Next Article in Journal
The Impact of Digital Transformation on Supply Chain Capabilities and Supply Chain Competitive Performance
Previous Article in Journal
Assessment of Machining of Hastelloy Using WEDM by a Multi-Objective Approach
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 15
Issue 13
10.3390/su151310101
sustainability-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

Analysis of Tea Plantation Suitability Using Geostatistical and Machine Learning Techniques: A Case of Darjeeling Himalaya, India

by
Netrananda Sahu
1,*,
Pritiranjan Das
1,2,
Atul Saini
3,
Ayush Varun
1,
Suraj Kumar Mallick
4,
Rajiv Nayan
5,
S. P. Aggarwal
5,
Balaram Pani
6,
Ravi Kesharwani
1 and
Anil Kumar
1
1
Department of Geography, Delhi School of Economics, University of Delhi, New Delhi 110007, India
2
Department of Geography, Shaheed Bhagat Singh Evening College, University of Delhi, New Delhi 110017, India
3
Delhi School of Climate Change & Sustainability, Institution of Eminence, University of Delhi, New Delhi 110007, India
4
Department of Geography, Shaheed Bhagat Singh College, University of Delhi, New Delhi 110017, India
5
Department of Commerce, Ramanujan College, University of Delhi, New Delhi 110019, India
6
Department of Chemistry (Environmental Science), Bhaskarcharya College of Applied Science, University of Delhi, New Delhi 110075, India
*
Author to whom correspondence should be addressed.
Sustainability 2023, 15(13), 10101; https://doi.org/10.3390/su151310101
Submission received: 21 May 2023 / Revised: 20 June 2023 / Accepted: 22 June 2023 / Published: 26 June 2023
(This article belongs to the Topic Machine Learning and Big Data Analytics for Sustainability and Resilience)
Browse Figures
Review Reports Versions Notes

Abstract

:
This study aimed to identify suitable sites for tea cultivation using both random forest and logistic regression models. The study utilized 2770 sample points to map the tea plantation suitability zones (TPSZs), considering 12 important conditioning factors, such as temperature, rainfall, elevation, slope, soil depth, soil drainability, soil electrical conductivity, base saturation, soil texture, soil pH, the normalized difference vegetation index (NDVI), and land use land cover (LULC). The data were normalized using ArcGIS 10.2 and the models were calibrated using 70% of the total data, while the remaining 30% of the data were used for validation. The final TPSZ map was classified into four different categories: highly suitable zones, moderately suitable zones, marginally suitable zones, and not-suitable zones. The study revealed that the random forest (RF) model was more precise than the logistic regression model, with areas under the curve (AUCs) of 85.2% and 83.3%, respectively. The results indicated that well-drained soil with a pH range between 5.6 and 6.0 is ideal for tea farming, highlighting the importance of climate and soil properties in tea cultivation. Furthermore, the study emphasized the need to balance economic and environmental considerations when considering tea plantation expansion. The findings of this study provide important insights into tea cultivation site selection and can aid tea farmers, policymakers, and other stakeholders in making informed decisions regarding tea plantation expansion.

1. Introduction

Darjeeling tea (Camellia sinensis L.) is world-famous for its unique taste [1]. It was the first Indian product to obtain a geographical indication (GI) tag. The tea plantations of Darjeeling are the major contributors to the region’s economy. The expansion of the tea plantation areas is, therefore, essential for long-term socioeconomic sustainability [2]. The increased demand for Darjeeling tea and the limited land available for tea plantations have presented difficulties in recent times. Along with socioeconomic development, tea garden development plays a critical role in attaining self-sufficiency for the dependent population [3].
Land suitability analysis (LSA) is a technique for determining the inherent and potential capabilities, as well as the suitability, of various goals [4]. Geospatial data are successfully investigated using geographic information system (GIS)-based multi-criteria evaluation (MCE) methodologies to permit rigorous and flexible land suitability analysis, which include the analytical hierarchical process (AHP), the frequency ratio (FR), the weight of evidence (WoE), and the evidence belief function (EBF) [5]. Physical criteria such as geology, soil characteristics, geomorphology, atmospheric conditions, vegetation conditions, and economic and sociocultural circumstances are considered when evaluating a site’s suitability [6]. The integration of geospatial techniques with the multi-criteria decision-making (MCDM) method has been employed in numerous studies to effectively address intricate challenges pertaining to land management.
These techniques are extensively used in LSA to identify the potential lands for watershed management [7], plantation [3], and agriculture [8,9,10,11,12]. Jayansinghe et al. (2020) assessed the land suitability for tea crops in Sri Lanka using the AHP and the decision-making trial and evaluation laboratory (DEMATEL) model, wherein all the conditioning factors were integrated to generate a land suitability map [13]. In their study on sustainable tea-production-suitability areas in Bangladesh, Das et al. (2020) used phenological datasets from remote sensing and geospatial datasets of soil-plant biophysical properties, and based their findings on expert opinions [14]. The classification of the suitability zones was finally determined using a weighted overlay spatial analysis technique. This method included integrating reclassified raster layers for all important criteria, as well as the results of the multi-criteria analysis. In their study on tea suitability along the Laos–China border, Chanhda et al. (2010) used a combination of multi-criteria analysis and system dynamics techniques to analyze forest land utilization and land suitability, with the goal of anticipating future land use patterns for tea cultivation. [15]. Kamkar et al. (2014) opined that selecting locations for agricultural land use necessitates taking into account geophysical constraints, different topological conditions, and climatic conditions [16]. Accordingly, it is important to take into account current and accurate land use/land cover (LULC), as well as other geographical, environmental, and environmental thematic layers, when determining the best tea-plantation suitability.
New-age machine learning techniques have also been employed for land suitability assessment (LSA) studies. In their study on wheat production in Iran, Fereydoon et al. (2014) employed a two-class support vector machine (SVM) model, wherein 22 models of soil profile information were used, including relief, slope, precipitation, temperature, calcium carbonate content, organic carbon content, pH, and gypsum content [17]. The models were tested on a nonlinear class boundary that provided results of R = 0.84 and RMSE = 3.72. Dahikar and Rode (2014) demonstrated the employability of an artificial neural network (ANN) for ascertaining the cropland suitability zones [18]. They used attributes such as pH, potassium content, sulfur content, manganese content, iron content, soil depth, temperature, rainfall, and humidity. Mokarram (2015) employed several machine learning algorithms such as Bagging, adoptive boosting (AdaBoost), and RotForest for wheat land suitability analysis [19]. The thematic layers included in the study included topography, alkalinity, salinity, soil texture, calcium carbonate content, soil depth, pH, gypsum content, and wetness. The study found that the RotBoost algorithm provided the highest-accuracy score.
In their study on the suitability evaluation of tea cultivation in the Xinming Township in Huangshan City, Anhui Province, China, Xing et al. (2022) compared the following machine-learning-based models: logistic regression (LR), extreme-gradient boosting (XGBoost), AdaBoost, gradient-boosting decision tree (GBDT), random forest (RF), Gaussian Naïve Bayes (GNB), and multilayer perceptron (MLP). These models were employed as computational models to ascertain the precision of weight calculations for the evaluation factors [20]. Xing et al. selected twelve factors and, using the RF model, calculated the evaluation factors’ weights and obtained suitability evaluation results.
Wei and Zhou (2023) compared different machine-learning and deep-learning models for evaluating suitable areas for the best tea in Yunnan, China. For the computation of the evaluation variables, climate, terrain, and green-cropping system variables were used [21]. Six machine-learning models were used in the study. Among all the models, the FA + ResNet model demonstrated the best performance, with an accuracy score of 0.94 and a macro F1 score of 0.93.
In the current study, due to the study area’s great elevation variations and the abundance of natural resources, there is very little land that is accessible for tea-garden expansion. In land-suitability mapping, the priority is to identify suitable lands with the highest productivity and the lowest input. Therefore, the analysis of agricultural land’s suitability is a reliable and appropriate technique. The primary goal of the current study was to use a random forest and logistic regression model to examine the tea-plantation-suitability zones (TPSZs) in the Darjeeling district.

