Skip to main content
NCBI home page
Springer Nature - PMC COVID-19 Collection logoLink to Springer Nature - PMC COVID-19 Collection
. 2023 May 24:1–12. Online ahead of print. doi: 10.1007/s42965-023-00304-x

Monitoring tea plantations during 1990–2022 using multi-temporal satellite data in Assam (India)

Bikash Ranjan Parida 1,✉, Trinath Mahato 1, Surajit Ghosh 2
PMCID: PMC10206575  PMID: 37362781

Abstract

Background

Tea is a valuable economic plant grown extensively in several Asian countries. The accurate mapping of tea plantations is critical for the growth and development of the tea industry. In eastern India, tea plantations have a significant role in its economy. Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia are major tea-producing districts in Assam. Due to the rapid increase in tea plantations and the burgeoning population, a detailed mapping and regular monitoring of tea plantations are imperative for understanding land use alteration.

Objectives

The present study aims to analyse the dynamics of tea plantations from 1990 to 2022 at a decadal scale, using satellite data, such as Landsat-5 and Sentinel-2.

Methods

A supervised classifier called Random Forest (RF) was deployed in the Google Earth Engine (GEE) platform to classify tea plantations.

Results

The results showed significant growth in tea plantations in the district of Dibrugarh (112%), whereas the remaining districts had a growth rate of 45–89%. During 32 years (1990–2022), about 1280.47 km2 (78.71%) of areas of tea plantations expanded across five districts of Assam. Precision and recall were used to measure the accuracy of tea plantations classification, which exhibited considerably high F1 scores (0.80 to 0.96).

Conclusions

This study helps to demonstrate the application of remote sensing techniques to evaluate the dynamics of tea plantations which can help policymakers to manage the tea estates and underlying changes in land cover.

Keywords: Tea plantation, Random Forest Classifier, Assam, Google Earth Engine (GEE), Land Use Land Cover (LULC)

Introduction

Globally, tea (Camellia sinensis L.) consumption is expanding faster than coffee and has become one of the most popular beverages. Green tea is regarded as the best among all tea varieties, having certain medical advantages in terms of antioxidant activity and a greater concentration of gallate content (Lou et al. 2013). It is a non-alcoholic beverage consumed by over 66% of the world’s population (Fu and Weng 2016). Tea originated in China, and after that, it spread all over the globe, including India, Indonesia, Sri Lanka, Kenya and some parts of South America. Globally, the tea has a cultivated area of around 4.9 million hectares, with a total production of 5.96 million tons in 2018. In 2018, China had 2.98 million hectares of tea cultivation land, accounting for 61.7% of the world’s tea plantation area (Zhu et al. 2019). After China, India is second in tea production, followed by Kenya, Sri Lanka, and Turkey. In India, tea grower states are Assam, West Bengal, Tamil Nadu, Kerala, and Tripura. Assam is widely recognised among all these states for its tea and quality. Assam cultivates its indigenous tea variety known as Assam-type var. Assamica (Reddy et al. 2015). Almost 50% of the tea is produced in India, and about 17% of the world’s total production comes from Assam (Duncan et al. 2016). Tea plantations are very delicate and, as a result, they must be constantly monitored and managed to ensure environmental sustainability.

The influence of both international and domestic markets, as well as local factors, such as the expansion of tea plantations, were studied to determine the consequences of land use land cover (LULC) conversion over the last several decades (Nasir and Shamsuddoha 2011; Parida and Kumari 2021a). It was reported that between 1874 and 2010, the forest area dropped by almost 69.5%, while tea plantations expanded by 30.7% in the Himalaya Piedmont zone of India’s West Bengal state (Prokop 2018). A similar trend of LULC transition was seen in the Bagapani Corridor, Kamrup district of Assam, where cropland decreased from 0.60 (2011) to 0.15 (2020) km2. A substantial rise in the settlement and tea plantations was noticed, and tea plantations have increased from 1.05 km2 in 2011 to 1.11 km2 in 2020 (Sarma et al. 2021). In the upper Dihing East-West Corridor tea plantation in Assam, it was reported that tea plantations increased from 1.94 km2 in 2011 to 1.96 km2 in 2020 (Dikshit and Dikshit 2014). In the Coonoor watershed in the Nilgiris district of Tamil Nadu, it was found that dense forest cover declined from 2000 to 2018 by 1158.48 km2, whereas tea plantations increased by 223.29 km2 (Saravanan et al. 2021). The rapid rise in tea plantations and the decline of forest cover led to an environmental crisis and disruption in the forest habitat. Rise in the tea plantation along the elephant corridor was also noticed in the Buxa tiger reserve, where it creates human-elephant conflict (HEC) and causes a threat to biodiversity conservation and human beings (Nad et al. 2022). Buxa tiger reserve in West Bengal has faced a decline in the forest cover by 184.04 km2 from 1990 to 2019 and a rise in the tea plantations by 87.05 km2 (Nad et al. 2022). In the Fujian Province of China from 1997 to 2009, conversions of Cunninghamia lanceolata forests to tea plantations were reported by 280.82 km2 and cultivated lands to tea plantations by 96.65 km2. During that period, tea plantations and built-up showed an evident change in area by 360.81 and 155.14 km2, respectively (You et al. 2017). A similar trend of conversion of forest cover to other LULC classes was also seen in Menghai County, China, from 2010 to 2015, where forest area decreased from 3286.66 to 3122.04 km2, whilst tea plantation increased from 278.17 km2 to 371.52 km2 in 5 years with an annual growth rate of 6.7% (Xu et al. 2018).

