance than geodiversity variables (Figure 4 and Figure 6). In this regard, Soil Adjusted Total Vegetation Index (SATVI), Ferrous Iron (FeI), Enhanced Vegetation Index (EVI), Saturation Index-T (SI-T), Canopy Response Salinity Index (CRSI), Carbonate Index (CaI), Normalized Difference Vegetation Index (NDVI), Normalized Difference red-edge 1 (NDre1), and Green Normalized Difference Vegetation Index (GNDVI) were the most important remotely sensed variables among the spectral variables in mid-winter for modeling and predicting the diversity (Figure 4), whereas Topographic Ruggedness Index (TRI), Mass Balance Index (MBI), and Plan Curvature (PLC) were the most important geodiversity variables (Figure 4).
Regarding the most important variables in mid-summer (July), land surface temperature (LST), Canopy Response Salinity Index (CRSI), Hue Index (HI), Soil Adjusted Total Vegetation Index (SATVI), Normalized Difference red-edge 1 (Ndre1), and Ferrous iron (FeI) were the most important spectral variables while rainfall (Rain), altitude (ALT), Topographic Wetness Index (TWI), nitrogen content (N), electrical conductivity (EC), and slope (SLP) were the most important geodiversity variables that influenced the prediction of Shannon’s, Simpson’s, and richness’ indexes, as well as vegetation cover (Figure 6).
In contrast, in mid-spring (April), the geodiversity variables were more important compared to the remotely sensed variables in modeling and predicting the vegetation cover, richness, and species diversity. For instance, the geodiversity variables, including Mass Balance Index (MBI), rainfall (Rain), Topographic Wetness Index (TWI), Topographic Ruggedness Index (TRI), lithology, altitude (ALT), silt, Multiresolution Index of the Ridge Top Flatness (MRRTF), and profile curvature (PRC), and remotely sensed variables, including land surface temperature (LST), Calcareous Sedimentary Rocks Index (CalcI), and Salinity Index (S3), were the important variables (Figure 5).

4. Discussion

In this study, test results for the best-performing diversity models indicated that the full set of predictor variables was more accurate than the subsets of variables taken separately (spectral- or geodiversity data) in multi-seasonal and single-date models (Table 3), which is in agreement with findings of Chrysafis et al. [73,74] and Senf et al. [75]. As for this study, the accuracy of the February, April, and July models was significantly enhanced when remotely sensed and geodiversity variables were combined. This is related to the relative importance of the input variables in each model, which leads to different contributions to model fitting. The reason is that RF is not a parsimonious method and uses all of the variables in the datasets to make predictions [76]. In addition to that, the predictions of plant diversity by the use of combined geodiversity and time series of remotely sensed data had a greater accuracy on a multi-seasonal model compared to single dates. This may be related to the fact that multi-seasonal satellite imagery captures the whole spectrum of phenological variations in various species, boosting the spectral differentiation [74,77,78]. The effectiveness of multi-seasonal models of this study is consistent with prior research findings. For example, the combination of two remotely sensed images recorded in winter and summer periods increased forest tree diversity prediction performance in northern Greece [79].
Concerning the single dates, the prediction of species diversity was found to be more accurate on the scenes acquired in April compared to those of February and July. This may be due to the optimal vegetation and high photosynthesis of April that enabled the capturing of the vegetation changes in the reflectance of fully formed leaves. The reduced photosynthetic activity of both evergreen and deciduous plants, as well as the water deficit in the preceding months, might explain the image’s low accuracy in July. Additionally, the poor accuracy in February might be because the leaf emergence did not start for many of the deciduous plants in the area. To back up these findings, a previous study has shown that the detection of species diversity is most accurate during the change over seasons (February in our instance), while spectral imaging precision is the highest during the growing season [79]. In contrast, Torresani et al. [80] found a greater link between tree species diversity and variance in reflectance as of July in south Tyrol (Italy). Additionally, Arekhi et al. [81] stated that the image taken in the summer helped to develop the most appropriate model by examining the link between various Landsat TM spectral bands (original and synthetic) and tree diversity (Shannon’s index) in a temperate forest from Turkey.
Vegetation indices (VI) were found to be important parameters for monitoring plant diversity in the study area. The spectral diversity measured by the coefficient of variation in vegetation indices (CV) measures the amount of variation stored in each vegetation index, and this variability is thought to imply increased environmental heterogeneity related to diverse species composition and species diversity [82]. The results indicated a significant relationship between spectral diversity, calculated as the coefficient of variation in vegetation indices, and species diversity, calculated as Shannon’s diversity index, Simpson’s diversity index, species richness, and vegetation cover. The CV demonstrates that the three seasons hold separate mechanisms in controlling this variation. In comparison to winter and summer, in the spring there is a stronger relationship between the increased CV of vegetation indices and the increase in plant diversity metrics that reflects the heterogeneity of the vegetation spectrum caused by some variability in species characteristics, including phenological and physiological traits of plant species. As found in this study, remotely sensed vegetation indices help in estimating the plant diversity. This result is consistent with previous studies [83,84,85]. Moreover, the results of this study showed that the relationship between Shannon’s and Simpson’s diversity index and the CV of vegetation indices as spectral diversity were stronger than the species richness and vegetation cover. Both Simpson and Shannon diversity indices are metrics that consider the abundance of species in a given plot. However, species richness values may appear to be high due to the rare species that are present in the plots. Compared to the spectral diversity that depends on the relative abundance of each species, it has been demonstrated that Shannon and Simpson indices, which take species abundance into account, are more accurate.
Concerning the variable importance of the models used in this study, diversity indices of different seasons were not equally related to the landscape features. For instance, the most important variables affecting plant diversity metrics were topographic variables and vegetation indices in winter. In spring, topographic variables had the most important role, and in summer, vegetation indices and edaphic variables were among the most important factors affecting plant diversity metrics [86]. Due to the mountainous nature of the studied area, it is likely that complex topographic features have shaped a variety of microclimates over a short distance [87], which could provide suitable microhabitats for a wide range of plant species holding different phenological and morphological characteristics [88]. Therefore, in this study, the topographic variables may have influenced plant diversity in summer and spring. Previous research confirmed the important role of topographical variables on plant diversity. Robinson et al. [89] stated that elevation is essential for diversity estimations. Moreover, across the United States, Read et al. [90] have found that tree diversity was positively associated with elevation, slope, and aspect.
A higher importance of vegetation indices on plant diversity in winter can be due to the role of winter cold in limiting vegetation and subsequently plant diversity in the mountainous rangelands. In this regard, some vegetation indices, such as EVI and SATVI, showed these changes to be better than others. This may be due to a higher sensitivity of EVI to seasonal changes in canopy’s biophysical attributes than other vegetation indices, which is in line with the previous study by Galvão et al. [91]. Moreover, SATVI has been demonstrated to capture temporal changes in grassland vegetation cover more efficiently than other vegetation indices because it relates with both live and senescent vegetation cover [39,92,93], potentially identifying changes related to vegetation. Interestingly, the variables characterizing the severe vegetation loss in July included high land surface temperature (LST), soil salinity, and deficit of rainfall, which are completely related to the condition of the arid regions in July.
Protecting biodiversity and ecosystem functioning is important for the delivery of basic services to human society. However, there are limitations in monitoring species diversity and figuring out the environmental factors that affect it in various geographic setups. The accelerating loss of biodiversity makes this problem even more important. It should be noted that ecosystem functions are rarely quantified, especially across large areas [94]. To solve this problem, satellite remote sensing, with relatively high temporal and spatial resolutions, offers global coverage and continuous measurements of plant diversity in space [95].
Among the limitations of this study were the scale of the study area and the relatively small number of samples, which were not especially well distributed throughout the region due to the severe topographic barriers in the study area. New opportunities to simultaneously measure plant structure and terrain morphology at fine spatial scales are presented by recent developments in airborne laser scanning (ALS). Therefore, it is necessary to address these issues in future studies by directly measuring environmental factors with grain sizes as small as tens of meters, which can reveal links between biodiversity and geodiversity at various scales.
Although the samples of this study were sufficient for conventional machine learning models, they were not optimal for deep learning. As a result, having more samples would enable the use of deep learning, which may prove more accurate. Therefore, these issues could be considered in future studies. Because grasslands located in different biomes clearly differ in the properties of plant species that inhabit them, the remotely sensed evaluation of their spectral diversity still needs to be biome-specific. It is also important to conduct similar studies on different scales to check whether the spectral diversity is scale-dependent or not.