2. Study Area

The strategic geographical position of Darjeeling (Figure 1) holds significant importance, as it shares borders with Sikkim to the north, Bhutan to the east, and Nepal to the west. This configuration establishes Darjeeling as a crucial region, encompassing international boundaries and interstate border areas. Darjeeling tea is known for its unique flavor; it is often referred to as the “champagne of teas”.
Darjeeling’s climate is suitable for tea growth. The environment in the region is cool and moist, ideal for growing tea. The geographical extension of this area is 26°27′ N to 27°13′ N and 87°59′ E to 88°53′ E. The average annual temperature exhibits variation, ranging from 24 °C in the low-lying areas to less than 12 °C at higher elevations. In the summer season, temperatures on the ridge can reach approximately 16 °C to 17 °C, while during winter, they tend to drop to around 5 °C to 6 °C. The southern slope of the ridge receives much higher rainfall (4000–5000 mm) than the leeward sides (2000–2500 mm). Darjeeling tea is cultivated within the elevation range of 3000 to 7000 feet above sea level. This altitude contributes to the tea’s distinct fragrance and aroma. Darjeeling’s soil is rich in minerals, which contributes to the tea’s flavor. Darjeeling tea is traditionally farmed on tiny, family-owned tea estates. To assure the best quality, the tea is grown using traditional methods and the leaves are hand-picked.
Darjeeling tea is harvested four times a year, with each season yielding a unique flavor character. The first flush, which occurs in March and April, produces a light and delicate tea. The second flush, which occurs in May and June, produces a fuller-bodied tea with a musky aroma. The monsoon flush, which occurs in July and August, produces a tea with a robust flavor. The autumn flush, which occurs in October and November, produces a tea with a milder flavor. Overall, Darjeeling tea cultivation is a highly specialized process that requires skill and expertise [22]. The result is a tea that is highly prized and sought after by tea connoisseurs around the world.

3. Materials and Methods

3.1. Database

To carry out the TPSZ analysis, different types of data were used. The data were acquired from different sources (Table 1). The LULC map for 2022 was prepared using Landsat 8 OLI data in Google Earth Engine [23], using the support vector machine (SVM) method [24,25]. The LULC map was used for TPSZ analysis after validation by the Kappa co-efficient method [26,27]. Tea-plantation sample points were collected randomly using the LULC map and field surveys.

3.2. Conditioning Factors for Tea Cultivation

3.2.1. Rainfall

Rainfall is one of the most important factors that affect tea cultivation, as it plays a vital role in determining the growth and yield of tea plants. Tea plants require a specific amount of rainfall for optimal growth and yield, which varies depending on the stage of growth. During the vegetative stage, tea plants require regular and adequate rainfall, typically ranging from 1500–2500 mm annually, for the growth of new leaves and shoots. However, during the reproductive stage, tea plants require less rainfall, ranging from 1000–1500 mm annually (Figure 2A), to support the flowering and development of quality shoots [28,29]. Insufficient rainfall can lead to water stress, reduced photosynthesis, and poor growth and yield of tea plants. On the other hand, excessive rainfall can lead to waterlogging, soil erosion, and the spread of diseases, resulting in poor-quality tea leaves.

3.2.2. Temperature

Temperature is one of the most important factors that affect tea cultivation, as it affects the growth, development, and quality of tea plants. According to a study [30], tea plants have a specific temperature range for optimal growth and yield, which varies depending on the stage of growth. During the vegetative stage, temperatures ranging from 20–30 °C are ideal for the growth of tea plants. However, during the reproductive stage, temperatures ranging from 10–20 °C are considered ideal for flowering and the development of quality shoots (Figure 2B).
Extreme temperatures, either too high or too low, can have adverse effects on tea plants, resulting in stunted growth, reduced yield, and poor quality of tea leaves. For instance, temperatures above 35 °C can result in excessive transpiration, leading to water stress and reduced photosynthesis, while temperatures below 10 °C can result in frost damage [31].

3.2.3. Elevation

Elevation is another critical factor that affects tea cultivation, as it affects the growth, development, and quality of tea plants. Tea plants have a specific elevation range for optimal growth and yield, which varies depending on the climatic conditions of the region. Tea plants grow best at elevations ranging from 600 to 2000 m above sea level (Figure 2C), with the ideal elevation varying depending on the region [32]. Higher elevations provide cooler temperatures and lower humidity, which result in slower growth and better-quality tea leaves with more complex flavors and aromas.
On the other hand, lower elevations provide warmer temperatures and higher humidity, which result in faster growth and higher yields but may also result in lower-quality tea leaves with a less-complex flavor profile.

3.2.4. Slope

Slope is another important factor that affects tea cultivation, as it affects the water and nutrient availability of tea plants. Slope plays a significant role in tea cultivation, as it affects soil erosion, water-holding capacity, and nutrient availability. Gentle slopes (less than 20 degrees) are considered ideal for tea cultivation, as they promote water infiltration, reduce soil erosion, and provide sufficient water and nutrients to tea plants. In contrast, steep slopes (greater than 20 degrees) are not suitable for tea cultivation (Figure 2D), as they promote soil erosion, reduce water infiltration, and result in poor growth and yield of tea plants [33].
In addition, the slope affects the microclimate of a tea plantation, with south-facing slopes receiving more sunlight and heat, resulting in faster growth and higher yields. However, the south-facing slopes may also experience higher water stress, requiring irrigation to maintain optimal soil-moisture levels.

3.2.5. Soil Depth

Tea plants require well-drained soil with good water-holding capacity, and the depth of the soil determines the root system’s development and the availability of nutrients and water. The depth of the soil significantly influences the growth and yield of tea plants [34]. Tea plants grown in deeper soils have higher shoot growth, greater leaf area, and higher yield than plants grown in shallow soils (Figure 3A). Researchers have attributed this to the greater availability of soil moisture and nutrients in deeper soils, which promote better root development and plant growth. Soil depth significantly affects the accumulation of catechins and caffeine in tea leaves. Tea plants grown in deeper soils have higher concentrations of these compounds, which are important contributors to the taste and health benefits of tea [35].
Soil depth is crucial in tea cultivation, as it affects the growth, yield, and quality of tea plants. Tea farmers should consider the depth of the soil when selecting sites for tea cultivation, and implement appropriate soil management practices to ensure optimal plant growth and yield.

3.2.6. Soil Drainability

Soil drainability is an essential factor in tea cultivation, as it affects the growth and yield of plants [36]. Drainage is necessary for tea cultivation because tea plants require well-drained soil to thrive. Tea plants are susceptible to root rot and other soil-borne diseases if they are grown in poorly drained soils, which can ultimately reduce the yield and quality of the tea produced. Tea plants grown in well-drained soils (Figure 3B) show significantly better growth and yield than those grown in poorly drained soils. A previous study also found that soil drainage has a significant impact on the nutrient uptake of tea plants. Plants grown in well-drained soils have higher nutrient uptake and show better leaf quality than those grown in poorly drained soils [37]. Therefore, tea growers need to ensure that their soil has adequate drainage before planting tea.