Remote sensing technology has been widely employed for spatial and temporal monitoring of LULC (Sahoo et al. 2022) and tea health conditions during the last few decades (1990–2022) (Prokop 2018; Phan et al. 2020; Moni et al. 2022). Spectral indices such as normalized difference vegetation index (NDVI), leaf area index (LAI), Simple Ratio (SR), Triangular Vegetation Index (TVI), Leaf Area Index (LAI), Tasseled Cap Transformation (TCT), and Normalized Difference Vegetation Index (NDWI) were evaluated as a method for detecting and identifying tea farms along with the tea patch texture and tone changes (Zhu et al. 2019; Rao et al. 2007). Remote sensing data includes a wide range of high spectral, temporal, and spatial resolutions, which were widely utilised for tea plantation mapping worldwide, including detecting conversion of LULC (Dihkan et al. 2013; Parida and Kumar 2020, Ranjan and Parida 2020). Because of their adequate geographic coverage with improving spatial resolution, satellite images are employed for tea plantations mapping and monitoring (e.g., Landsat series, IRS series, SPOT series, and Sentinel). Tea plantations in north Bengal regions in West Bengal (India) were detected and analysed using Sentinel-2 Multi Spectral Instrument (MSI) and Landsat-5 Thematic Mapper (TM) sensors (Parida and Kumari 2021a). High spatiotemporal multispectral data were also employed to study the tea plantation’s phenological features in northern Zhejiang from April to May 2018 (Li et al. 2019). Based on the spectral indices, such as NDVI and LAI, tea yield and waterlogging conditions of tea plantations were analysed by Rossell Tea Company Limited, Nagrijuli tea estate Assam, India (Rao et al. 2007). Hyperspectral and Synthetic Aperture Radar (SAR) data are also employed for tea plantations identification, and studies indicated that the fusion between Hyperspectral and SAR has improved overall accuracy (Chen et al., 2020). Hu et al. (2016) offered an object-based fusion technique, and the findings revealed that synergistic usage could achieve greater performance in comparison to a single data source.

Tea yield is frequently affected by rapid changes in climatic conditions and extreme weather events (Ahmed et al. 2014; Dutta 2014; Duncan et al. 2016). It was reported that a monthly average temperature greater than 26.6 °C negatively influences tea yield, and an extra one degree of monthly average temperature leads to a drop in tea yields by 3.8% (Duncan et al. 2016). It also indicated that heavy precipitation during monsoon season significantly negatively impacts tea yield. Tea plantations are sensitive to soil moisture as well as severe events (i.e., drought conditions), which results in yield losses by reducing tea crown density (Das et al. 2021). A study reported that tea cultivation in Sri Lanka needs an optimum temperature of about 22℃. A reduction in precipitation by 100 mm could decrease the tea productivity by 30–80 kg, whilst an increase in ambient CO2 concentration increased the yield by about 33–37% due to an increase in the rate of photosynthesis (Wijeratne et al. 2007). Furthermore, drought conditions negatively impact tea yield and quality by reducing plant growth. The standardised precipitation index (SPI), land surface temperature (LST), soil moisture index (SMI), NDWI, NDVI, and LAI derived from Landsat data revealed that during drought periods (i.e., 2018-19, 2019-20, and 2020-21) in Sylhet division (Bangladesh), caused yield losses by 7.72%, 11.92%, and 12.52%, respectively (Das et al. 2021).

Few studies have employed satellite data and machine learning (ML) algorithms to examine tea plantations (Phan et al. 2020; Qu et al. 2022). In this context, the ML algorithm, especially Random Forest (RF), was utilised in the Google Earth Engine (GEE) cloud platform to map tea plantations in the Anhui Province of eastern China (Qu et al. 2022). The RF is a well-suited and dependable classifier for extracting tea plantations during the spring tea flushes in March and April (Wang et al. 2019). Based on the multi-sensors data, Tridawati et al. (2020) employed an RF algorithm to classify tea plantations and demonstrated that the classification accuracy was lower, with a kappa coefficient of 0.77. A similar study was conducted in the Tanuyen District of Lai Chau Province, Vietnam; forecasting tea health and yield through machine learning (ML) methods such as support vector machine (SVM) and random forest (RF) are very effective in estimating future yields, with the accuracy of R2 = 0.66 and 0.73, respectively (Phan et al. 2020). Spring frost conditions can reduce the tea yield and production because it damages the bud’s tea plants and the young shoot. In Yuezhou Longjing, China, tea damage indicated that the minimum temperatures and frost led to enormous economic loss (Lou et al. 2013). To improve the accuracy of tea plantation identification, typically, multi-temporal, multi-feature and phenological patterns of tea plantations are deployed in the Shihe district of China (Zhu et al. 2019). It was found that the feature significance analysis in RF classification effectively minimises feature dimensions and increases classification efficiency (Zhu et al. 2019). In addition to satellite data, field instruments such as proximal sensors (i.e., handheld spectroradiometer with 512-channel) have been widely used in detecting complex information like spectral behaviours and tea productivity (Lin and Sun 2020). A spectral range of 325–1075 nm is typically deployed for discriminating tea varieties, age, growth stage, health and diseases and shade conditions.

Worldwide, tea cultivation areas are increasing with their demand in the domestic and international markets. Hence, tea plantation identification is critical for the adoption of sustainable practices and frequent monitoring (Kumar et al. 2013). Moreover, tea plantation needs extensive maintenance and a large number of labourers and investments (Nasir and Shamsuddoha 2011). In this study, the extent of tea plantations in the state of Assam (India) was considered as it represents a primary tea-producing state in India, along with the pressure of anthropogenic activities, such as land use conversion and substantial management consequences. The expansion of tea plantations has benefited the development of the local economy in Assam, but it has also resulted in various environmental challenges, such as diminished soil fertility and soil erosion (Alom et al. 2020; Alam et al. 2022). The objective of this study is to monitor tea plantations at decadal time scales (1990–2022) in the major tea-growing districts of Assam (i.e., Sibsagar, Tinsukia, Jorhat, Sonitpur, and Dibrugarh) for evaluating the expansion of tea plantations. The satellite data of Landsat-5 and Sentinel-2 were employed to identify and map tea patches using Random Forest (RF) classifier.