5. Conclusions

This study provides insights into the usefulness of geodiversity and seasonal remotely sensed data in forecasting plant diversity patterns in dry rangelands. A non-linear model was used to predict vegetation cover, diversity, and richness. The link between plant diversity and reflectance is season-dependent, according to this study. The results complement the current knowledge, which is typically based on assessing the species diversity in the growing season. The accuracy of the models was enhanced by the use of satellite images taken during the peak of photosynthetic output and vegetation optimum (i.e., April). On the other hand, the use of satellite images taken in the middle of winter, when the leaf emergence for all deciduous species from the area is still incomplete, has led to a level of diversity, which is less than the actual level of diversity. Moreover, in vegetation diversity modeling, remotely sensed data might be more relevant than geodiversity data in winter and summer. Despite that, the geodiversity data had a greater contribution to the species diversity in the spring. Such evidence supports the integration of geodiversity and remotely sensed data for plant diversity monitoring and mapping across landscapes. The results of such studies can be synthesized to provide documented information on plant diversity, which, in turn, will promote the conservation in rangeland ecosystems and will make it easier for ecologists to comprehend the temporal and spatial patterns of biodiversity.

Author Contributions

Conceptualization, S.R. and S.A.B.; formal analysis S.R.; investigation, S.R. and S.K.; methodology, S.R., A.A., S.M., V.N. and S.P.; software, S.R.; supervision, S.R. and S.A.B.; validation, S.R. and S.A.B.; visualization, S.R. and V.N.; writing, S.R.; writing—reviewing S.R., V.N. and S.A.B. All authors have read and agreed to the published version of the manuscript.

Funding

This work received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data supporting the findings of this study are available from the first author (S.R.) upon reasonable request.

Acknowledgments

We thank to Hamed Gholizadeh for valuable and profound suggestions in the manuscript.

Conflicts of Interest

The authors declare no competing interest.

Appendix A

Table
Table A1. Selected variables for modeling plant diversity in single-date and multi-seasonal models for each time milestone (February, April, and July).
Table A1. Selected variables for modeling plant diversity in single-date and multi-seasonal models for each time milestone (February, April, and July).
Time ScaleModelVariable
FebruaryGeodiversityTRI, TPI, MRVBF, MRRTF, MBI, Silt, ALT, pH, OM, EC, TWI, Rain, PRC, PLC, N, ASP
Remotely sensedNDWI, NDVIre2, NDre1, VSDI, SI, SATVI, MSI, HI, GNDVI, FeI, EVI, CRSI, CI, CalcI, CaI
All dataVSDI, SI, SATVI, NDWI, NDVIre2, NDre1, MSI, HI, GNDVI, FeI, EVI, CRSI, CI, CalcI, CaI, TRI, TPI, MRVBF, MRRTF, MBI, Silt, ALT, pH, OM, LST, CN, EC, CLY, TWI, Rain, PRC, PLC, N, ASP
AprilGeodiversityASP, ALT, PLC, PRC, Rain, Silt, TWI, EC, N OM, MBI, MRRTF, MRVBF, TPI, TRI, Lithology, Cly
Remotely sensedEVI, BI2, CalcI, CRSI, FeI, HI, NDVIre2, NMDI, SATVI, S3, TCIre1, LST, NDre2
All dataNDVI, ASP, ALT, CN, EVI, EC, N, OM, PH, PLC, PRC, Rain, Silt, TWI, MBI, MRRTF, MRVBF, TPI, TRI, BI2, CalcI, CRSI, FeI, HI, NDVIre2, NMDI, SATVI, S3, TCIre1, Lithology, Clay, LST, NDre2
JulyGeodiversityASP, ALT, Clay, EC, N, OM, pH, PLC, PRC, RN, SLP, SLT, TWI, NMDI, GRVI, Lithology, MBI, MRRTF, MRVBF, TPI
Remotely sensedEVI, FeI, GSI, HI, N, NDVI, LST, NDre2, S3, V, VDSI, NDVIre2, TCIre1, SATVI, CRSI
All dataASP, ALT, Clay, EC, EVI, FeI, GSI, HI, N, NDVI, OM, PH, LST, PLC, PRC, RN, SLP, SLT, TWI, NDre2, NMDI, GRVI, MSI, TVI, S3, V, VDSI, Lithology, NDVIre2, TCIre1, SATVI, MBI, MRRTF, MRVBF, TPI, CN, CRSI
Multi seasonalGeodiversityASP, ALT, Cly, EC, N, OC, pH, PLC, PRC, Rain, SLP, Silt, TWI, MBI, MRRTF, MRVBF, TPI, TRI, CN
Remotely sensedBI2_April, CalcI_April, CaI_April, EVI_April, HI_April, LST_April, NDVI_April, NDre2_April, NMDI_April, NDWI_April, CRSI_April, MSAVI_April, SAVI_July, SMMI_July, TVI_July, V_July, VDSI_July, NDVIre1_July, TCIre1_July, SAVI_July, CRSI_July, LSTJ, LST_July, BI2_April, CalcI_April, FeI_April, GSI_April, NDVIre2_April, NMDI_April, SMMI_April, TCIre1_April, LST_April, BI_Feb, BI2_Feb, CaI_Feb, CalcI_Feb, CRSI_Feb, EVI_Feb, FeI_Feb, GNDVI_Feb, GSI_Feb, MSI_Feb, NDre2_Feb, NDVIre2_Feb, NDWI_Feb, SATVI_Feb, SMMI_Feb
All dataASP, ALT, BI, CaI, CalcI, CLY, EC, EVI, HI, LST, N, NDVI, OC, PH, PLC, PRC, Rain, SLP, Silt, TWI, NDre2_July, NMDI_July, NDWI_July, CRSI_July, LSWI_July, MSAVI_July, SAVI_July, SMMI_July, TVI_July, V_July, VDSI_July, Lithology, NDVIre2_July, TCIRE1_July, SATVI_July, MBI, MRRTF, MRVBF, TPI, TRI, CN, CRSI_July, LST_July, BI2_April, CalcI_April, FeI_April, GSI_April, NDVIre2_April, NMDI_April, SMMI_April, TCIre1_April, LST_April, BI_Feb, BI2_Feb, CaI_Feb, CalcI_Feb, CRSI_Feb, EVI_Feb, FeI_Feb, GNDVI_Feb, GSI_Feb, MSI_Feb, NDre2_Feb, NDVIre2_Feb, NDWI_Feb, SATVI_Feb, SMMI_Feb