3.2.7. Soil Electrical Conductivity (EC)

Soil electrical conductivity (EC) is an important parameter that affects the growth and development of tea plants. It is a measure of the ability of the soil to conduct electricity, which is influenced by factors such as soil texture, moisture content, and the presence of minerals and salts. Tea plants are particularly sensitive to changes in soil EC, as high levels of salinity can lead to reduced growth, lower yields, and poor quality of the harvested leaves [38]. High soil EC levels (Figure 3C) harm tea plant growth, particularly in terms of reducing shoot length and leaf area. Tea plants grown in soil with high EC have lower photosynthetic rates and chlorophyll content, which can ultimately lead to lower tea quality [39]. Farmers and growers need to carefully monitor soil EC levels and take appropriate measures to prevent excessive salinization of the soil, such as reducing the use of fertilizers and irrigation water with high salt content.

3.2.8. Soil Base Saturation

Soil base saturation is a crucial factor in tea cultivation, as it determines the availability of essential nutrients and influences soil pH. The base saturation refers to the relative allocation of basic cations within the cation exchange capacity (CEC) of the soil, such as calcium, magnesium, and potassium. In tea cultivation, a base saturation level of around 50–70% is considered ideal for optimal growth and productivity [40]. Soil base saturation (Figure 3D) significantly influences tea yield and quality. A previous study was conducted in the Darjeeling hills of India, where tea cultivation is a major economic activity. The researchers found that tea plants grown in soils with base saturation levels between 50–70% had higher leaf yield, better quality, and higher levels of important biochemical compounds such as catechins and caffeine. On the other hand, tea plants grown in soils with low base saturation levels (<50%) showed reduced growth, low yield, and poor quality.

3.2.9. Soil Texture

Soil texture is one of the most important factors affecting tea cultivation. The texture of the soil can greatly influence the growth, yield, and quality of tea plants. Tea plants require well-drained, deep soils that are rich in organic matter and have good water-holding capacity. The texture of soil also affects the availability of nutrients and the activity of soil microorganisms that are essential for plant growth. Tea plants grown in sandy loam soils have higher shoot growth and yield than those grown in clay soils [41]. The sandy loam soils (Figure 4A) also have higher levels of organic matter and available nitrogen, phosphorus, and potassium. Tea plants grown in sandy soils have higher levels of catechins and caffeine than those grown in clay soils. Therefore, soil with a sandy loam texture is considered ideal for tea cultivation, as it provides good drainage, good water-holding capacity, and good nutrient availability.

3.2.10. Soil pH

Soil pH is a critical factor in tea cultivation, as it influences the availability of essential nutrients to the tea plant, affecting its growth, yield, and quality. The significant effect of soil pH on the growth, yield, and quality of tea plants grown in soil with a pH range of 4.5 to 5.5 produces higher yields and better-quality tea than those grown in alkaline soils (Figure 4B) with a pH range of 6.0 to 7.0 [42]. Furthermore, a previous study suggested that maintaining optimal soil pH is crucial for sustainable tea production, as it not only improves plant growth and yield but also reduces the use of chemical fertilizers and pesticides, leading to a more environmentally friendly approach to tea cultivation.

3.2.11. Normalized Difference Vegetation Index (NDVI)

The normalized difference vegetation index (NDVI) is widely used for monitoring and assessing vegetation health, productivity, and stress (Figure 4C). Tea cultivation is a major agricultural activity in many countries, and the NDVI is a useful tool for monitoring and managing tea plantations. Several studies have highlighted the importance of the NDVI in tea cultivation [39]. For instance, a study conducted in China found that the NDVI can effectively monitor the spatial–temporal variability of tea growth and yield. Another study, conducted in India, showed that the NDVI can be used to assess the effect of different nutrient management practices on tea growth and yield.

3.2.12. Land Use/Land Cover (LULC)

Land use/land cover (LULC) mapping is an important tool for understanding the distribution and extent of different land cover types, including tea cultivation. Tea cultivation is a major land use in many countries and LULC mapping can provide insights into the spatial distribution of tea plantations, as well as their environmental and socioeconomic impacts. Several studies have highlighted the importance of LULC mapping in tea cultivation. LULC mapping (Figure 4D) can help to identify areas where tea plantations are expanding and encroaching on forested areas [43]. Another study conducted in India showed that LULC mapping can help to assess the impact of different land use types on tea productivity and quality [44].
LULC mapping is an important tool for tea farmers and researchers to understand the spatial distribution and impact of tea cultivation. Its applications in tea cultivation have been widely documented and it has been shown to have practical benefits in improving tea productivity and sustainability.

3.3. Methods

3.3.1. Logistic Regression Model (LR)

The logistic regression (LR) model is a statistical approach used to model the relationship between an independent variable and a dependent variable [45,46]. This technique is commonly used in classification problems, where the goal is to predict the probability of a binary outcome based on one or more input variables. It is a supervised learning method that is widely applied in various fields. Logistic regression is categorized into two main types: simple logistic regression and multivariable logistic regression. Simple logistic regression involves only one predictor variable, while multivariable logistic regression is applied when there are several predictors, including both categorical and continuous variables. Logistic regression is a popular technique for binary classification problems and is widely used in the field of statistical modelling [47].
Using a technique called maximum likelihood estimation, the logistic regression model is trained by identifying the coefficient values that maximize the probability of examining the training data. Once trained, the model may be used for predicting the likelihood of an outcome for new input data by simply entering the values of the input variables into the model equation and using the logistic function. The benefit of logistic regression is that, unlike traditional linear regression, where the variables must all have normal distributions, the variables can be continuous, discrete, or any combination of the two, and they are not required to have normal distributions. This is accomplished by adding an appropriate link function to the standard linear regression model. In our study, the binary dependent variable represents the presence or absence of a tea plantation. To analyze the possibility of tea plantation, the input value of the dependent variable must be either 0 or 1.
The equation below shows the relationship between the occurrence and its dependency on multiple variables:
p = 1 / 1 + e − z
where p is the probability of an event occurring and z is the linear combination. In our study, the value p is the estimated probability of a suitable tea plantation site. The value of p ranges from 0 to 1 on an S-shaped curve. Logical regression entails fitting an equation of the following kind to the data:
z = b 0 + b 1 x 1 + b 2 x 2 + b 3 x 3 + b n x n
where b 0 is the intercept of the model, the b i (0, 1, 2, …, n) are the slope coefficients of the LR model, and the x i (0, 1, 2, …, n) are the independent variables.
This model assessed the spatial relationship between tea cultivation samples and conditioning factors. The spatial databases of each factor were converted raster to ASCII format in ArcGIS 10.2 software for use in IBM SPSS V28 software. Therefore, logistic regression mathematical equations were used to formulate each factor and obtain the coefficient values (Table 1). Finally, the probability assessment of tea plantation occurrence was determined by analyzing spatial data via mathematical equations. (Equations (1) and (2)). In addition, logistic regression mathematical equations (Equations (1) and (3)) were used for all factors to generate an outcome of tea-plantation-suitability zones. The coefficient values of different conditioning factors were used for mapping tea-cultivation-suitability zones in ArcGIS 10.2, based on Equation (3) in the raster calculator (Figure 5).
z p = 0.001798 × R a i n f a l l + 0.000295 × T e m p e r a t u r e + − 0.07205 × E l e v a t i o n + 0.04237 × S l o p e + D e p t h c                             + D r a i n a b i l i t y c + E c c + S a t u r a t i o n c + T e x t u r e c + p H c + 0.024655 × N D V I + L U L C c − 17.640
where Depth is soil depth, Drainability is soil drainability, Ec is soil electrical conductivity, Saturation is base saturation, Texture is soil texture, pH is soil pH, and the coefficient values are listed in Table 2. A positive coefficient indicates that the tea plantation is more suitable at the level of a predictor than at the reference level of the factors. A negative coefficient indicates that the tea plantation is less suitable at the level of predictors than at the reference level of the factors (Table 2).