Study area

Assam is located in eastern India and is well-known for its tea, landscape, and culture. The tea industry of Assam lies in the Brahmaputra plains, which comprises the districts, namely, Udalgiri, Nagaon, Golaghat, Sonitpur, Jorhat, Sibsagar, Dibrugarh, Tinsukia. Assam tea industries are spread across an area of 21,620 km2. Geographically, Assam covers an area of 78,438 km2 and stretches between 89° 42' E to 96° E longitude and 24° 8' N to 28° 2' N latitude. Assam primarily has a ‘Tropical Monsoon Rainforest Climate’, lies in the temperate zone and gets significant rainfall and humidity. During June, the monsoon reaches its peak rainfall of 250–300 mm/day. Winter dominates from late October to late February, with a minimum temperature of 6–8 °C. Summer begins during mid-May and has a maximum temperature of 35–38 °C (Dutta et al. 2021). These climatic conditions provide the most desirable seasons for tea plantations: spring and autumn, with moderate temperatures and little rain. Assam is located on either side of the Brahmaputra River and shares borders with Bangladesh and Myanmar. Agriculture employs half of the state’s population and accounts for around one-third of the state’s total GDP. The study area contains a variety of geological and geographical characteristics (Bhattacharyya et al. 2015). The creation of distinct types of soils has been fostered by climatic circumstances and diverse plant types. There are four primary soil types identified in the study area: alluvial soils, Piedmont soils, hill soils, and lateritic soils. The soils in the research region are high in organic matter and nitrogen.

Assam is one of the world’s most prolific tea-producing states. The state is separated into three geographical parts: the upper Surma river valley (Barak River) in the south, the Brahmaputra river valley in the north, and the mountainous hills between Meghalaya, Nagaland, and Manipur in the east. Geologically, the Brahmaputra and Barak valleys are located on old alluvial sediments encompassing a wide range of alluvial deposits from the Neogene and Paleogene periods consisting of hard sandstone, loose and soft sand, coal seams, shales, conglomerates, limestone, and sandy clays are widespread in these deposits.

In this study, five districts were chosen to analyse the temporal variations of tea plantations: Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia (Fig. 1). The five districts account total area of 18,020 km2. In Assam, tea cultivation occurs four times a year: the first flush (Spring Season) from March to April, the second flush (Summer Season) from May to June, the third flush (Monsoon flush) from July to September, and the fourth flush (Autumn flush) from October to December. Most tea plantations in Assam are located in the lowlands of the Brahmaputra valley, at elevations ranging from 60 to 100 m above mean sea level and with a mild slope.

Fig. 1.

Fig. 1

Open in a new tab

Location of Assam has been highlighted in yellow inside the India map. The False Color Composite (FCC) of five tea-producing states has been shown (Data Source: Sentinel-2 A acquired between March to April 2022)

Materials and methods

Satellite data

A detailed study workflow for tea plantation mapping in 1990, 2000, 2010, and 2022 is shown in Fig. 2. This workflow consists of two satellite data, namely Landsat-5 and Sentinel-2, to study tea plantation dynamics and structure. The Landsat-5 data has been used to map tea plantations from 1990 to 2010 at decadal intervals, whereas the Sentinel-2 data was preferred for the year 2022 due to its fine spatial resolution. The detailed sensor characteristics are given in (Table 1).

Fig. 2.

Fig. 2

Open in a new tab

Methodology structure for deriving tea plantation extent in 1990, 2000, 2010, and 2022

Table 1.

Datasets used for the spatiotemporal analysis of tea plantations for the years 1990, 2000, 2010, and 2022

Data used Spatial resolution (m) Wavelength (µm) Acquisition dates Purpose
Landsat-5 TM 30

0.45–0.52

0.52–0.60

0.63–0.69

0.76–0.90

 March - April Tea plantation mapping for 1990, 2000, and 2010.
Sentinel-2 A MSI 10, 20, 60

0.45–0.52

0.54–0.57

0.65–0.68

0.69–0.71

0.78–0.90

 March - April Tea plantation mapping for 2022
Open in a new tab

Landsat-5 Thematic Mapper (TM) was launched by the National Aeronautics and Space Administration (NASA), and it is the longest-serving earth observation satellite (1984–2013), extensively used to monitor long-term land cover change and time series analysis. Sentinel-2 A was the first optical Earth Observing Satellite launched in 2015 by the European Space Agency (ESA) as part of the European Copernicus Programme. It has been widely used to monitor vegetation and agricultural studies (Parida et al. 2022). Sentinel-2 A consists of high-resolution bands, such as red edge bands, that enable to distinguish early changes from plant health and are utilised for analysing biochemical and biophysical compositions (Parida and Kumari 2021b).

Methods

Satellite imageries of Landsat-5 TM and Sentinel-2 A MSI were obtained from the United States Geological Survey (USGS) and processed via the Google Earth Engine (GEE) cloud platform. For 1990, 2000, and 2010, Landsat-5 TM Top of Atmosphere (TOA) bands are utilised (Fu and Weng 2016). Similarly, atmospherically adjusted Sentinel-2 MSS bands were processed for the year 2022. GEE is efficient enough to perform various image analysis as it provides a simple geospatial image data viewer, access to massive datasets, and regional datasets accessible in the Earth Engine Data Catalog. To call image collections of different satellite imageries, the ee.ImageCollection was used along with the name of the satellite dataset tag as “ee.ImageCollection(“LANDSAT/LT05/C01/T1_TOA”)” for the Landsat-5 TM, for the Sentinel-2, the tag of “ee.ImageCollection (“COPERNICUS/S2_SR”)” was used. Since both datasets have different geographical resolutions, we applied the RF classification separately for each year. The areas of tea plantations in the study area are measured after post-classification techniques applied on a classified map.