References

  1. Kissling, W.D.; Walls, R.; Bowser, A.; Jones, M.O.; Kattge, J.; Agosti, D.; Amengual, J.; Basset, A.; van Bodegom, P.M.; Cornelissen, J.H.C.; et al. Towards global data products of Essential Biodiversity Variables on species traits. Nat. Ecol. Evol. 2018, 2, 1531–1540. [Google Scholar] [CrossRef] [Green Version]
  2. Rahmanian, S.; Hejda, M.; Ejtehadi, H.; Farzam, M.; Pyšek, P.; Memariani, F. Effects of livestock grazing on plant species diversity vary along a climatic gradient in northeastern Iran. Appl. Veg. Sci. 2020, 23, 551–561. [Google Scholar] [CrossRef]
  3. Gotelli, N.J.; Colwell, R.K. Quantifying biodiversity: Procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 2001, 4, 379–391. [Google Scholar] [CrossRef] [Green Version]
  4. Rahmanian, S.; Hejda, M.; Ejtehadi, H.; Farzam, M.; Memariani, F.; Pyšek, P. Effects of livestock grazing on soil, plant functional diversity, and ecological traits vary between regions with different climates in northeastern Iran. Ecol. Evol. 2019, 9, 8225–8237. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  5. He, K.S.; Bradley, B.A.; Cord, A.F.; Rocchini, D.; Tuanmu, M.-N.; Schmidtlein, S.; Turner, W.; Wegmann, M.; Pettorelli, N. Will remote sensing shape the next generation of species distribution models? Remote Sens. Ecol. Conserv. 2015, 1, 4–18. [Google Scholar] [CrossRef] [Green Version]
  6. Rocchini, D.; Boyd, D.S.; Féret, J.-B.; Foody, G.M.; He, K.S.; Lausch, A.; Nagendra, H.; Wegmann, M.; Pettorelli, N. Satellite remote sensing to monitor species diversity: Potential and pitfalls. Remote Sens. Ecol. Conserv. 2016, 2, 25–36. [Google Scholar] [CrossRef]
  7. Pettorelli, N.; Laurance, W.F.; O’Brien, T.G.; Wegmann, M.; Nagendra, H.; Turner, W. Satellite remote sensing for applied ecologists: Opportunities and challenges. J. Appl. Ecol. 2014, 51, 839–848. [Google Scholar] [CrossRef]
  8. Jetz, W.; Cavender-Bares, J.; Pavlick, R.; Schimel, D.; Davis, F.W.; Asner, G.P.; Guralnick, R.; Kattge, J.; Latimer, A.M.; Moorcroft, P.; et al. Monitoring plant functional diversity from space. Nat. Plants 2016, 2, 16024. [Google Scholar] [CrossRef] [Green Version]
  9. Cavender-Bares, J.; Gamon, J.A.; Hobbie, S.E.; Madritch, M.D.; Meireles, J.E.; Schweiger, A.K.; Townsend, P.A. Harnessing plant spectra to integrate the biodiversity sciences across biological and spatial scales. Am. J. Bot. 2017, 104, 966–969. [Google Scholar] [CrossRef] [Green Version]
  10. Luque, S.; Pettorelli, N.; Vihervaara, P.; Wegmann, M. Improving biodiversity monitoring using satellite remote sensing to provide solutions towards the 2020 conservation targets. Methods Ecol. Evol. 2018, 9, 1784–1786. [Google Scholar] [CrossRef]
  11. Schimel, D.; Townsend, P.A.; Pavlick, R. Prospects and pitfalls for spectroscopic remote sensing of biodiversity at the global scale. In Remote Sensing of Plant Biodiversity; Gamon, J.A., Townsend, P.A., Eds.; Springer: Cham, Switzerland, 2020; pp. 503–518. [Google Scholar]
  12. Lefsky, M.A.; Cohen, W.B.; Parker, G.G.; Harding, D.J. Lidar remote sensing for ecosystem studies: Lidar, an emerging remote sensing technology that directly measures the three-dimensional distribution of plant canopies, can accurately estimate vegetation structural attributes and should be of particular interest to forest, landscape, and global ecologists. BioScience 2002, 52, 19–30. [Google Scholar]
  13. Wedeux, B.; Dalponte, M.; Schlund, M.; Hagen, S.; Cochrane, M.; Graham, L.; Usup, A.; Thomas, A.; Coomes, D. Dynamics of a human-modified tropical peat swamp forest revealed by repeat lidar surveys. Glob. Chang. Biol. 2020, 26, 3947–3964. [Google Scholar] [CrossRef] [PubMed]
  14. Newton, A.C.; Cayuela, L.; Echeverría, C.; Armesto, J.J.; Del Castillo, R.F.; Golicher, D.; Geneletti, D.; Gonzalez-Espinosa, M.; Huth, A.; López-Barrera, F.; et al. Toward integrated analysis of human impacts on forest biodiversity: Lessons from Latin America. Ecol. Soc. 2009, 14, 2. [Google Scholar] [CrossRef] [Green Version]
  15. Stereńczak, K.; Laurin, G.V.; Chirici, G.; Coomes, D.A.; Dalponte, M.; Latifi, H.; Puletti, N. Global airborne laser scanning data providers database (GlobALS)—A new tool for monitoring ecosystems and biodiversity. Remote Sens. 2020, 12, 1877. [Google Scholar] [CrossRef]
  16. Pettorelli, N.; Ryan, S.; Mueller, T.; Bunnefeld, N.; Jędrzejewska, B.; Lima, M.; Kausrud, K. The Normalized Difference Vegetation Index (NDVI): Unforeseen successes in animal ecology. Clim. Res. 2011, 46, 15–27. [Google Scholar] [CrossRef]
  17. Tucker, C.J. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 1979, 8, 127–150. [Google Scholar] [CrossRef] [Green Version]
  18. Rocchini, D.; Chiarucci, A.; Loiselle, S.A. Testing the spectral variation hypothesis by using satellite multispectral images. Acta Oecologica 2004, 26, 117–120. [Google Scholar] [CrossRef]
  19. Tucker, C.J.; Sellers, P.J. Satellite remote sensing of primary production. Int. J. Remote Sens. 1986, 7, 1395–1416. [Google Scholar] [CrossRef] [Green Version]
  20. Mittelbach, G.G.; Steiner, C.F.; Scheiner, S.M.; Gross, K.L.; Reynolds, H.L.; Waide, R.B.; Willig, M.R.; Gough, L. What is the observed relationship between species richness and productivity? Ecology 2001, 82, 2381–2396. [Google Scholar] [CrossRef]
  21. Farwell, L.S.; Gudex-Cross, D.; Anise, I.E.; Bosch, M.J.; Olah, A.M.; Radeloff, V.C.; Razenkova, E.; Rogova, N.; Silveira, E.M.; Smith, M.M.; et al. Satellite image texture captures vegetation heterogeneity and explains patterns of bird richness. Remote Sens. Environ. 2021, 253, 112175. [Google Scholar] [CrossRef]
  22. Schmidtlein, S.; Fassnacht, F.E. The spectral variability hypothesis does not hold across landscapes. Remote Sens. Environ. 2017, 192, 114–125. [Google Scholar] [CrossRef]