3.3.2. Random Forest (RF)

The random forest (RF) model is widely recognized as a prominent algorithm in the domain of supervised machine learning, suitable for addressing classification as well as regression tasks [48,49]. The algorithm constructs decision trees by leveraging multiple instances, employing majority voting for classification and averaging values for regression tasks. An important feature of RF is its ability to handle datasets with both continuous and categorical variables for both regression and classification tasks. RF has been shown to achieve excellent performance in classification problems. This model is also very useful for remote-sensing data classification [50]. The random forest approach [51] was developed by Leo Breiman; it is a compilation of unpruned decision trees designed for classification or regression purposes. These trees are built from randomly picked training data samples, and random features are chosen during the induction process. To make a prediction, the ensemble integrates the predictions of its members using a summing method. This typically involves taking the majority vote of the ensemble’s members for classification tasks, or the average prediction for regression tasks [52].
In this study, the tea-cultivation status of the study area was collected via the tea-cultivation and non-tea-cultivation field investigation and the remote sensing techniques. The tea-cultivation land value was denoted as 1 and the non-tea-cultivation land as 0 in ArcMap 10.2. The values from all variables were collected using the “multi values to point” algorithm. R statistical software provides limited tuning options for random forests [53,54], including “ntree”, “mytr”, “sampsize”, “nodsize”, and “maxnodes”. The “ntree’”parameter is used to specify the number of trees to be used in the algorithm. Using a proper number of trees can help to reduce errors, but using an excessive number of trees might be inefficient, specifically when dealing with large datasets. The “mytr” is the number of variables picked at random as candidates at each split. When performing classification, a vector called “sampsize” can be used to specify the number of samples to be drawn from each stratum during stratified sampling. The vector’s length should correspond to the number of strata, and each element represents the desired sample size for the corresponding stratum. In this study, “ntree” is 100 number of trees and the “mytr” variable sample split is 1 [55]. The data were split randomly, where 70% of the data were used as training data to calibrate the model and the remaining 30% of the data were used as testing data for model validation. Finally, we stacked all the variables as a raster layer and prepared it for the final output (Figure 5).

3.4. ROC (Receiver Operating Characteristic) Curve

The receiver operating characteristic (ROC) curve is a visual representation of a classification model’s performance across various classification thresholds. It plots the relationship between the true positive rate (TPR) (Equation (4)) and the false positive rate (FPR) (Equation (5)). TPR is computed as the ratio of true positives (TP) to the sum of true positives and false negatives (FN) using Equation (4), while FPR is calculated as the ratio of false positives (FP) to the sum of false positives and true negatives (TN) using Equation (5).
T P R = T P T P + F N
F P R = F P F P + T N
In these equations, TP represents acceptably classified positive cases, TN represents correctly classified negative cases, FP represents wrongly classified negative instances, and FN represents wrongly classified positive instances.
The area under the ROC curve (AUC) is a commonly used metric for evaluating the performance of a classification model. AUC values range between 0 and 1, where higher values show better performance. AUC values from 0.9 to 1 are considered outstanding, values from 0.8 to 0.9 are considered excellent, values from 0.7 to 0.8 are considered fair, values from 0.6 to 0.7 are considered poor, and values from 0.5 to 0.6 are considered failed. The AUC provides a comprehensive measure of the model’s capability to discriminate between positive and negative instances across different classification thresholds [56,57].

4. Results

The study aimed to map the tea-plantation-suitability zones (TPSZs) in a particular region and to compare the performance of two different models, random forest and logistic regression, in predicting the suitability of an area for tea plantation. In this section, we elaborate on the results obtained in the study, providing a comprehensive analysis of the study findings. To begin with, the study collected data on various factors that are essential for tea plantations, including topography, climatic conditions, soil characteristics, and land use. There are six factors that play crucial roles, i.e., Lulc, NDVI, elevation, temperature, slope. and soil pH, for this tea-plantation-suitability-zone analysis (Figure S1). The data were collected from different sources, including satellite imagery, ground surveys, and meteorological data records. The collected data were then preprocessed to ensure that they were clean, accurate, and ready for analysis. To ensure that the random forest and logistic regression models were accurately calibrated, all factors were adjusted to the same scale and extension. The layers were prepared using ArcGIS 10.2, and the random forest model utilized R-4.3.0 statistical software, while the logistic regression model used IBM SPSS V28 software.
The study involved using the calibrated models to prepare a map of the tea-plantation-suitability zones (TPSZs). The TPSZs were divided into four categories—highly suitable zones, moderately suitable zones, marginally suitable zones, and not-suitable zones (Figure 6)—based on the predicted suitability values. The study involved using the calibrated models to map the TPSZs in the study area.

4.1. Tea-Plantation-Suitability Zones (TPSZs)

It was found that the total area covered by the TPSZs was 3149 km2. The TPSZs were divided into four categories—highly suitable zones, moderately suitable zones, marginally suitable zones, and not-suitable zones (Figure 6)—based on the predicted suitability values. The highly suitable zones covered 17.59% (553.84 km2) for the random forest model and 20.87% (657.14 km2) for the logistic regression model. The moderately suitable zones covered 16.59% (522.29 km2) for the random forest model and 28.74% (905.11 km2) for the logistic regression model. The marginally suitable zones covered 13.35% (420.47 km2) for the random forest model and 15.80% (497.58 km2) for the logistic regression model. The not-suitable zones covered 52.47% (1652.40 km2) for the random forest model and 34.59% (1089.17 km2) for the logistic regression model (Figure 7).
The study also revealed that the highly suitable and moderately suitable zones covered a significant proportion of the TPSZs, indicating that these areas have high potential for tea plantation. These zones could be prioritized for tea-plantation development, as they are likely to yield higher productivity and profitability. On the other hand, the marginally suitable and not-suitable zones may not be suitable for tea plantation development and, therefore, alternative land use options should be explored in these areas.

4.2. Accuracy Assessment of Models

The sample data were divided into two sets: 70% for model calibration and 30% for model validation. The models were calibrated using the training data and the accuracy of the models was assessed using the testing data. The performance of the models was measured using the area under the c(AUC) of the receiver operating characteristic (ROC) curve, which is a standard metric for assessing the performance of binary classifiers. The results showed that both models were accurate in predicting the suitability of an area for tea plantation, with AUC values of 85.2% and 83.3% for the random forest and logistic regression models, respectively. The Kappa value was 0.6916 and the confusion matrix (see the Supplementary Information for confusion statistics) showed that in the predicted map of tea plantation, positive occurrence points accurately showed 650 points for tea plantation and 865 points for no-tea plantation and negative occurrence points showed 197 points for tea plantation and 76 points for no-tea plantation (Figure 8). However, the random forest model performed better than the logistic regression model in terms of accuracy and precision, as indicated by the higher AUC value (Figure 8).

4.3. Intercomparison of Models for Tea-Plantation Suitability

This study’s findings provide valuable insights into the suitability of an area for tea plantation and the performance of two different models in predicting such suitability. The results showed that both models were accurate in predicting the suitability of an area for tea plantation, but the random forest model performed better than the logistic regression model. The ROC curve serves as validation for the suitability maps, and our findings indicate that the random forest model yielded better results than the logistic regression model, with an improvement in accuracy of 1.9%. The comparatively good performance of random forest could be attributed to the fact that random forest is an ensemble method that combines multiple decision trees, thereby improving the accuracy and robustness of the model. The study highlights the high suitability of non-forest areas for tea plantations but emphasizes the need to consider the potential positive and negative impacts on both the environment and the economy.
This study’s results also highlight the importance of data preprocessing and calibration in ensuring the accuracy and reliability of the models. The collected data were preprocessed to ensure that they were clean, accurate, and ready for analysis. The models were then calibrated using the calibration dataset, which helped to optimize the models’ parameters and improve their performances. The validation dataset was used to assess the accuracy of the models, and the AUC values showed that the models were accurate in predicting the suitability of an area for tea plantation.