All the image processing was carried out in GEE using the JavaScript API (https://code.earthengine.google.com/). To classify tea plantations for the years 1990, 2000, 2010, and 2022, we applied the supervised classification approach and the RF classifier. RF classifier is a machine learning technique based on the ensemble learning method available in GEE generally uses six input parameters, which are the number of variables used at each node, number of trees used in the classification, minimum leaf population, random seed variable for construction of decision trees, bagged fraction of the input variables for each decision trees, and out-of-bag mode. The overall accuracy of the classification increases as the number of trees increases until convergence begins without overfitting. The flowchart has been shown in (Fig. 2) to understand the adopted methodology. Furthermore, the F1 score is used to determine the accuracy of the classification on GEE and divide the reference data into two parts such as the training and validation samples collected manually based on visual interpretation of high-resolution satellite images using Google Earth Pro. In the present study, 1287 samples were used from Google Earth Pro for the year 1990, 2000, and 2010, whereas 132 GPS locations taken during the field visit in 2022 was used to perform an accuracy assessment. This specific method was extensively reported and applied in the various literature for better accuracy, validation, and training datasets were separately selected. They are used to compute overall accuracy, producers’ accuracy, precision, and recall (Kumar and Parida 2021).

For image classification, accuracy assessment is the most important final step. It helps to provide the quantitative measure of how effectively segments (or pixels) are assigned to the correct LULC classes. In other words, classification accuracy is defined by the degree of similarity between the referenced and produced maps. Various indices and approaches for accuracy assessment have been developed in past years. The commonly used overall measures of accuracy are based on the Kappa coefficient and overall accuracy (OA). The kappa coefficient in accuracy assessment varies from − 1 to 1. Negative kappa values signify that the classification is significantly worse than random classification. Kappa coefficient values close to 1 signify the classification is correct. In the accuracy assessment, OA indicates the likelihood that a randomly selected location on the map is correctly classified. To assess the effectiveness of the binary classes (tea and non-tea), this study used F1-Score, precision and recall for accuracy assessment. The F1 score is typically employed when two classes are generated from satellite images based on recall and accuracy (Maxwell and Warner 2020). The precision is determined by the 1-commission error (user’s accuracy), while the recall is determined by the 1-omission error (producer’s accuracy) (Eqs. 1, 2, and 3).

graphic file with name M1.gif Eq. 1
graphic file with name M2.gif Eq. 2
graphic file with name M3.gif Eq. 3

Where FN, FP, and TP denote the number of false negatives, false positives, and true positives, respectively. The metric for each category was calculated first, followed by the average score across all categories.

Results

Spatial-temporal distribution of tea plantations in 1990, 2000, 2010, and 2022

To quantify the tea extent maps, four years of satellite imagery at ten years interval were used to demonstrate the spread of tea plantations across five districts of Assam, namely Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia. From 1990 to 2022, the pattern of tea plantations in those five districts indicated an increasing area under the tea plantations (Fig. 3). Notably, an increase in areas is evident in three districts Sonitpur, Dibrugarh, and Tinsukia. This is due to the rise in population and the worldwide increase in demand for Assam tea because of the better quality of tea. Over the years, a considerable level of investment in the Assam tea industry was noticed.

Fig. 3.

Fig. 3

Open in a new tab

Satellite-derived tea extent maps at decadal periods from 1990 to 2022

According to tea statistics derived from tea maps of 1990, the Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia districts accounted for 280.73, 243.89, 384.12, 376.43, and 351.88 km2 of tea plantations respectively, with the Sibsagar district having the largest area and the Jorhat district having the lowest of tea gardens (Table 2). In 2000, tea gardens were increased, with an area of 337.53, 233.11, 386.25, 576.53, and 397.94 km2 in Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia districts, respectively. In 2010, tea plantations in the districts showed a growing trend with an area of 463.85, 349.02, 424.93, 817.79, and 457.14 km2 in Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia districts, respectively. In 2022, the area of tea plantations was 531.65, 433.53, 557.34, 800.03, and 594.97 km2 in Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia, respectively. However, Dibrugarh reported a decline in tea gardens by 17.76 km2 due to urbanisation from 2010 to 2022. As per the satellite estimates, tea plantation areas were 1637.05 km2 in 1990 and 2917.52 km2 in 2022, indicating that tea plantations rose by 1280.47 km2 (or 78.71%) over the last three decades (1992–2022). The most extensive growth in tea plantings was 112.53% in Dibrugarh (423.06 km2) and 89.38% in Sonitpur (250.92 km2) (Table 2). It can be attributed to an increase in population in these regions as well as a rise in market demand for tea. Dibrugarh has the most significant rise in tea plantations from 1990 to 2022, owing to an increase in tea demand worldwide and massive investment by the tea industry. Thus, Dibrugarh is one of the districts that produce more tea than any other Assam district. The tea Industry plays an important role in the development of the Assam economy, and due to high economic benefits, a large number of populations started cultivating tea leading to the rise in small tea growers in the past decades over the study area and especially in the districts Dibrugarh and Tinsukia. Due to the continuous inflow of immigrant labourers from other states, the state of Assam is facing a rise in population as well as tea plantations. Classification accuracy of tea plantations is measured using the F1 score. It was seen that the F1 score ranges from 0.80 to 0.96, indicating a high level of classification accuracy, which is measured between 0 and 1 (Tables 3 and 4).

Table 2.