  23. Rocchini, D.; Luque, S.; Pettorelli, N.; Bastin, L.; Doktor, D.; Faedi, N.; Feilhauer, H.; Féret, J.-B.; Foody, G.M.; Gavish, Y.; et al. Measuring β-diversity by remote sensing: A challenge for biodiversity monitoring. Methods Ecol. Evol. 2018, 9, 1787–1798. [Google Scholar] [CrossRef] [Green Version]
  24. Xu, C.; Zeng, Y.; Zheng, Z.; Zhao, D.; Liu, W.; Ma, Z.; Wu, B. Assessing the Impact of Soil on Species Diversity Estimation Based on UAV Imaging Spectroscopy in a Natural Alpine Steppe. Remote Sens. 2022, 14, 671. [Google Scholar] [CrossRef]
  25. Rossi, C.; Kneubühler, M.; Schütz, M.; Schaepman, M.E.; Haller, R.M.; Risch, A.C. Remote sensing of spectral diversity: A new methodological approach to account for spatio-temporal dissimilarities between plant communities. Ecol. Indic. 2021, 130, 108106. [Google Scholar] [CrossRef]
  26. Silveira, E.M.; Radeloff, V.C.; Martinuzzi, S.; Pastur, G.J.M.; Rivera, L.O.; Politi, N.; Lizarraga, L.; Farwell, L.S.; Elsen, P.R.; Pidgeon, A.M. Spatio-temporal remotely sensed indices identify hotspots of biodiversity conservation concern. Remote Sens. Environ. 2021, 258, 112368. [Google Scholar] [CrossRef]
  27. Reddy, C.S. Remote sensing of biodiversity: What to measure and monitor from space to species? Biodivers. Conserv. 2021, 30, 2617–2631. [Google Scholar] [CrossRef]
  28. Oldeland, J.; Wesuls, D.; Rocchini, D.; Schmidt, M.; Jürgens, N. Does using species abundance data improve estimates of species diversity from remotely sensed spectral heterogeneity? Ecol. Indic. 2010, 10, 390–396. [Google Scholar] [CrossRef]
  29. Féret, J.B.; Asner, G.P. Mapping tropical forest canopy diversity using high-fidelity imaging spectroscopy. Ecol. Appl. 2014, 24, 1289–1296. [Google Scholar] [CrossRef] [Green Version]
  30. Schäfer, E.; Heiskanen, J.; Heikinheimo, V.; Pellikka, P. Mapping tree species diversity of a tropical montane forest by unsupervised clustering of airborne imaging spectroscopy data. Ecol. Indic. 2016, 64, 49–58. [Google Scholar] [CrossRef]
  31. Wang, R.; Gamon, J.A.; Schweiger, A.K.; Cavender-Bares, J.; Townsend, P.A.; Zygielbaum, A.I.; Kothari, S. Influence of species richness, evenness, and composition on optical diversity: A simulation study. Remote Sens. Environ. 2018, 211, 218–228. [Google Scholar] [CrossRef]
  32. Fuhlendorf, S.D.; Fynn, R.W.; McGranahan, D.A.; Twidwell, D. Heterogeneity as the basis for rangeland management. In Rangeland Systems; Briske, D.D., Ed.; Springer: Cham, Switzerland, 2017; pp. 169–196. [Google Scholar]
  33. Karami, S.; Khosravi, A.R. A Floristic Study of Kuh-e Dakal in Mamasani County, Fars Province. J. Taxon. Biosyst. 2019, 11, 1–12. [Google Scholar]
  34. Scudiero, E.; Skaggs, T.H.; Corwin, D.L. Regional-scale soil salinity assessment using Landsat ETM+ canopy reflectance. Remote Sens. Environ. 2015, 169, 335–343. [Google Scholar] [CrossRef]
  35. Chen, Y.; Qiu, Y.; Zhang, Z.; Zhang, J.; Chen, C.; Han, J.; Liu, D. Estimating salt content of vegetated soil at different depths with Sentinel-2 data. PeerJ 2020, 8, e10585. [Google Scholar] [CrossRef]
  36. Gitelson, A.A.; Kaufman, Y.J.; Merzlyak, M.N. Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ. 1996, 58, 289–298. [Google Scholar] [CrossRef]
  37. Rouse, J.W., Jr.; Haas, R.H.; Deering, D.W.; Schell, J.A.; Harlan, J.C. Monitoring the Vernal Advancement and Retrogradation (Green Wave Effect) of Natural Vegetation; NASA/GSFCT Type III Final Report; No. E75-10354; NASA: Greenbelt, MD, USA, 1974.
  38. Khalid, H.W.; Khalil, R.M.Z.; Qureshi, M.A. Evaluating spectral indices for water bodies extraction in western Tibetan Plateau. Egypt. J. Remote Sens. Space Sci. 2021, 24, 619–634. [Google Scholar]
  39. Marsett, R.C.; Qi, J.; Heilman, P.; Biedenbender, S.H.; Watson, M.C.; Amer, S.; Weltz, M.; Goodrich, D.; Marsett, R. Remote sensing for grassland management in the arid southwest. Rangel. Ecol. Manag. 2006, 59, 530–540. [Google Scholar] [CrossRef]
  40. Qi, J.; Chehbouni, A.; Huete, A.R.; Kerr, Y.H.; Sorooshian, S. A modified soil adjusted vegetation index. Remote Sens. Environ. 1994, 48, 119–126. [Google Scholar] [CrossRef]
  41. Yue, J.; Tian, J.; Tian, Q.; Xu, K.; Xu, N. Development of soil moisture indices from differences in water absorption between shortwave-infrared bands. ISPRS J. Photogramm. Remote Sens. 2019, 154, 216–230. [Google Scholar] [CrossRef]
  42. Jordan, C.F. Derivation of leaf-area index from quality of light on the forest floor. Ecology 1969, 50, 663–666. [Google Scholar] [CrossRef]
  43. Raya, S.S.; Singhb, J.P.; Dasa, G.; Panigrahyb, S. Use of high resolution remote sensing data for generating site-specific soil mangement plan. Red 2004, 550, 727. [Google Scholar]
  44. Amen, A.; Blaszczynski, J. Integrated Landscape Analysis; US Department of the Interior, Bureau of Land Management, National Science and Technology Center: Denver, CO, USA, 2001.
  45. Nield, S.J.; Boettinger, J.L.; Ramsey, R.D. Digitally mapping gypsic and natric soil areas using Landsat ETM data. Soil Sci. Soc. Am. J. 2007, 71, 245–252. [Google Scholar] [CrossRef] [Green Version]
  46. Xiao, J.; Shen, Y.; Tateishi, R.; Bayaer, W. Development of topsoil grain size index for monitoring desertification in arid land using remote sensing. Int. J. Remote Sens. 2006, 27, 2411–2422. [Google Scholar] [CrossRef]
  47. Triki Fourati, H.; Bouaziz, M.; Benzina, M.; Bouaziz, S. Modeling of soil salinity within a semi-arid region using spectral analysis. Arab. J. Geosci. 2015, 8, 11175–11182. [Google Scholar] [CrossRef]
  48. Rock, B.N.; Hoshizaki, T.; Miller, J.R. Comparison of in situ and airborne spectral measurements of the blue shift associated with forest decline. Remote Sens. Environ. 1988, 24, 109–127. [Google Scholar] [CrossRef]
  49. Allbed, A.; Kumar, L.; Aldakheel, Y.Y. Assessing soil salinity using soil salinity and vegetation indices derived from IKONOS high-spatial resolution imageries: Applications in a date palm dominated region. Geoderma 2014, 230, 1–8. [Google Scholar] [CrossRef]