5. Discussion

In this study, we utilized both random forest and logistic regression models to identify suitable sites for tea plantations. We conducted extensive mapping of 2770 tea plantation locations using a combination of satellite imagery and field surveys. Both the random forest and logistic regression models were employed to analyze the TPSZs, using the aforementioned 12 factors. Before the analysis, we standardized the cell size and row column to ensure an equal number of data points. We further normalized the data using ArcGIS 10.2. The discrete data layer was subsequently classified based on the tea plantation site, and both models adopted a grid-based analysis methodology.
The climate plays a pivotal role in influencing the distribution, growth, and yield of tea plantations. Optimal tea cultivation occurs in warm and moist climatic conditions, characterized by a consistent temperature throughout the year. Adequate and well-distributed annual rainfall, typically ranging from 1500 to 2000 mm, is necessary for successful tea cultivation. However, extended dry seasons, excessively low temperatures (below 17.5 °C), or extremely high temperatures (above 26.5 °C) can have detrimental effects on tea plant foliage in the Darjeeling Himalaya [58]. Consequently, tea cultivation has been facing serious issues posed by inadequate water supply and large dam construction [2]. Furthermore, the increasing intensity of temperature also impacts crucial biological processes such as photosynthesis and reproductive growth in tea plantations [30,31].
Soil characteristics play a critical role in tea plantations. The physical, chemical, and biological properties of the soil collectively contribute to the success of tea plantations. Notably, soil texture and pH emerged as the most significant physicochemical factors, with an acidic soil pH ranging between 5.6 and 6.0 being deemed ideal for tea cultivation in the study area. However, in recent decades, increasing soil pH along with sediment-based soil texture has created a barrier for the quality of tea plantation in the low-elevated areas of northwestern Darjeeling Himalaya. Moreover, base saturation (i.e., Ca, Mg, K, and Na) has been found to be less in the northeast section and, consequently, the quality of tea production was also somewhat damaged. Similar kinds of results have been found in Bangladesh tea gardens [40] and in Malaysia [59,60]. Regarding drainability, soils with a sandy or coarser grain texture exhibit superior permeability and optimal drainability; they are found in the low-elevated areas. except in the northwestern and northeastern sections. Therefore, comparatively low-elevated but gentle-to-moderate slope areas provide optimal conditions for tea growth, and the northwestern and northeastern sections are comparatively less productive for tea growth.
However, the current land-use scenario—specifically, agriculture and vegetation status—were determined where are favorable conditions for tea growth [2]. The analysis revealed that the dominant LULC—specifically, agricultural land and vegetation cover—have sometimes created a problem in tea plantations. Dense forest cover or intensive agriculture activities near tea gardens consume more water, and recharge in the ground creates a problem for the growth of tea.
This study has some limitations. The RF model tends to overfit the training data if not properly tuned. Overfitting can lead to misleading predictions and reduced model performance. Both the RF and LR models require a sufficient amount of high-quality data for accurate prediction. The RF and LR models are best suited for site suitability analysis and can be used in different studies, such as studies of flood susceptibility, landslide susceptibility, forest fire susceptibility, landfill suitability, and urban built-up suitability [61].
This study was able to identify substantial potential areas for expanding tea plantations in Darjeeling Himalaya, based on suitable surface and environmental conditions such as slope, rainfall, and temperature. Moreover, the soil conditions in the study area were found to be conducive to tea plantations.

6. Conclusions

This study employed random forest (RF) and logistic regression (LR) models to assess the suitability of tea-plantation sites based on several factors, including temperature, rainfall, elevation, soil characteristics, and land use/land cover. The models were calibrated and validated using standardized data, and GIS techniques were employed to map tea-plantation-suitability zones (TPSZs).
The study found that both the RF and LR models were effective in classifying TPSZs into four categories: highly suitable zones, moderately suitable zones, marginally suitable zones, and not-suitable zones. The total area covered by the TPSZs was 3149 km2. The RF model identified 17.59% of the total area as highly suitable, 16.59% as moderately suitable, 13.35% as marginally suitable, and 52.47% as not suitable. The LR model identified 20.87% of the total area as highly suitable, 28.74% as moderately suitable, 15.80% as marginally suitable, and 34.59% as not suitable. The study used 2770 sample points for TPSZ mapping, with 70% of the locations used for model calibration and 30% for validation. The area under the curve (AUC) for the RF model was 85.2%, while for the LR model, it was 83.3%. These high AUC values indicated the accuracy and justification of the models in predicting tea-plantation potential.
The study found that climate factors such as temperature and rainfall played a crucial role in determining tea-plantation suitability. Soil characteristics, including soil depth, texture, pH, drainability, electrical conductivity, and base saturation, were also significant determinants. The analysis of land use/land cover using the NDVI and LULC indices highlighted the dominant presence of forest cover and tea plantations in the study area. The RF model outperformed the LR model, demonstrating a 1.9% improvement in accuracy. The suitability maps generated by the RF model provided valuable insights into tea-plantation potential, while considering the balance between environmental and economic impacts. This study recommends the protection of natural reserve areas, while promoting tea plantation expansion in non-forest regions.
In summary, the RF and LR models facilitated the identification and classification of suitable tea-plantation sites based on multiple factors. The study’s findings can assist stakeholders in making informed decisions regarding tea-plantation site selection and management, while considering environmental and economic aspects.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/su151310101/s1, Figure S1: Importance of different factors for Tea plantation suitability analysis.

Author Contributions

Conceptualization, N.S.; data curation, P.D., A.S. and A.V.; formal analysis, N.S., P.D. and A.V.; funding acquisition., N.S., A.S. and S.K.M.; methodology, N.S. and P.D.; supervision, N.S. and R.N.; visualization, N.S., P.D., A.V., R.K. and A.K.; validation, P.D.; writing—original draft, P.D., A.V. and A.S.; project administration, N.S. and R.N.; writing—review and editing, N.S., A.S.; S.K.M., R.N., S.P.A. and B.P. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the Teaching Learning Centre, Ramanujan College, New Delhi, India, (project code: MP02/2022-23). The mentioned project is funded under the PMMMNMTT scheme, Department of Higher Education, Ministry of Education, Government of India.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Publicly available datasets were analyzed in this study. Analyzed data are available on request from the corresponding author.