District-wise tea plantation statistics (km2) for 1990, 2000, 2010, and 2022

Districts Area in 1990 (km2) Area in 2000 (km2) Area in 2010 (km2) Area in 2022 (km2) Area change (1990–2022) (km2) % change
(1990–2022)
Sonitpur 280.73 337.53 463.85 531.65 + 250.92 89.38
Jorhat 243.89 233.11 349.02 433.53 + 189.64 77.77
Sibsagar 384.12 386.25 424.93 557.34 + 173.22 45.10
Dibrugarh 376.43 576.53 817.79 800.03 + 423.6 112.53
Tinsukia 351.88 397.94 457.14 594.97 + 243.09 69.10
Total 1637.05 1931.36 2512.73 2917.52 + 1280.47 78.71
Open in a new tab

Table 3.

Accuracy assessment using F1 score of tea classification from 1990 to 2022 of Sonitpur and Jorhat

Years Sonitpur Jorhat
Recall Precision F1 Score Recall Precision F1 Score
1990 0.94 0.95 0.94 0.92 0.93 0.92
2000 0.91 0.94 0.94 0.87 0.90 0.88
2010 0.87 0.90 0.88 0.89 0.88 0.88
2022 0.92 0.91 0.91 0.90 0.91 0.90
Open in a new tab

Table 4.

Accuracy assessment using F1 score of tea classification from 1990 to 2022 of Sibsagar, Dibrugarh, and Tinsukia

Years Sibsagar Dibrugarh Tinsukia
Recall Precision F1 Score Recall Precision F1 Score Recall Precision F1 Score
1990 0.95 0.94 0.94 0.88 0.90 0.89 0.94 0.93 0.93
2000 0.92 0.93 0.92 0.83 0.78 0.80 0.86 0.87 0.86
2010 0.92 0.93 0.92 0.97 0.96 0.96 0.93 0.91 0.91
2022 0.94 0.91 0.92 0.95 0.88 0.91 0.94 0.87 0.90
Open in a new tab

Validating tea plantation patterns across different districts in Assam using images from google earth pro and satellite imageries

For validating tea plantation classified images from 1990 to 2022, three locations across three districts were randomly selected as marked in Fig. 4. The corresponding three locations were zoomed and shown in Fig. 5. High-resolution imageries from Google Earth Pro are utilised to validate the classified maps (Fig. 5). The tea plantation is represented by symmetrical green patches in the True Colour Composite (TCC), with a red circle indicating the changes that have occurred over the past 32 years (1990–2022). Sonitpur’s tea plantations area was 280.73 km2 in 1990 and expanded to 531.65 km2 in 2022 due to a significant expansion of both small and medium size tea plantation growers. Tea plantation findings show steady growth in tea plantations over time, and the classified map is comparable to Google Earth pro high-resolution imageries (Fig. 5a) across all five districts. In Dibrugarh, there is a slight decline in tea plantations in 2022 due to the district’s expansion of concrete structures, similar to the False Colour Composite (FCC) and Google Earth Pro TCC images (Fig. 5b). However, from 1990 to 2022, the tea plantation in Dibrugarh rose by 423.6 km2 (Table 2). Tinsukia is Assam’s leading tea plantation district and produces the finest quality tea. Tinsukia tea industry has experienced a drastic spatial change from 1990 (351.88 km2) to 2022 (594.97 km2), which is primarily due to a rise in the number of small tea producers in the districts of Dibrugarh and Tinsukia. Tinsukia has many small tea farmers, as represented by the red circle markings (Fig. 5c). Both FCC and TCC imageries clearly show a considerable increase in small tea producers from 1990 to 2022, with desolate land in 1990 images changed to small-level tea plantations. These small tea growers significantly contribute to tea quality because they manage tea gardens on a small scale with proper care and monitoring. In the study area, most tea plantations are found at low elevations (45–60 m above the mean sea level).

Fig. 4.

Fig. 4

Open in a new tab

The three locations marked by circles were zoomed and shown in Fig. 5. The FCC indicates the RGB composite created for the year 2022 from Sentinle-2

Fig. 5.

Fig. 5

Open in a new tab

Satellite-derived tea plantation from 1990 to 2022 in Sonitpur (a), Dibrugarh (b) and Tinsukia (c) district and its comparison with FCC (satellite data) and TCC (Google Earth Pro)

Discussion

Assam is India’s leading tea-producing state, and hence, the tea business is critical to the state’s economy. Its total tea production in the 2020 fiscal year was 618.20 million kilograms, but its statistics show that Assam’s tea production dropped by 8% in 2021 (Tea Board of India 2021) due to Covid-19 induced lockdown (Parida et al. 2021). The major findings exhibited an increasing pattern of tea plantations in Assam, specifically in the districts Sonitpur, Jorhat, Sibsagar, Dibrugarh, and Tinsukia, which aids in quantifying the long-term spatial extent of tea plantations for decision-making, policy-making, management practices, and conservation techniques. The increasing pattern was observed when the three decades (1990–2020) of LULC changes were analysed. The main findings showed that satellite images provide critical information about the long-term dynamics of tea plantations over four decades 1990, 2000, 2010, and 2022, with the total expansion of tea plantations by 1280.47 km2 from 1990 to 2022. The most expansion is seen in the districts of Dibrugarh and Sonitpur by 423.6 and 250.92 km2, respectively, along with an increase in demography in the region. As the scope of tea plantations grows in these districts, other land uses, such as agricultural fields, shifting cultivation, and forest covers, are being turned into tea plantations. A similar pattern was also reported in Siliguri Urban Agglomerations (UA) located in the Darjeeling district in West Bengal, where an expansion of tea plantation area by 2.92% (from 53.38 km2 to 68.82 km2) was reported during the last 29 years (1990–2019) based on Landsat satellite data. It is promoting economic development in Siliguri and the surrounding due to increasing tea estates. However, during the same period, agricultural land decreased by almost 30% (Roy and Kasemi 2022). Natural forest cover was also transformed into tea plantations, agricultural operations, and built-up such as dwellings, markets, and agricultural enterprises (Kumari et al. 2022). In the Nilgiris district of Tamil Nadu, forest cover declined by 516.98 km2 (38%) from 2000 to 2015, while tea plantations increased by 196.75 km2 (70%) which revealed that tea plantations have taken over around 187.85 km2 of the forest land (Nunna 2020).