  50. Escadafal, R.; Girard, M.C.; Courault, D. Munsell soil color and soil reflectance in the visible spectral bands of Landsat MSS and TM data. Remote Sens. Environ. 1989, 27, 37–46. [Google Scholar] [CrossRef]
  51. Williams, D. Landsat-7 Science Data User’s Handbook; National Aeronautics and Space Administration: Greenbelt, MD, USA, 2009; p. 186. [Google Scholar]
  52. Wilson, J.P.; Gallant, J.C. Digital Terrain Analysis. In Terrain Analysis: Principles and Applications; Wilson, J.P., Gallant, J.C., Eds.; John Wiley Sons: New York, NY, USA, 2000; pp. 1–27. [Google Scholar]
  53. Gallant, J.C.; Dowling, T.I. A multi resolution index of valley bottom flatness for mapping depositional areas. Water Resour. Res. 2003, 39, 1347–1360. [Google Scholar] [CrossRef]
  54. Moore, I.D.; Gessler, P.; Nielsen, G.; Peterson, G. Soil attribute prediction using terrain analysis. Soil Sci. Soc. Am. J. 1993, 57, 443–452. [Google Scholar] [CrossRef]
  55. Weiss, J.B.; Suyama, K.L.; Lee, H.H.; Scott, M.P. Jelly belly: A Drosophila LDL receptor repeat-containing signal required for mesoderm migration and differentiation. Cell 2001, 107, 387–398. [Google Scholar] [CrossRef] [Green Version]
  56. Riley, S.J.; DeGloria, S.D.; Elliot, R. Index that quantifies topographic heterogeneity. Intermt. J. Sci. 1999, 5, 23–27. [Google Scholar]
  57. Lu, G.Y.; Wong, D.W. An adaptive inverse-distance weighting spatial interpolation technique. Comput. Geosci. 2008, 34, 1044–1055. [Google Scholar] [CrossRef]
  58. Tuanmu, M.N.; Jetz, W. A global, remote sensing-based characterization of terrestrial habitat heterogeneity for biodiversity and ecosystem modelling. Glob. Ecol. Biogeogr. 2015, 24, 1329–1339. [Google Scholar] [CrossRef]
  59. Kim, J.H. Multicollinearity and misleading statistical results. Korean J. Anesthesiol. 2019, 72, 558–569. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  60. Pielou, E.C. Shannon’s formula as a measure of specific diversity: Its use and misuse. Am. Nat. 1966, 100, 463–465. [Google Scholar] [CrossRef]
  61. Simpson, E.H. Measurement of diversity. Nature 1949, 163, 688. [Google Scholar] [CrossRef]
  62. Hooper, D.U.; Solan, M.; Symstad, A.; Diaz, S.; Gessner, M.O.; Buchmann, N.; Degrange, V.; Grime, P.; Hulot, F.; Mermillod-Blondin, F.; et al. Species diversity, functional diversity and ecosystem functioning. In Biodiversity and Ecosystem Functioning: Synthesis and Perspectives; Oxford University Press: Oxford, UK, 2002; pp. 195–208. [Google Scholar]
  63. Pearson, K.X. Contributions to the mathematical theory of evolution.—II. Skew variation in homogeneous material. Philos. Trans. R. Soc. Lond. A 1895, 186, 343–414. [Google Scholar]
  64. Bowman, D.T. Contemporary issues. J. Cotton Sci. 2001, 5, 137–141. [Google Scholar]
  65. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef] [Green Version]
  66. Rahmanian, S.; Pourghasemi, H.R.; Pouyan, S.; Karami, S. Habitat potential modelling and mapping of Teucrium polium using machine learning techniques. Environmental Monitoring and Assessment 2021, 193, 1–21. [Google Scholar] [CrossRef]
  67. Nasiri, V.; Sadeghi, S.M.M.; Moradi, F.; Afshari, S.; Deljouei, A.; Griess, V.C.; Maftei, C.; Borz, S.A. The Influence of Data Density and Integration on Forest Canopy Cover Mapping Using Sentinel-1 and Sentinel-2 Time Series in Mediterranean Oak Forests. ISPRS Int. J. Geo-Inf. 2022, 11, 423. [Google Scholar] [CrossRef]
  68. Liaw, A.; Wiener, M. Classification and regression by randomForest. R News 2002, 2, 18–22. [Google Scholar]
  69. Kuhn, M. A Short Introduction to the caret Package. R Found Stat Comput 2015, 1, 1–10. [Google Scholar]
  70. Wei, P.; Lu, Z.; Song, J. Variable importance analysis: A comprehensive review. Reliab. Eng. Syst. Saf. 2015, 142, 399–432. [Google Scholar] [CrossRef]
  71. Uyanık, T.; Karatuğ, Ç.; Arslanoğlu, Y. Machine learning approach to ship fuel consumption: A case of container vessel. Transp. Res. Part D: Transp. Environ. 2020, 84, 102389. [Google Scholar] [CrossRef]
  72. Wang, W.; Lu, Y. Analysis of the mean absolute error (MAE) and the root mean square error (RMSE) in assessing rounding model. IOP Conf. Ser. Mater. Sci. Eng. 2018, 324, 012049. [Google Scholar] [CrossRef]
  73. Chrysafis, I.; Mallinis, G.; Tsakiri, M.; Patias, P. Evaluation of single-date and multi-seasonal spatial and spectral information of Sentinel-2 imagery to assess growing stock volume of a Mediterranean forest. Int. J. Appl. Earth Obs. Geoinf. 2019, 77, 1–14. [Google Scholar] [CrossRef]
  74. Chrysafis, I.; Korakis, G.; Kyriazopoulos, A.P.; Mallinis, G. Predicting Tree Species Diversity Using Geodiversity and Sentinel-2 Multi-Seasonal Spectral Information. Sustainability 2020, 12, 9250. [Google Scholar] [CrossRef]
  75. Senf, C.; Leitão, P.J.; Pflugmacher, D.; van der Linden, S.; Hostert, P. Mapping land cover in complex Mediterranean landscapes using Landsat: Improved classification accuracies from integrating multi-seasonal and synthetic imagery. Remote Sens. Environ. 2015, 156, 527–536. [Google Scholar] [CrossRef]
  76. Ehrlinger, J. ggRandomForests: Exploring random forest survival. arXiv 2016, arXiv:1612.08974. [Google Scholar]
  77. Wolter, P.T.; Mladenoff, D.J.; Host, G.E.; Crow, T.R. Using multi-temporal landsat imagery. Photogramm. Eng. Remote Sens 1995, 61, 1129–1143. [Google Scholar]
  78. Hill, R.A.; Wilson, A.K.; George, M.; Hinsley, S.A. Mapping tree species in temperate deciduous woodland using time-series multi-spectral data. Appl. Veg. Sci. 2010, 13, 86–99. [Google Scholar] [CrossRef]
  79. Kampouri, M.; Kolokoussis, P.; Argialas, D.; Karathanassi, V. Mapping of Forest Tree Distribution and Estimation of Forest Biodiversity using Sentinel-2 Imagery in the University Research Forest Taxiarchis in Chalkidiki, Greece. Geocarto Int. 2019, 34, 1273–1285. [Google Scholar] [CrossRef]