Acknowledgments

The authors acknowledge the data shared by the Climate Research Unit, University of East Anglia; Google Earth Engine; R Project; BHUVAN, the Indian Space Research Organisation; the International Soil Reference and Information Centre; and the National Aeronautics and Space Administration, USA. We are thankful to the Teaching Learning Centre, Ramanujan College, New Delhi, India, for the major project (project code: MP02/2022-23). The mentioned project is funded under the PMMMNMTT scheme, Department of Higher Education, Ministry of Education, Government of India.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Tea Board of India. 68th Annual Report 2021–2022. Available online: https://www.vegetableindia.com/AR/2021-2022.pdf (accessed on 12 January 2023).
  2. Pramanik, M.K. Site Suitability Analysis for Agricultural Land Use of Darjeeling District Using AHP and GIS Techniques. Model. Earth Syst. Environ. 2016, 2, 56. [Google Scholar] [CrossRef] [Green Version]
  3. Zolekar, R.; Bhagat, V. Use of IRS P6 LISS-IV Data for Land Suitability Analysis for Cashew Plantation in Hilly Zone. Asian J. Geoinform. 2014, 14, 23–35. [Google Scholar]
  4. A Framework for Land Evaluation. FAO Soils Bulletin 52, 1st ed.; FAO and Agriculture Organization of the United Nations: Rome, Italy, 1976. [Google Scholar]
  5. Mendas, A.; Delali, A. Integration of MultiCriteria Decision Analysis in GIS to Develop Land Suitability for Agriculture: Application to Durum Wheat Cultivation in the Region of Mleta in Algeria. Comput. Electron. Agric. 2012, 83, 117–126. [Google Scholar] [CrossRef]
  6. Mokarram, M.; Aminzadeh, F. Gis-Based Multicriteria Land Suitability Evaluation Using Ordered Weight Averaging with Fuzzy Quantifier: A Case Study in Shavur Plain, Iran. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. ISPRS Arch. 2010, 38, 508–512. [Google Scholar]
  7. Steiner, F.; McSherry, L.; Cohen, J. Land Suitability Analysis for the Upper Gila River Watershed. Landsc. Urban Plan. 2000, 50, 199–214. [Google Scholar] [CrossRef]
  8. Bandyopadhyay, S.; Jaiswal, R.K.; Hegde, V.S.; Jayaraman, V. Assessment of Land Suitability Potentials for Agriculture Using a Remote Sensing and GIS Based Approach. Int. J. Remote Sens. 2009, 30, 879–895. [Google Scholar] [CrossRef]
  9. Jafari, S.; Zaredar, N. Land Suitability Analysis Using Multi Attribute Decision Making Approach. Int. J. Environ. Sci. Dev. 2010, 1, 441–445. [Google Scholar] [CrossRef]
  10. Chandio, I.A.; Matori, A.-N.; Lawal, D.U.; Sabri, S. GIS—Based Land Suitability Analysis Using AHP for Public Parks Planning in Larkana City. Mod. Appl. Sci. 2011, 5, 177–189. [Google Scholar] [CrossRef] [Green Version]
  11. Akinremi, O.O.; McGinn, S.M.; Cutforth, H.W. Seasonal and Spatial Patterns of Rainfall Trends on the Canadian Prairies. J. Clim. 2001, 14, 2177–2182. [Google Scholar] [CrossRef]
  12. Akıncı, H.; Özalp, A.Y.; Turgut, B. Agricultural Land Use Suitability Analysis Using GIS and AHP Technique. Comput. Electron. Agric. 2013, 97, 71–82. [Google Scholar] [CrossRef]
  13. Jayasinghe, S.L.; Kumar, L.; Hasan, M.K. Relationship between Environmental Covariates and Ceylon Tea Cultivation in Sri Lanka. Agronomy 2020, 10, 476. [Google Scholar] [CrossRef] [Green Version]
  14. Das, A.C.; Noguchi, R.; Ahamed, T. Integrating an Expert System, GIS, and Satellite Remote Sensing to Evaluate Land Suitability for Sustainable Tea Production in Bangladesh. Remote Sens. 2020, 12, 4136. [Google Scholar] [CrossRef]
  15. Chanhda, H.; Wu, C.; Ye, Y.; Ayumi, Y. GIS Based Land Suitability Assessment along Laos-China Border. J. For. Res. 2010, 21, 343–349. [Google Scholar] [CrossRef]
  16. Kamkar, B.; Dorri, M.A.; Teixeira da Silva, J.A. Assessment of Land Suitability and the Possibility and Performance of a Canola (Brassica Napus L.)—Soybean (Glycine Max L.) Rotation in Four Basins of Golestan Province, Iran. Egypt. J. Remote Sens. Sp. Sci. 2014, 17, 95–104. [Google Scholar] [CrossRef] [Green Version]
  17. Sarmadian, F.; Keshavarzi, A.; Rooien, A.; Zahedi, G.; Javadikia, H. Support Vector Machines Based-Modeling of Land Suitability Analysis for Rainfed Agriculture. J. Geosci. Geomat. 2014, 2, 165–171. [Google Scholar]
  18. Dahikar, S.S.; Rode, S.V. Agricultural Crop Yield Prediction Using Artificial Neural Network Approach Miss. Int. J. Innov. Res. Electr. Electron. Instrum. Control. Eng. 2014, 2, 683–686. [Google Scholar]
  19. Mokarram, M.; Hamzeh, S.; Aminzadeh, F.; Zarei, A. Using Machine Learning for Land Suitability Classification. West Afr. J. Appl. Ecol. 2015, 23, 63–73. [Google Scholar]
  20. Xing, W.; Zhou, C.; Li, J.; Wang, W.; He, J.; Tu, Y.; Cao, X.; Zhang, Y. Suitability Evaluation of Tea Cultivation Using Machine Learning Technique at Town and Village Scales. Agronomy 2022, 12, 2010. [Google Scholar] [CrossRef]
  21. Wei, G.; Zhou, R. Comparison of Machine Learning and Deep Learning Models for Evaluating Suitable Areas for Premium Teas in Yunnan, China. PLoS ONE 2023, 18, e0282105. [Google Scholar] [CrossRef]
  22. Kumar, R. An Introduction to Cultivation of Darjeeling Tea (Camellia sinensis L.). Farming Manag. 2018, 3, 66–79. [Google Scholar] [CrossRef]
  23. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-Scale Geospatial Analysis for Everyone. Remote Sens. Environ. 2017, 202, 18–27. [Google Scholar] [CrossRef]
  24. Dihkan, M.; Guneroglu, N.; Karsli, F.; Guneroglu, A. Remote Sensing of Tea Plantations Using an SVM Classifier and Pattern-Based Accuracy Assessment Technique. Int. J. Remote Sens. 2013, 34, 8549–8565. [Google Scholar] [CrossRef]
  25. Shi, D.; Yang, X. Support Vector Machines for Land Cover Mapping from Remote Sensor Imagery. In Monitoring and Modeling of Global Changes: A Geomatics Perspective; Li, J., Yang, X., Eds.; Springer: Dordrecht, The Netherlands, 2015; pp. 265–279. ISBN 978-94-017-9813-6. [Google Scholar]
  26. Mallick, S.K.; Das, P.; Maity, B.; Rudra, S.; Pramanik, M.; Pradhan, B.; Sahana, M. Understanding Future Urban Growth, Urban Resilience and Sustainable Development of Small Cities Using Prediction-Adaptation-Resilience (PAR) Approach. Sustain. Cities Soc. 2021, 74, 103196. [Google Scholar] [CrossRef]
  27. Sim, J.; Wright, C.C. The Kappa Statistic in Reliability Studies: Use, Interpretation, and Sample Size Requirements. Phys. Ther. 2005, 85, 257–268. [Google Scholar] [CrossRef] [Green Version]
  28. Duncan, J.M.A.; Saikia, S.D.; Gupta, N.; Biggs, E.M. Observing Climate Impacts on Tea Yield in Assam, India. Appl. Geogr. 2016, 77, 64–71. [Google Scholar] [CrossRef]
  29. Deka, N.; Goswami, K.; Anurupam, K. What Will Drive the Small Tea Growers towards Environment-Friendly Cultivation? Implications from the Tea Sector in Assam, India. Clim. Dev. 2022, 14, 443–458. [Google Scholar] [CrossRef]
  30. Chiu, W. Factors Affecting the Production and Quality of Partially Fermented Tea in Taiwan. Acta Hortic. 1990, 57–64. [Google Scholar] [CrossRef]
  31. Yan, Y.; Jeong, S.; Park, C.-E.; Mueller, N.D.; Piao, S.; Park, H.; Joo, J.; Chen, X.; Wang, X.; Liu, J.; et al. Effects of Extreme Temperature on China’s Tea Production. Environ. Res. Lett. 2021, 16, 044040. [Google Scholar] [CrossRef]
  32. Chen, P.; Cai, J.; Zheng, P.; Yuan, Y.; Tsewang, W.; Chen, Y.; Xiao, X.; Liao, J.; Sun, B.; Liu, S. Quantitatively Unravelling the Impact of High Altitude on Oolong Tea Flavor from Camellia Sinensis Grown on the Plateaus of Tibet. Horticulturae 2022, 8, 539. [Google Scholar] [CrossRef]
  33. Rahaman, S.A.; Aruchamy, S. Land Suitability Evaluation of Tea (Camellia sinensis L.) Plantation in Kallar Watershed of Nilgiri Bioreserve, India. Geographies 2022, 2, 701–723. [Google Scholar] [CrossRef]
  34. Yüksek, T.; Yüksek, F. Effects of Altitude, Aspect, and Soil Depth on Carbon Stocks and Properties of Soils in a Tea Plantation in the Humid Black Sea Region. Land Degrad. Dev. 2021, 32, 4267–4276. [Google Scholar] [CrossRef]
  35. Venkatesan, S.; Murugesan, S. Influence of Tea Cultivation on Soil Characteristics with Special Reference to Potassium. Int. J. Soil Sci. 2005, 1, 58–63. [Google Scholar] [CrossRef] [Green Version]
  36. Nath, A.J.; Kumar, R.; Devi, N.B.; Rocky, P.; Giri, K.; Sahoo, U.K.; Bajpai, R.K.; Sahu, N.; Pandey, R. Agroforestry Land Suitability Analysis in the Eastern Indian Himalayan Region. Environ. Chall. 2021, 4, 100199. [Google Scholar] [CrossRef]
  37. Arafat, Y.; Ud Din, I.; Tayyab, M.; Jiang, Y.; Chen, T.; Cai, Z.; Zhao, H.; Lin, X.; Lin, W.; Lin, S. Soil Sickness in Aged Tea Plantation Is Associated With a Shift in Microbial Communities as a Result of Plant Polyphenol Accumulation in the Tea Gardens. Front. Plant Sci. 2020, 11, 601. [Google Scholar] [CrossRef]
  38. Di, S.; Zong, M.; Li, S.; Li, H.; Duan, C.; Peng, C.; Zhao, Y.; Bai, J.; Lin, C.; Feng, Y.; et al. The Effects of the Soil Environment on Soil Organic Carbon in Tea Plantations in Xishuangbanna, Southwestern China. Agric. Ecosyst. Environ. 2020, 297, 106951. [Google Scholar] [CrossRef]
  39. Huang, Y.; Jiang, H.; Wang, W. Research on Tea Tree Growth Monitoring Model Using Soil Information. Plants 2022, 11, 262. [Google Scholar] [CrossRef]