Along with the increasing pattern of the area under tea plantations, it is evident that tea production also increased (Tea Board of India 2021), which could be attributed to the rising demand for tea in national and international markets. Specifically, the Northeast region of India has seen a change in land use policy as a result of population dynamics and an understanding of tea cultivation methods (Bose 2019; Dikshit and Dikshit 2014). As per the New Land Use Policy (NLUP), the population is likely to grow in the areas of tea estates (Dikshit and Dikshit 2014). Tea plantation regions also possess a higher population density than cities due to economic activities throughout the year. Tea plantation is a labour-intensive industry that directly employs 1.2 million people in Assam state alone and supports over 10 million people who rely on tea plantation labour, with women accounting for approximately half of the workforce (Ekka et al. 2021). The majority of Assam’s tea plantations are located in the state’s eastern region. Apart from Sonitpur, Jorhat, Sibsagar, and Udalguri, the largest tea-producing districts were identified as Dibrugarh and Tinsukia.

Some limitations were noticed when extracting irregular patterns of tea plantations from fragmented scenes in satellite data. The majority of tea plantations are also shaded by trees, resulting in spectral flexibility in tea plantations. The signature of tea plantations differs from that of other forests, plantations, and croplands owing to the distinct phenology and planting pattern of tea plantations (Beringer et al. 2020). Moreover, the coarser spatial resolution (30 m) of Landsat-5 (TM) makes it impossible to distinguish the similar spectral signatures between tea plantations and diverse landcover classifications (e.g., orchard, nursery, forest, shrub, and cropland) (Xu et al. 2018). In comparison, Sentinel-2 satellite data with a spatial resolution of 10 m can easily distinguish between identical spectra of tea and other vegetation. Thereby, training sites from Sentinel-2 produced better clusters to separate tea and non-tea classes and consistently produced better F1 scores for the year 2022.

Conclusions

Satellite-derived accurate information on the dynamics of LULC and significant changes in tea plantations is required to assess environmental change, land management, and planning. For tracking the dynamics of tea plants from 1990 to 2022, high-resolution satellite data, such as Sentinel-2, were examined using supervised classification. The spectral signature of tea plantations differs from other LULC classes, such as forests, plantations, and cropland, due to the structure of leaf canopy and the unique phenology and planting design of tea plantations, but small-scale tea plantations are difficult to segregate from other green vegetation pixels because tea plantations are covered by tree canopy. The study’s primary findings exhibited that tea expanded by 1280.47 km2 (78.71%) during the previous 32 years, from 1990 (1637.05 km2) to 2022 (2917.52 km2). From 1990 to 2022, there was significant growth in population in the districts of Sonitpur and Dibrugarh due to the expansion of tea plantations. In Assam, tea plantations are heavily invested in improving management and yields, and many tea research projects are underway. The findings of this high-resolution satellite investigation give detailed information for management strategies and health monitoring. This research has the potential to assist tea researchers in evaluating regional-level monitoring and health analyses over time.

Acknowledgements

Authors thank the Google Earth Engine (GEE) team for developing an efficient cloud-based remote sensing platform that can handle big datasets. We are thankful to S. Bar for technical support and B. Kanu for helping during the field data collection.

Declarations

Conflict of interest

The authors declare no competing interests.