  80. Torresani, M.; Rocchini, D.; Sonnenschein, R.; Zebisch, M.; Marcantonio, M.; Ricotta, C.; Tonon, G. Estimating tree species diversity from space in an alpine conifer forest: The Rao’s Q diversity index meets the spectral variation hypothesis. Ecol. Inform. 2019, 52, 26–34. [Google Scholar] [CrossRef] [Green Version]
  81. Arekhi, M.; Yılmaz, O.Y.; Yılmaz, H.; Akyüz, Y.F. Can tree species diversity be assessed with Landsat data in a temperate forest? Environ. Monit. Assess. 2017, 189, 586. [Google Scholar] [CrossRef]
  82. Hernández-Stefanoni, J.L.; Gallardo-Cruz, J.A.; Meave, J.A.; Rocchini, D.; Bello-Pineda, J.; López-Martínez, J.O. Modeling α-and β-diversity in a tropical forest from remotely sensed and spatial data. Int. J. Appl. Earth Obs. Geoinf. 2012, 19, 359–368. [Google Scholar] [CrossRef]
  83. Hurlbert, A.H.; Haskell, J.P. The effect of energy and seasonality on avian species richness and community composition. Am. Nat. 2003, 161, 83–97. [Google Scholar] [CrossRef] [PubMed]
  84. Feeley, K.J.; Gillespie, T.W.; Terborgh, J.W. The Utility of Spectral Indices from Landsat ETM+ for Measuring the Structure and Composition of Tropical Dry Forests 1. Biotropica J. Biol. Conserv. 2005, 37, 508–519. [Google Scholar]
  85. Lassau, S.A.; Cassis, G.; Flemons, P.K.; Wilkie, L.; Hochuli, D.F. Using high-resolution multi-spectral imagery to estimate habitat complexity in open-canopy forests: Can we predict ant community patterns? Ecography 2005, 28, 495–504. [Google Scholar] [CrossRef]
  86. Prăvălie, R.; Sîrodoev, I.; Nita, I.A.; Patriche, C.; Dumitraşcu, M.; Roşca, B.; Tişcovschi, A.; Bandoc, G.; Săvulescu, I.; Mănoiu, V.; et al. NDVI-based ecological dynamics of forest vegetation and its relationship to climate change in Romania during 1987–2018. Ecol. Indic. 2022, 136, 108629. [Google Scholar] [CrossRef]
  87. Scherrer, D.; Körner, C. Topographically controlled thermal-habitat differentiation buffers alpine plant diversity against climate warming. J. Biogeogr. 2011, 38, 406–416. [Google Scholar] [CrossRef]
  88. Opedal, Ø.H.; Armbruster, W.S.; Graae, B.J. Linking small-scale topography with microclimate, plant species diversity and intra-specific trait variation in an alpine landscape. Plant Ecol. Divers. 2015, 8, 305–315. [Google Scholar] [CrossRef] [Green Version]
  89. Robinson, C.; Saatchi, S.; Clark, D.; Astaiza, J.H.; Hubel, A.F.; Gillespie, T.W. Topography and three-dimensional structure can estimate tree diversity along a tropical elevational gradient in Costa Rica. Remote Sens. 2018, 10, 629. [Google Scholar] [CrossRef]
  90. Read, Q.D.; Zarnetske, P.L.; Record, S.; Dahlin, K.M.; Costanza, J.K.; Finley, A.O.; Gaddis, K.D.; Grady, J.M.; Hobi, M.L.; Latimer, A.M.; et al. Beyond counts and averages: Relating geodiversity to dimensions of biodiversity. Glob. Ecol. Biogeogr. 2020, 29, 696–710. [Google Scholar] [CrossRef] [Green Version]
  91. Galvão, L.S.; Breunig, F.M.; Teles, T.S.; Gaida, W.; Balbinot, R. Investigation of terrain illumination effects on vegetation indices and VI-derived phenological metrics in subtropical deciduous forests. GIScience Remote Sens. 2016, 53, 360–381. [Google Scholar] [CrossRef]
  92. Goirán, S.B.; Aranibar, J.N.; Gomez, M.L. Heterogeneous spatial distribution of traditional livestock settlements and their effects on vegetation cover in arid groundwater coupled ecosystems in the Monte Desert (Argentina). J. Arid. Environ. 2012, 87, 188–197. [Google Scholar] [CrossRef]
  93. Hagen, M.; Kissling, W.D.; Rasmussen, C.; De Aguiar, M.A.M.; Brown, L.E.; Carstensen, D.W.; Alves-Dos-Santos, I.; Dupont, Y.L.; Edwards, F.K.; Genini, J.; et al. Biodiversity, species interactions and ecological networks in a fragmented world. In Advances in Ecological Research; Academic Press: Cambridge, MA, USA, 2012; Volume 46, pp. 89–210. [Google Scholar]
  94. Pettorelli, N.; Bühne, H.S.T.; Tulloch, A.; Dubois, G.; Macinnis-Ng, C.; Queirós, A.M.; Keith, D.A.; Wegmann, M.; Schrodt, F.; Stellmes, M.; et al. Satellite remote sensing of ecosystem functions: Opportunities, challenges and way forward. Remote Sens. Ecol. Conserv. 2018, 4, 71–93. [Google Scholar] [CrossRef]
  95. Turner, W.; Spector, S.; Gardiner, N.; Fladeland, M.; Sterling, E.; Steininger, M. Remote sensing for biodiversity science and conservation. Trends Ecol. Evol. 2003, 18, 306–314. [Google Scholar] [CrossRef]
Remotesensing 15 00387 g001 550
Figure 1. Geographic location of the study area used to collect species occurrence. Upper left panel—map of Iran showing the Fars province and the study area; lower right panel—study area showing the soil and vegetation plot samples.
Figure 1. Geographic location of the study area used to collect species occurrence. Upper left panel—map of Iran showing the Fars province and the study area; lower right panel—study area showing the soil and vegetation plot samples.
Remotesensing 15 00387 g001
Remotesensing 15 00387 g002 550
Figure 2. Datasets, metrics, and methodological steps used in this study.
Figure 2. Datasets, metrics, and methodological steps used in this study.
Remotesensing 15 00387 g002
Remotesensing 15 00387 g003a 550Remotesensing 15 00387 g003b 550
Figure 3. Spatial distribution of richness for February (a), April (b), and July (c); vegetation cover for February (d), April (e), and July (f); Shannon for February (g), April (h), and July (i); and Simpson for February (j), April (k), and July (l).
Figure 3. Spatial distribution of richness for February (a), April (b), and July (c); vegetation cover for February (d), April (e), and July (f); Shannon for February (g), April (h), and July (i); and Simpson for February (j), April (k), and July (l).
Remotesensing 15 00387 g003aRemotesensing 15 00387 g003b
Remotesensing 15 00387 g004 550