  40. Adhikary, S.; Khan, M.Z.; Arobe, S.; Dey, S.; Billah, S.M. Soil Chemical Analysis of Kazi and Kazi Organic Tea Garden and Compared to Ordinary Tea Gardens of Bangladesh. Open J. Soil Sci. 2019, 9, 91–102. [Google Scholar] [CrossRef] [Green Version]
  41. Huu Chien, H.; Tokuda, M.; Van Minh, D.; Kang, Y.; Iwasaki, K.; Tanaka, S. Soil Physicochemical Properties in a High-Quality Tea Production Area of Thai Nguyen Province in Northern Region, Vietnam. Soil Sci. Plant Nutr. 2019, 65, 73–81. [Google Scholar] [CrossRef] [Green Version]
  42. Li, S.; Li, H.; Yang, C.; Wang, Y.; Xue, H.; Niu, Y. Rates of Soil Acidification in Tea Plantations and Possible Causes. Agric. Ecosyst. Environ. 2016, 233, 60–66. [Google Scholar] [CrossRef]
  43. Prokop, P. Tea Plantations as a Driving Force of Long-Term Land Use and Population Changes in the Eastern Himalayan Piedmont. Land Use Policy 2018, 77, 51–62. [Google Scholar] [CrossRef]
  44. Begum, M.; Pramanick, N.; Mitra, D.; Chanda, A.; Hazra, S.; Mukhopadhyay, A. The Journey of Darjeeling Tea Gardens Over Decades in the Eyes of the Satellite. In Handbook of Climate Change Across the Food Supply Chain; Leal Filho, W., Djekic, I., Smetana, S., Kovaleva, M., Eds.; Springer International Publishing: Cham, Switzerland, 2022; pp. 21–37. [Google Scholar]
  45. Cox, D.R. The Regression Analysis of Binary Sequences. J. R. Stat. Soc. Ser. B 1958, 20, 215–232. [Google Scholar] [CrossRef]
  46. Chapman, S.J. Review of Discovering Statistics Using IBM SPSS Statistics. J. Polit. Sci. Educ. 2018, 14, 145–147. [Google Scholar] [CrossRef]
  47. Lee, S.; Pradhan, B. Landslide Hazard Mapping at Selangor, Malaysia Using Frequency Ratio and Logistic Regression Models. Landslides 2007, 4, 33–41. [Google Scholar] [CrossRef]
  48. Saha, S.; Mondal, P. Estimation of the Effectiveness of Multi-Criteria Decision Analysis and Machine Learning Approaches for Agricultural Land Capability in Gangarampur Subdivision, Eastern India. Artif. Intell. Geosci. 2022, 3, 179–191. [Google Scholar] [CrossRef]
  49. Park, Y.-S.; Chung, Y.-J.; Moon, Y.-S. Hazard Ratings of Pine Forests to a Pine Wilt Disease at Two Spatial Scales (Individual Trees and Stands) Using Self-Organizing Map and Random Forest. Ecol. Inform. 2013, 13, 40–46. [Google Scholar] [CrossRef]
  50. Pal, M. Random Forest Classifier for Remote Sensing Classification. Int. J. Remote Sens. 2005, 26, 217–222. [Google Scholar] [CrossRef]
  51. Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef] [Green Version]
  52. Speiser, J.L.; Miller, M.E.; Tooze, J.; Ip, E. A Comparison of Random Forest Variable Selection Methods for Classification Prediction Modeling. Expert Syst. Appl. 2019, 134, 93–101. [Google Scholar] [CrossRef]
  53. Genuer, R.; Poggi, J.-M.; Tuleau-Malot, C. VSURF: An R Package for Variable Selection Using Random Forests. R J. 2015, 7, 19–33. [Google Scholar] [CrossRef] [Green Version]
  54. Vincenzi, S.; Zucchetta, M.; Franzoi, P.; Pellizzato, M.; Pranovi, F.; De Leo, G.A.; Torricelli, P. Application of a Random Forest Algorithm to Predict Spatial Distribution of the Potential Yield of Ruditapes Philippinarum in the Venice Lagoon, Italy. Ecol. Modell. 2011, 222, 1471–1478. [Google Scholar] [CrossRef]
  55. Liaw, A.; Wiener, M. Classification and Regression by RandomForest. R News 2002, 2, 18–22. [Google Scholar]
  56. Hanley, J.A.; McNeil, B.J. The Meaning and Use of the Area under a Receiver Operating Characteristic (ROC) Curve. Radiology 1982, 143, 29–36. [Google Scholar] [CrossRef] [Green Version]
  57. Hoo, Z.H.; Candlish, J.; Teare, D. What Is an ROC Curve? Emerg. Med. J. 2017, 34, 357–359. [Google Scholar] [CrossRef] [Green Version]
  58. Tirkey, L.P.; Nepal, P. Tea plantations in the Darjeeling Hills geo-ecological impact and livelihood implications. Hydro Nepal J. Water Energy Environ. 2012, 10, 53–59. [Google Scholar] [CrossRef]
  59. Hanoon, M.S.; Ahmed, A.N.; Zaini, N.A.; Razzaq, A.; Kumar, P.; Sherif, M.; Sefelnasr, A.; El-Shafie, A. Developing machine learning algorithms for meteorological temperature and humidity forecasting at Terengganu state in Malaysia. Sci. Rep. 2021, 11, 18935. [Google Scholar] [CrossRef]
  60. Essam, Y.; Ahmed, A.N.; Ramli, R.; Chau, K.W.; Idris Ibrahim, M.S.; Sherif, M.; Sefelnasr, A.; El-Shafie, A. Investigating photovoltaic solar power output forecasting using machine learning algorithms. Eng. Appl. Comput. Fluid Mech. 2022, 16, 2002–2034. [Google Scholar] [CrossRef]
  61. Mallick, S.K.; Rudra, S.; Maity, B. Land suitability assessment for urban built-up development of a city in the Eastern Himalayan foothills: A study towards urban sustainability. Environ. Dev. Sustain. 2022, 24, 1–26. [Google Scholar] [CrossRef]
Sustainability 15 10101 g001 550
Figure 1. Location map of the study area.
Figure 1. Location map of the study area.
Sustainability 15 10101 g001
Sustainability 15 10101 g002 550
Figure 2. Suitability thematic layers for tea plantations: (A) rainfall, (B) temperature, (C) elevation and (D) slope.
Figure 2. Suitability thematic layers for tea plantations: (A) rainfall, (B) temperature, (C) elevation and (D) slope.
Sustainability 15 10101 g002
Sustainability 15 10101 g003 550
Figure 3. Suitability thematic layers for tea plantations: (A) soil depth, (B) soil drainability, (C) soil electrical conductivity, and (D) base saturation.
Figure 3. Suitability thematic layers for tea plantations: (A) soil depth, (B) soil drainability, (C) soil electrical conductivity, and (D) base saturation.
Sustainability 15 10101 g003
Sustainability 15 10101 g004 550
Figure 4. Suitability thematic layers for tea plantations: (A) soil texture, (B) soil pH, (C) NDVI, and (D) LULC.
Figure 4. Suitability thematic layers for tea plantations: (A) soil texture, (B) soil pH, (C) NDVI, and (D) LULC.
Sustainability 15 10101 g004
Sustainability 15 10101 g005 550
Figure 5. Methodological flowchart of this study.
Figure 5. Methodological flowchart of this study.
Sustainability 15 10101 g005
Sustainability 15 10101 g006 550
Figure 6. Tea-cultivation site suitability zones using random forest model (A) and logistic regression (B).
Figure 6. Tea-cultivation site suitability zones using random forest model (A) and logistic regression (B).
Sustainability 15 10101 g006
Sustainability 15 10101 g007 550
Figure 7. Area distribution of tea-plantation-suitability zones.
Figure 7. Area distribution of tea-plantation-suitability zones.
Sustainability 15 10101 g007
Sustainability 15 10101 g008 550
Figure 8. Receiver operating characteristic curve and confusion matrix.
Figure 8. Receiver operating characteristic curve and confusion matrix.
Sustainability 15 10101 g008
Table
Table 1. Different data sources for the tea-plantation-suitability analysis (accessed on 5 January 2023).
Table 1. Different data sources for the tea-plantation-suitability analysis (accessed on 5 January 2023).
Thematic LayersDatasetResolutionSources
RainfallCRU0.5 km × 0.5 kmhttps://crudata.uea.ac.uk
TemperatureCRU0.5 km × 0.5 kmhttps://crudata.uea.ac.uk
ElevationSRTM DEM30 m × 30 mhttps://earthexplorer.usgs.gov
Slope Calculated from SRTM DEM30 m × 30 mhttps://earthexplorer.usgs.gov
Soil depthBHUVAN5 km × 5 kmhttps://bhuvan-app3.nrsc.gov.in
Soil drainabilityISRIC0.5° × 0.5°https://data.isric.org
Soil electrical conductivityISRIC0.5° × 0.5°https://data.isric.org
Base saturationISRIC0.5° × 0.5°https://data.isric.org
Soil textureISRIC0.5° × 0.5°https://data.isric.org
Soil pHSoil grids250 m × 250 mhttps://soilgrids.org
NDVILandsat 8 OLI30 m × 30 mhttps://earthexplorer.usgs.gov
LULCLandsat 8 OLI30 m × 30 mhttps://earthexplorer.usgs.gov
Table
Table 2. Coefficient values of different discrete data factors.
Table 2. Coefficient values of different discrete data factors.
FactorsClassNo. of Pixels in a Class% of Pixels in a ClassNo. of Pixels of Tea Cultivation% of Pixels of Tea CultivationCoefficient of
Logistic Regression
Soil depth4.3–5.0310913.662689.690.29
5.03–5.6825198.4672926.370.477
5.68–6.34473815.91134348.570.462
6.34–6.9921,43471.9742515.370.306
Soil textureSandy loam11,28637.9049517.920.07
Silt loam762225.601666.010.190
Clay loam10,58135.54208275.350.412
Loam2870.96200.720.057
Soil pHbelow 5.22670.9000.00−0.76
5.2–5.515,12250.821635.89−1.01
5.5–5.6708523.8158521.15−0.34
5.6–6.0728124.47201872.960.09
Soil drainabilityModerately well-drained457015.361696.110.745
Well-drained18,02460.57246589.121.096
Somewhat excessively drained644221.651320.441.145
Excessively drained7202.4200.000.866
Soil electrical conductivity1.52–1.987862.6400.000.172
1.98–2.4619636.6054319.630.069
2.46–3.2210,26534.49210576.100.016
3.22–3.9916,73456.251184.270.006
Soil base saturation3.0–3.714,06247.2735012.65−0.501
3.7–4.5698123.4695534.5314.021
4.5–5.3786126.42146152.8214.593
5.3–6.18472.8500.0014.238
LULCAgricultural land5601.8800.000.722
Built-up area3441.1600.00−0.567
Sand depositions16365.5000.000.019
Tea cultivation299010.052770100.001.381
Vegetation20,28868.2100.000.206
Waterbodies392713.2000.00−0.410
Coefficient of logistic regression value for continuous data: temperature 0.000295, slope 0.04237, rainfall 0.001798, NDVI 0.024655, and elevation −0.07205.
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

Sahu, N.; Das, P.; Saini, A.; Varun, A.; Mallick, S.K.; Nayan, R.; Aggarwal, S.P.; Pani, B.; Kesharwani, R.; Kumar, A. Analysis of Tea Plantation Suitability Using Geostatistical and Machine Learning Techniques: A Case of Darjeeling Himalaya, India. Sustainability 2023, 15, 10101. https://doi.org/10.3390/su151310101

AMA Style

Sahu N, Das P, Saini A, Varun A, Mallick SK, Nayan R, Aggarwal SP, Pani B, Kesharwani R, Kumar A. Analysis of Tea Plantation Suitability Using Geostatistical and Machine Learning Techniques: A Case of Darjeeling Himalaya, India. Sustainability. 2023; 15(13):10101. https://doi.org/10.3390/su151310101

Chicago/Turabian Style

Sahu, Netrananda, Pritiranjan Das, Atul Saini, Ayush Varun, Suraj Kumar Mallick, Rajiv Nayan, S. P. Aggarwal, Balaram Pani, Ravi Kesharwani, and Anil Kumar. 2023. "Analysis of Tea Plantation Suitability Using Geostatistical and Machine Learning Techniques: A Case of Darjeeling Himalaya, India" Sustainability 15, no. 13: 10101. https://doi.org/10.3390/su151310101

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