References

  1. Ahmed S, Stepp JR, Orians C, Griffin T, Matyas C, Robbat A, Cash S, Xue D, Long C, Unachukwu U, Buckley S, Small D, Kennelly E. Effects of extreme climate events on tea (Camellia sinensis) functional quality validate indigenous farmer knowledge and sensory preferences in Tropical China. PLoS ONE. 2014;9(10):e109126. doi: 10.1371/journal.pone.0109126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Alam MS, Akter S, Shemul S, Chowdhury TR, Ara S. Assessing the quality and heavy metal contamination of soil in tea gardens around Magurchara gas blowout in Bangladesh using multivariate and soil quality index methods. J Hazard Mater Adv. 2022;7:100127. doi: 10.1016/j.hazadv.2022.100127. [DOI] [Google Scholar]
  3. Alom S, Das R, Ahmed CS, Baruah U, Das S. Evaluation of Soil Quality in Tea Plantation under Present Climatic Condition. Plant Archives. 2020;20:7352–7358. [Google Scholar]
  4. Beringer T, Kulak M, Müller C, Schaphoff S, Jans Y. First process-based simulations of climate change impacts on global tea production indicate large effects in the world’s major producer countries. Environ Res Lett. 2020;15:034023. doi: 10.1088/1748-9326/ab649b. [DOI] [Google Scholar]
  5. Bhattacharyya PN, Sarmah SR, Dutta P, Tanti AJ. Emergence in mapping microbial diversity in tea (Camellia sinensis (L.) O. Kuntze) soil of Assam, North-East India: a novel approach. Eur J Biotechnol Bioscience. 2015;3:2321–9122. [Google Scholar]
  6. Bose P. Oil palm plantations vs. shifting cultivation for indigenous peoples: analyzing Mizoram’s New Land Use Policy. Land Use Policy. 2019;81:115–123. doi: 10.1016/j.landusepol.2018.10.022. [DOI] [Google Scholar]
  7. Chen Y, Tian S. Feature-level fusion between Gaofen-5 and Sentinel-1a data for tea plantation mapping. Forests. 2020;11(12):1–21. doi: 10.3390/f11121357. [DOI] [Google Scholar]
  8. Das AC, Noguchi R, Ahamed T. An assessment of drought stress in tea estates using optical and thermal remote sensing. Remote Sens. 2021;13:2730. doi: 10.3390/rs13142730. [DOI] [Google Scholar]
  9. 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. doi: 10.1080/01431161.2013.845317. [DOI] [Google Scholar]
  10. Dikshit KR, Dikshit JK (2014) North-East India: land, people and economy. Springer Dordrecht, pp 289–293. 10.1007/978-94-007-7055-3
  11. Duncan JMA, Saikia SD, Gupta N, Biggs EM. Observing climate impacts on tea yield in Assam, India. Appl Geogr. 2016;77:64–71. doi: 10.1016/j.apgeog.2016.10.004. [DOI] [Google Scholar]
  12. Dutta R. Climate change and its impact on tea in northeast India. J Water Clim Change. 2014;5:625–632. doi: 10.2166/wcc.2014.143. [DOI] [Google Scholar]
  13. Dutta S, Rehman S, Chatterjee S, Sajjad H (2021) Analyzing seasonal variation in the vegetation cover using NDVI and rainfall in the dry deciduous forest region of Eastern India. Forest Resources Resilience and Conflicts, pp 33–48. ISBN 9780128229316, Elsevier
  14. Ekka B, Joseph GA. A review on the role of Youth in the Socio-Economic Development of Tea Garden Community in Assam keywords: Tea Garden Community of India and Assam, Youth, Behaviour, Entrepreneurship, Import-Export Socio-Economic Development Opportunities. Int Educational Sci Res J. 2021;7:16–24. [Google Scholar]
  15. Fu P, Weng Q. A time series analysis of urbanization induced land use and land cover change and its impact on land surface temperature with landsat imagery. Remote Sens Environ. 2016;175:205–214. doi: 10.1016/j.rse.2015.12.040. [DOI] [Google Scholar]
  16. Hu J, Ghamisi P, Schmitt A, Zhu XX (2016) Object based fusion of polarimetric SAR and hyperspectral imaging for land use classification, 2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), Los Angeles, CA, USA, 2016, pp. 1–5
  17. Kumar S, Parida BR. Hydroponic farming hotspot analysis using the Getis–Ord Gi* statistic and high-resolution satellite data of Majuli Island, India. Remote Sens Lett. 2021;12:408–418. doi: 10.1080/2150704X.2021.1895446. [DOI] [Google Scholar]
  18. Kumar A, Manjunath KR, Meenakshi, Bala R, Suda RK, Singh RD, Panigrahy S. Field hyperspectral data analysis for discriminating spectral behavior of tea plantations under various management practices. Int J Appl Earth Obs Geoinf. 2013;23:352–359. [Google Scholar]
  19. Kumari A, Parida BR, Ghosh S (2022) Monitoring Tea Plantations Dynamics Using Satellite Data between 2000 and 2020 in Dooars Regions of Himalaya. “Himalayan Ecosystems & Sustainability Vol. 1: Spatio-Temporal Monitoring of Forests and Climate” Edited by B. R. Parida, A. C. Pandey, M. D. Behara, and Navneet Kumar, CRC Press, Taylor & Francis. ISBN 9781032203140
  20. Li N, Zhang D, Li L, Zhang Y (2019) Mapping the spatial distribution of tea plantations using high-spatiotemporal-resolution imagery in northern Zhejiang, China. Forests 10:856
  21. Lin X, Sun DW. Recent developments in vibrational spectroscopic techniques for tea quality and safety analyses. Trends Food Sci Technol. 2020;104:163–176. doi: 10.1016/j.tifs.2020.06.009. [DOI] [Google Scholar]
  22. Lou W, Ji Z, Sun K, Zhou J. Application of remote sensing and GIS for assessing economic loss caused by frost damage to tea plantations. Precision Agric. 2013;14:606–620. doi: 10.1007/s11119-013-9318-5. [DOI] [Google Scholar]
  23. Maxwell AE, Warner TA (2020) Thematic classification accuracy assessment with inherently uncertain boundaries: An argument for center-weighted accuracy assessment metrics. Remote Sens 12:1905
  24. Moni R, Subrata K, Ritika N, Arun S, Nath J (2022) Mapping the spatial distribution of aboveground biomass of tea agroforestry systems using random forest algorithm in Barak valley, Northeast India. Agrofores Syst 96:1175–1188
  25. Nad C, Roy R, Roy TB. Human elephant conflict in changing land-use land-cover scenario in and adjoining region of Buxa tiger reserve, India. Environ Challenges. 2022;7:100384. doi: 10.1016/j.envc.2021.100384. [DOI] [Google Scholar]