Figure 4. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in February for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Figure 4. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in February for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Remotesensing 15 00387 g004
Remotesensing 15 00387 g005 550
Figure 5. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in April for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Figure 5. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in April for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Remotesensing 15 00387 g005
Remotesensing 15 00387 g006 550
Figure 6. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in July for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Figure 6. Ten of the most important variables in modeling and predicting the Simpson’s, Shannon’s, and richness’ indexes, as well as vegetation cover in July for RF models, based on remotely sensed and geodiversity variables. Remotely sensed variables are shown in red, and geodiversity variables are shown in blue.
Remotesensing 15 00387 g006
Table
Table 1. List of the spectral and geodiversity indices used in this study.
Table 1. List of the spectral and geodiversity indices used in this study.
Sentinel-2 FeaturesFormulaeReference
Canopy Response Salinity Index (CRSI) √ ( ( B 8 A ) × B 4 − B 3 × B 2 ) / ( B 8 A × B 4 + B 3 × B 2 ) [34]
Enhanced Vegetation Index (EVI) 2.5 × ( B 8 A − B 4 ) / ( ( B 8 A + 6 × B 4 − 7.5 × B 2 + 1 ) ) [35]
Green-Red Vegetation Index (GRVI)B3 − B4/B3 + B4[21]
Green Normalized Difference Vegetation Index (GNDVI)B8 − B3/B8 + B3[36]
Normalized Difference Vegetation Index (NDVI)B8 − B4/B8 + B4[37]
Normalized Difference Vegetation Index red-edge 2 (NDVIre2)B8A − B6/B8A + B6[35]
Normalized Difference red-edge 1 (NDre1)B8 − B5/B8 + B5[35]
Normalized Multiband Drought Index (NMDI)(B8A − B11 − B12)/(B8A + B11 + B12)[35]
Normalized Difference Water Index (NDWI)B8A − B11/B8A + B11[38]
Soil-Adjusted Total Vegetation Index (SATVI) ( ( B 11 − B 4 ) ) / ( ( B 11 + B 4 + L ) ) × ( 1 + L ) − B 12 / 2 [39]
Modified Soil Adjusted Vegetation Index (MASVI) 2 B 8 + 1 − √ ( ( ( B 8 + 1 ) ^ 2 − 8 ( B 8 − B 4 ) ) / 2 ) [40]
Normalized Shortwave-infrared Difference SM Index 3 (NSDSI3)B11 − B12/B11 + B12[41]
Triangular Chlorophyll Index red-edge 1 (TCIrel)1.2 × (B5 − B3) − 1.5 × (B4 − B3) × ( B 5 / B 4 ) [35]
Transformed Vegetation Index (TVI)((B8 − B4/B8 + B4) + 0.5)0.5 ×   100 [21]
Vegetation (V)B8/B4[42]
Brightness Index (BI) x = √ ( 〖 B 4 〗 ^ 2 + 〖 〖 B 3 〗 ^ 2 + B 2 〗 ^ 2 ) / 3 [43]
Carbonate index (CaI)B4/B3[44]
Calcareous Sedimentary rocks index (CalcI)B11 − B3/B11 + B3[45]
Coloration index (CI)B4 − B3/B4 + B3[43]
Ferrous iron (FeI)B4/B11[44]
Geological response index (GeoI)B11 − B12/B11 + B12[45]
Grain size index (GSI)(B4 − B2)/(B4 + B3 + B2)[46]
Hue Index (HI)(2 × B4 − B3 − B2)/(B3 − B2)[43]
Intensity index 1 (Int1)B3 + B4/2[47]
Intensity index 2 (Int2)B3 + B4 + B8A/2[35]
Moisture Stress Index (MSI)B11/B8[48]
Salinity Index (S3)B3×B4/B2[49]
Saturation Index-T (SI-T)B4 − B2/B4 + B2[43]
Second Brightness Index (BI2) √ ( ( B 4 × B 4 ) + ( B 3 × B 3 ) + ( B 8 × B 8 ) / 3 ) [50]
Soil Moisture Monitoring Index (SMMI) √ ( 〖 B 8 A 〗 ^ 2 + B 〖 11 〗 ^ 2   ) / √ 2 [35]
Visible and Shortwave Drought Index (VSDI)1 − (B12, B4 − 2×B2)[35]
Landsat-8 FeaturesFormulasReference
Land Surface Temperature (LST)-[51]
Geodiversity variables
Altitude (ALT)Elevation above sea level[52]
Slope degree (SLP)Gradient of the slope steepness[53]
AspectDirection of the line of the steepest descent (r = width/height)[53]
Plan curvature (PLC)Rate of change of aspect along a contour[53]
Profile curvature (PRC)Rate of change of slope down a slope line[53]
Stream power index (SPI)A measure of the topographic control on the sediment transport (USLE’s LS factor)[54]
Multiresolution Index of the Ridge Top Flatnes (MRRTF)Measure of flatness and lowness[53]
Topographic position index (TPI)Difference between a cell elevation value and the average elevation of the neighborhood around that cell[55]
Topographic Ruggedness Index (TRI)The amount of elevation difference between adjacent cells of a digital elevation model[56]
Ultiresolution index of valley bottom flatness (MRVBF)Measure of flatness and lowness[53]
Mass Balance Index (MBI)Balance between soil mass deposited and eroded[51]
Mean annual rainfall (Rain)The monthly average rainfall
pHPotential of hydrogen: scale used to specify the acidity or basicity of an aqueous solution.
Electrical conductivity (EC)The total of soil anions and cations
Sand percentage (Sand)Share of sand in the soil particle size distribution
Clay percentage (Clay)Share of clay in the soil particle size distribution-
Silt percentage (Silt)Share of silt in the soil particle size distribution-
Soil organic carbon content (OC) Decayed plant residues and microorganisms
Soil organic matter content (OM)Detritus of plant and animal, soil microbes, and substances that soil microbes synthesize at various stages of decomposition.
Soil Nitrogen content (N)Both organic and inorganic forms of Nitrogen in the soil
Table
Table 2. Results of general linear models testing the effects of geodiversity and spectral variables on plant species diversity (Shannon H and Simpson), richness, and vegetation cover in the selected seasons.
Table 2. Results of general linear models testing the effects of geodiversity and spectral variables on plant species diversity (Shannon H and Simpson), richness, and vegetation cover in the selected seasons.
ShannonSimpson
FebruaryAprilJulyFebruaryAprilJuly
RR2p valueRR2p valueRR2p valueRR2p valueRR2p valueRR2p value
GNDVI0.150.020.080.270.110.0020.150.030.0580.060.010.130.260.080.0050.130.010.15
GRVI0.130.010.130.280.070.010.140.0250.090.180.030.060.250.050.0250.210.050.02
MASVI0.320.080.0060.480.18<0.0010.240.080.0070.340.130.0010.470.16<0.0010.370.14<0.001