  26. Nasir T, Shamsuddoha M. Tea Productions, Consumptions and exports: Bangladesh Perspective. Int J educational Res Technol. 2011;2:68–73. [Google Scholar]
  27. Nunna VK. Tea Acreage Estimation and Condition Assessment using Satellite Data in Nilgiris District, Tamil Nadu. In: Pawar P, Ronge B, Balasubramaniam R, Vibhute A, Apte S, editors. Techno-Societal 2018. Cham: Springer; 2020. [Google Scholar]
  28. Parida BR, Kumar P. Mapping and dynamic analysis of mangrove forest during 2009–2019 using landsat–5 and sentinel–2 satellite data along Odisha Coast. Trop Ecol. 2020;61:538–549. doi: 10.1007/s42965-020-00112-7. [DOI] [Google Scholar]
  29. Parida BR, Kumari A (2021a) Mapping tea plantations dynamics during 2000–2020 and monitoring biophysical attributes using multi-temporal satellite data in North Bengal (India). Arab J Geosci 14:2096
  30. Parida BR, Kumari A. Mapping and modeling mangrove biophysical and biochemical parameters using Sentinel-2A satellite data in Bhitarkanika National Park, Odisha. Model Earth Syst Environ. 2021;7:2463–2474. doi: 10.1007/s40808-020-01005-3. [DOI] [Google Scholar]
  31. Parida BR, Bar S, Roberts G, Mandal SP, Pandey AC, Kumar M, Dash J. Improvement in air quality and its impact on land surface temperature in major urban areas across India during the first lockdown of the pandemic. Environ Res. 2021;199:111280. doi: 10.1016/j.envres.2021.111280. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Parida BR, Kushwaha A, Ranjan AK. Synergy of Sentinel-2A and Near-proximal sensor data for deriving biochemical parameters of paddy at different growth stages. Environ Dev Sustain. 2022;24:1048–1068. doi: 10.1007/s10668-021-01482-1. [DOI] [Google Scholar]
  33. Phan P, Chen N, Xu L, Chen Z. Using multi-temporal MODIS NDVI data to monitor tea status and forecast yield: a case study at Tanuyen, Laichau, Vietnam. Remote Sens. 2020;12:1–23. doi: 10.3390/rs12111814. [DOI] [Google Scholar]
  34. 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. doi: 10.1016/j.landusepol.2018.05.035. [DOI] [Google Scholar]
  35. Qu L, Li M, Chen Z, Liu W, Zhi J, Zhang L. Mapping large area tea plantations using progressive random forest and Google Earth Engine. J Appl Remote Sens. 2022;16:1–22. doi: 10.1117/1.JRS.16.024509. [DOI] [Google Scholar]
  36. Ranjan AK, Parida BR. Estimating biochemical parameters of paddy using satellite and near-proximal sensor data in Sahibganj Province, Jharkhand (India) Remote Sens Applications: Soc Environ. 2020;18:100293. doi: 10.1016/j.rsase.2020.100293. [DOI] [Google Scholar]
  37. Rao NR, Kapoor M, Sharma N, Venkateswarlu K. Yield prediction and waterlogging assessment for tea plantation land using satellite image-based techniques. Int J Remote Sens. 2007;28:1561–1576. doi: 10.1080/01431160600904980. [DOI] [Google Scholar]
  38. Reddy CS, Jha CS, Diwakar PG, Dadhwal VK. Nationwide classification of forest types of India using remote sensing and GIS. Environ Monit Assess. 2015;187:1–30. doi: 10.1007/s10661-015-4990-8. [DOI] [PubMed] [Google Scholar]
  39. Roy B, Kasemi N. Monitoring land Use/Cover change and urban sprawl using Remote Sensing Data: a study of Siliguri and Raiganj Urban Agglomerations, India. In: Bhunia GS, Chatterjee U, Lalmalsawmzauva K, Shit PK, editors. Anthropogeomorphology. Geography of the physical environment. Cham: Springer; 2022. [Google Scholar]
  40. Sahoo S, Majumder A, Swain S, Pateriya B, Al-Ansari N. Analysis of Decadal Land Use Changes and its impacts on Urban Heat Island (UHI) using Remote sensing-based Approach: a Smart City Perspective. Sustainability. 2022;14(19):11892. doi: 10.3390/su141911892. [DOI] [Google Scholar]
  41. Saravanan S, Jennifer JJ, Singh L, Thiyagarajan S, Sankaralingam S. Impact of land-use change on soil erosion in the Coonoor watershed, Nilgiris mountain range, Tamil Nadu, India. Arab J Geosci. 2021;14:407. doi: 10.1007/s12517-021-06817-w. [DOI] [Google Scholar]
  42. Sarma PK, Sarma K, Kumar B, Aaranyak T. Tracing out of elephant corridors and landscape dynamics of Eastern Assam using geospatial tools: a case study in Tinsukia District of Assam, India Habitat View project Rhino reintroduction programme in Dudhwa NP UP View project. Eco Env & Cons. 2021;27:178–186. [Google Scholar]
  43. Tea Board of India (2021) Available online: www.indiatea.org
  44. Tridawati A, Wikantika K, Susantoro TM, Harto AB, Darmawan S, Yayusman LF, Ghazali MF (2020) Mapping the Distribution of Coffee Plantations from Multi-Resolution, Multi-Temporal, and Multi-Sensor Data Using a Random Forest Algorithm. Remote Sens 12(23):3933. 10.3390/rs12233933
  45. Wang B, Li J, Jin X, Xiao H. Mapping tea plantations from multi-seasonal Landsat-8 OLI Imageries using a Random Forest Classifier. J Indian Soc Remote Sens. 2019;47:1315–1329. doi: 10.1007/s12524-019-01014-5. [DOI] [Google Scholar]
  46. Wijeratne MA, Anandacoomaraswamy A, Amarathunga MKSLD, Ratnasiri J, Basnayake BRSB, Kalra N. Assessment of impact of climate change on productivity of tea (Camellia sinensis L.) plantations in Sri Lanka. J Natl Sci Foundation Sri Lanka. 2007;35:119–126. doi: 10.4038/jnsfsr.v35i2.3676. [DOI] [Google Scholar]
  47. Xu W, Qin Y, Xiao X, Di G, Doughty RB, Zhou Y, Zou Z, Kong L, Niu Q, Kou W. Quantifying spatial-temporal changes of tea plantations in complex landscapes through integrative analyses of optical and microwave imagery. Int J Appl Earth Obs Geoinf. 2018;73:697–711. [Google Scholar]
  48. You W, Ji Z, Wu L, Deng X, Huang D, Chen B, Yu J, He D. Modeling changes in land use patterns and ecosystem services to explore a potential solution for meeting the management needs of a heritage site at the landscape level. Ecol Ind. 2017;73:68–78. doi: 10.1016/j.ecolind.2016.09.027. [DOI] [Google Scholar]
  49. Zhu J, Pan Z, Wang H, Huang P, Sun J, Qin F, Liu Z (2019) An improved multi-temporal and multi-feature tea plantation identification method using sentinel-2 imagery. Sensors 19:2087 [DOI] [PMC free article] [PubMed]

Articles from Tropical Ecology are provided here courtesy of Nature Publishing Group

RESOURCES