Ndre10.150.020.090.360.15<0.0010.200.0460.0350.080.020.110.360.13<0.0010.160.020.116
NDVIre20.160.020.090.350.15<0.0010.190.0450.030.10.020.090.350.13<0.0010.160.010.126
SATVI0.240.060.010.440.17<0.0010.240.0840.0060.320.150.0020.470.18<0.0010.370.15<0.001
NDVI0.070.0060.460.280.060.020.190.040.0470.280.080.0080.350.110.0020.230.030.051
TCIre10.120.0020.370.340.13<0.0010.170.0070.210.140.010.150.380.14<0.0010.110.0080.21
VSDI0.140.010.120.330.13<0.0010.180.0160.140.120.040.040.330.120.0010.140.010.17
V0.070.0060.460.280.060.020.120.0160.140.280.080.0080.350.110.0020.160.010.16
TVI0.190.040.030.340.13<0.0010.130.0070.5120.130.030.070.370.14<0.0010.08−0.010.72
EVI0.006−0.010.780.190.020.090.120.0120.1680.250.070.010.320.080.0050.200.030.05
CRSI0.400.160.0010.490.19<0.0010.050.0130.800.320.140.0010.410.14<0.0010.13−0.010.83
RichnessVegetation cover
FebruaryAprilJulyFebruaryAprilJuly
RR2p valueRR2p valueRR2p valueRR2p valueRR2p valueRR2p value
GNDVI0.060.0010.340.100.020.130.100.0060.230.0050.010.640.040.0030.390.010.030.87
GRVI0.220.0490.030.330.120.0010.370.14<0.0010.110.070.20.160.040.030.080.0050.44
MASVI0.220.030.060.270.060.010.290.090.0050.060.010.610.060.0040.410.050.0080.52
Ndre10.100.0050.240.220.060.010.160.0230.100.050.010.40.090.0010.290.060.0080.52
NDVIre20.110.0050.230.210.060.010.170.0250.090.040.010.410.090.0020.270.070.0070.49
SATVI0.200.030.060.250.050.020.260.080.0080.090.030.510.060.0050.430.050.0070.51
NDVI0.150.020.090.210.060.140.170.030.060.110.040.930.040.0070.210.060.0130.97
TCIre10.120.0020.360.220.070.010.050.0090.570.040.010.50.070.0010.350.030.0090.57
VSDI0.090.0050.230.200.060.020.180.0160.140.040.020.420.100.0050.240.090.0030.38
V0.150.020.090.220.070.010.080.0020.270.010.010.930.040.0060.220.140.0080.52
TVI0.130.0040.250.210.050.030.020.0130.950.010.010.650.030.0080.540.220.030.07
EVI0.110.0030.270.150.020.110.120.0050.300.130.070.510.010.010.620.090.0070.50
CRSI0.210.030.060.240.050.030.260.0190.120.060.010.590.060.0040.420.180.0050.33
Table
Table 3. Accuracy assessment results of single-date and multi-seasonal RF models based on each of the geodiversity- or remotely sensed-only variables and their hyperparameter data.
Table 3. Accuracy assessment results of single-date and multi-seasonal RF models based on each of the geodiversity- or remotely sensed-only variables and their hyperparameter data.
Time Scale Multi-SeasonalApril
Diversity MetricsVariablesRMSER2MAETuning ParametersRMSER2MAETuning Parameters
RichnessAll0.0750.270.058mtry = 67; ntree = 6000.080.260.06mtry = 2; ntree = 500
Geodiversity0.0870.20.060mtry = 21; ntree = 5000.080.250.061mtry = 2; ntree = 500
Remotely sensed0.0910.200.062mtry = 49; ntree = 5500.080.190.062mtry = 2; ntree = 500
Vegetation coverAll0.1100.200.092mtry = 67; ntree = 5500.1120.170.085mtry = 32; ntree = 550
Geodiversity0.1190.190.089mtry = 11; ntree = 6000.1120.170.086mtry = 2; ntree = 500
Remotely sensed0.1170.190.089mtry = 49; ntree = 5500.110.210.085mtry = 2; ntree = 500
ShannonAll0.2900.280.237mtry = 34; ntree = 5000.3120.280.239mtry = 17; ntree = 550
Geodiversity0.2900.180.201mtry = 21; ntree = 5000.3380.260.27mtry = 10; ntree = 500
Remotely sensed0.3100.250.228mtry = 25; ntree = 5000.3680.200.293mtry = 2; ntree = 500
SimpsonAll0.0210.440.018mtry = 2; ntree = 5000.0230.430.019mtry = 2; ntree = 450
Geodiversity0.0200.400.017mtry = 2; ntree = 5000.0240.390.018mtry = 2; ntree = 450
Remotely sensed0.0200.260.019mtry = 4; ntree = 5000.0270.250.022mtry = 2; ntree = 450
February July
Diversity MetricsVariablesRMSER2MAETuning parametersRMSER2MAETuning parameters
RichnessAll0.0800.250.063mtry = 32; ntree = 5000.0830.220.064mtry = 9; ntree = 500
Geodiversity0.0850.240.063mtry = 10; ntree = 5000.0830.210.064mtry = 2; ntree = 450
Remotely sensed0.0800.180.063mtry = 2; ntree = 5000.0890.220.069mtry = 9; ntree = 450
Vegetation coverAll0.1180.170.093mtry = 2; ntree = 5000.1240.210.095mtry = 2; ntree = 500
Geodiversity0.1240.170.098mtry = 2; ntree = 4500.1270.200.094mtry = 2; ntree = 500
Remotely sensed0.1210.170.095mtry = 2; ntree = 4500.1220.200.095mtry = 2; ntree = 500
ShannonAll0.3300.270.248mtry = 28; ntree = 5000.3210.270.244mtry = 19; ntree = 500
Geodiversity0.3090.300.241mtry = 2; ntree = 4000.3250.270.243mtry = 2; ntree = 500
Remotely sensed0.4060.280.314mtry = 2; ntree = 4000.3360.250.264mtry = 2; ntree = 500
SimpsonAll0.0230.420.018mtry = 2; ntree = 5000.0250.380.021mtry = 2; ntree = 500
Geodiversity0.0230.530.017mtry = 2; ntree = 5000.0250.370.021mtry = 11; ntree = 500
Remotely sensed0.0230.510.018mtry = 2; ntree = 5000.0270.240.0226mtry = 9; ntree = 500
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Rahmanian, S.; Nasiri, V.; Amindin, A.; Karami, S.; Maleki, S.; Pouyan, S.; Borz, S.A. Prediction of Plant Diversity Using Multi-Seasonal Remotely Sensed and Geodiversity Data in a Mountainous Area. Remote Sens. 2023, 15, 387. https://doi.org/10.3390/rs15020387

AMA Style

Rahmanian S, Nasiri V, Amindin A, Karami S, Maleki S, Pouyan S, Borz SA. Prediction of Plant Diversity Using Multi-Seasonal Remotely Sensed and Geodiversity Data in a Mountainous Area. Remote Sensing. 2023; 15(2):387. https://doi.org/10.3390/rs15020387

Chicago/Turabian Style

Rahmanian, Soroor, Vahid Nasiri, Atiyeh Amindin, Sahar Karami, Sedigheh Maleki, Soheila Pouyan, and Stelian Alexandru Borz. 2023. "Prediction of Plant Diversity Using Multi-Seasonal Remotely Sensed and Geodiversity Data in a Mountainous Area" Remote Sensing 15, no. 2: 387. https://doi.org/10.3390/rs15020387

APA Style

Rahmanian, S., Nasiri, V., Amindin, A., Karami, S., Maleki, S., Pouyan, S., & Borz, S. A. (2023). Prediction of Plant Diversity Using Multi-Seasonal Remotely Sensed and Geodiversity Data in a Mountainous Area. Remote Sensing, 15(2), 387. https://doi.org/10.3390/rs15020387

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop