Introduction

The distribution of terrestrial ecosystems on global and regional scales, including their configuration, existence, and mechanisms, is influenced by climatic conditions (Delire et al., 2011). The natural cover of vegetation is a vulnerable and unstable part of any ecosystem. Vegetation cover is a key element of the Earth’s terrestrial ecosystems, as it helps in soil conservation, the energy exchange between tropics, and various biogeochemical processes on the Earth (Liu et al. 2018). The importance of natural vegetation cover in the study of climate stability management is important, as it regulates the carbon balance, improves global greenhouse gas (GHG) emissions, and connects the atmosphere, water, and soil (Chuai et al. 2013; Kumar and Kumar 2020; Pandey et al., 2021; Puppala and Singh, 2021). However, it is susceptible to change in the climate. One of the most critical aspects of environmental change in terrestrial ecosystems is plant patterns and their reactions to climate change; while studying the internal interaction between vegetation and surrounding changes, the temporal coverage of vegetation cover had immense significance, exposing the evolution process of the condition of the regional environment and future forecast trend of the region (Zhang et al., 2003). The coverage of natural vegetation is also another important and recognizable measure of seasonality and interannual change (Cui and Shi 2010; Pandey et al., 2021; Puppala and Singh, 2021) and a sign of global level changing of climate (Weiss et al. 2004).

The development in the technology of remote sensing science provides valuable sources of information about different ecosystems, including terrestrial ecosystems. With the use of remote sensing technology, it is possible to analyze historical data available in the form of satellite imagery to monitor and map the vegetation cover and find the changes that occur from different periods. The remote sensing technology uses different electromagnetic energy, ranging from visible range to microwave spectrum range, for detecting the surface features. It provides advanced statistical approaches for optimizing classifying the surface objects such as different spectral indices, namely, normalized difference vegetation index (NDVI) for vegetation, normalized difference-built index (NDBI) for built-up, band rationing, and principal component analysis. Most of the indices are calculated by using digital brightness values of the satellite imagery. For monitoring and mapping vegetative cover, the statistical approach plays a vital role in deriving NDVI spectral indices. It is one of the ideal indicators of vegetation growing (Piao et al. 2015), which represents the growing and spatial pattern of existing vegetation cover on the ground (Purevdorj et al. 1998; Liu et al. 1999; Pandey et al., 2021) and morphology of plants (Derrien et al. 1992; Defries and Townshend 1994). NDVI value ranges from − 1 to 1 in which the maximum positive values indicate the healthier vegetation cover, values near to 0, and decreasing negative values indicator of non-vegetation cover type surface feature, i.e., bare rock, soil surface, water, and ice (Schnur et al. 2010; Chuai et al. 2013). Many factors that affect the vegetation, such as metrological variables (i.e., precipitation and temperature), directly influence the vegetation mechanisms (Bengang and Shu, 2000; Nicholson et al. 1990; Pandey et al., 2021; Zhang et al. 2003) and spatial–temporal land use land cover change (LULC) factors (Melese, 2016). Changes in the land use land cover pattern greatly impact the terrestrial ecosystem, such as a change in the functioning of forestry, degradation of natural habitat, and agriculture, which have been rising issues globally due to national and regional development initiatives. Land fertility, such as forest cover, can decline as a result of LULC under agriculture, soil erosion, and overgrazing (Melese 2016), which are also the causes of climate change (Hailemariam et al., 2016). LULC research aims to understand better the mechanisms, patterns, and complexities that influence land-use transformations and the forces that drive those (Nduati et al., 2013). The decline of vegetation cover creates the threat of atmospheric conditions, increased desertification, displacement of people, and biodiversity loss (Fekadu, 2015; Gebru, 2016).

This study evaluated causes of change in the vegetation cover (VCC) in the Pindari watershed of the upper Himalayas in the Uttarakhand state of India based on climate parameters using remote sensing and statistical analysis (annual precipitation, minimum and maximum temperature from 1972 to 2018). We find that the main reason behind the VCC in the study site is deforestation and anthropogenic activities. This research is helpful for country planners to make the action plan for conserving the forest ecosystem at the ground level and making more effective master plans. The present study suggests analyzing the LULC in the upper Himalayan region for further analysis using remote sensing technology.

Study area

Uttarakhand is one of the Himalayan states of India, located in the northwestern part of the country. It is bordered to the northwest by the Indian state of Himachal Pradesh, to the northeast by the Tibet Autonomous Region of China, the southeast by Nepal, and the south and southwest by the Indian state of Uttar Pradesh. Uttarakhand experiences great metrological conditions (weather and climatic). The study area, Pindari valley (a buffer zone of Nanda Devi Biosphere Reserve), is situated in the Bageshwar district of Uttarakhand, Western Himalaya. It extends geographically between 30°6′15.429″ N 30°19′15.45″ N latitudes and 79°47′57.434″ E to 80°5′41.672″ E longitudes and encompasses an area of 391.05 km2. The region receives heavy rainfall during the monsoon season. The approximate variation in the precipitation is from 750 to 2000 mm. Figure 1 illustrates the study site of the Pindari watershed region in the Kumaun Himalayan part of Uttarakhand state of India. Pindari glacier is one of the essential glaciers of the watershed located in the Baleshwar district of Uttarakhand, which is placed in the upper lesser Himalayan range and along with neighbor to Chhanguch glacier.

Fig. 1
figure 1

Location of the Pindari watershed in the northern part of the lesser Himalayas

Full size image

Pindari river is one of the important rivers of Uttarakhand state that originates from the Pindari glacier. It is the biggest tributary of India’s most prominent river, the Ganges. The study site faces mixed climatic conditions, ranging from sunny to heavy snowfall (Pandey et al., 2018). Major peaks near the Pindari glacier are Pawalidwar (6663 m), Nandakhat (6611 m), Baljuri (5922 m), Nandakot (6860 m), and Changuch (6322 m, unclimbed peak).

Data and methodology

In this study, we used Landsat satellite data for extracting the vegetation cover of the study site. Table 1 shows the description of the remotely sensed datasets used for conducting this study.

Table 1 Description of the dataset used in this study
Full size table

This study includes multi-spectral Landsat satellite datasets of different years, which are used to classify the vegetation in the Pindari watershed of the upper Himalayas. Landsat 1 (MSS) was used for the year of 1972, Landsat 7 (ETM +) for 1999, Landsat 5 (TM) for 2011, and Landsat 8 (OLI) for 2018. For the metrological trend analysis, we included power data access viewer datasets from 1981 to 2018 collected from the official web page of NASA’s platform of power data access viewer. We used the preparation and temperature dataset from the power data access viewer portal. Finally, for the ground truth data for validating the classified output result, we used GPS devices and Google Earth high-resolution imagery.

Preprocessing

Preprocessing is the first step of putting remotely sensed data to use in an application. It includes radiometric and geometric correction. All the Landsat imageries were firstly applied to both operations before classifying the vegetation cover. We first applied the TOA (top of atmosphere) operation on all obtained Landsat imagery for the radiometric correction by following Eq. (1) shown below.

$${\rho }_{\lambda }=\frac{\pi *{L}_{\lambda }*{d}^{2}}{{ESUN}_{\lambda }*{cos\theta }_{s}}$$
(1)

where \(d\) is the Earth–Sun distance in astronomical units, \({\rho }_{\lambda }\) is the unitless planetary reflectance,\({\theta }_{s}\) is the solar zenith angle, \({L}_{\lambda }\) is the spectral radiance, and \({ESUN}_{\lambda }\) is the mean solar exo-atmospheric irradiances.

The reflectance measured by a space-based instrument traveling higher than the Earth’s atmosphere is known as top-of-atmosphere reflectance. These reflectance values will contain influences from clouds, aerosols, and chemicals in the atmosphere. TOA reflectance can be derived using Eq. (1).

After the radiometric correction, we geometrically correct the satellite imagery with the help of a toposheet at the scale of (1:50,000) issued by the national survey of India (survey of India). We adopted the sampling method of the nearest neighbor for geometrically correcting the satellite images. All the satellite imagery was corrected geometrically by fixing the root mean square error (RMSE) < 0.6 pixels. After both correction operations, the imageries were layer stacked; masked were applied using ERDAS Imagine geospatial software for extracting the ground cover vegetation of the study site, bands 4 to 7 for Landsat 1 (MSS) and bands 1 to 5 and band number 7 for Landsat 7 (ETM +); the same as for the Landsat 5 TM and Landsat 8 (OLI), bands 1 to 7 were stacked and then masked for extracting the area of interest.

NDVI classification

The NDVI data is synthesized based on different years of Landsat and sentinel datasets from the preprocessed period, and the NDVI dataset was retrieved. A method was used to maximize the reflectance energy recorded in the bands of near-infrared (NIR) and visible red bands using Eq. (2):

$$NDVI=\frac{{\mathrm{NIR}}_{B}-{\mathrm{RED}}_{B}}{{\mathrm{NIR}}_{B}+{\mathrm{RED}}_{B}}$$
(2)

NDVI is one of the most commonly used spectral indices for determining the ground vegetation cover. It calculates the greenness of plants using the Red and NIR spectral bands of satellite images, which can then be applied and analyzed over several seasons of selected years (Babalola and Akinsanola, 2016). The values of the NDVI range from 1 to − 1, with the maximum positive values representing healthier vegetation cover and values nearing 0 to negative representing non-vegetation land cover types. Here, in the present study, different thresholds were applied of the calculated NDVI of the year 1972, 1999, 2011, and 2018 based on collected ground truth data by making different field visits (2018) and Google Earth imagery points (1972, 1999, and 2011) for the previous year imagery purpose to extract the vegetation cover over the Pindari watershed. At different NDVI threshold values, such as below 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, and above 0.5, the NDVI approach is used depending on vegetation characteristics (Pandey et al., 2021). Every class (forest cover, grassland, and scrubland) has an NDVI threshold value assigned to it. The first NDVI threshold value is estimated, which is used for classification.

Analysis of trend using Sen’s slope test and Mann–Kendall test of the statistical approach

A trend analysis is described as a change in a variable that can be observed using both parametric and nonparametric statistical methods over time (Soni, 2017). The present study included the trend analysis on precipitation and temperature data from 1981 to 2018, obtained from NASA’s power data access viewer platform. The above variables’ long-term trend was analyzed by performing the rank-based nonparametric Mann–Kendall trend test (Mann, 1945; Kendall, 1948). The Mann–Kendall test revealed a trend of decreasing, increasing, or non-existent on the metrological variables of temperature and precipitation in the study timeline. The equations for the test of Mann–Kendall S, V (S), and standard Z are performed by applying Eqs. (3) to (6):

$$S= {\sum }_{i=j}^{n-1}{\sum }_{j=i+l}^{n}sig\left(X_{j}-X_{i}\right),$$
(3)
$$sign(y_{j}-y_{i})= \left\{\begin{array}{c}1, if\left(y_{j}-y_{i}\right)>0\\ 0 , if\left(y_{j}-y_{i}\right)=0\\ -1, if\left(y_{j}-y_{i}\right)<0\end{array}\right.$$
(4)
$$V \left(S\right)= \frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-{\sum }_{p=1}^{q}{t}_{p}({t}_{p}-1)(2{t}_{p}+5)\right],$$
(5)
$$Z=\left\{\begin{array}{c}\frac{S-1}{\sqrt{VAR\left(S\right)}}if\;S>0\\0\;if\;S=0\\\frac{S+1}{\sqrt{VAR\;\left(S\right)}}if\;S<0\end{array}\right.$$
(6)

Here, Xj and Xi represent the sequential data of the different climatic variables. The length of a time series is represented as n, \({t}_{p}\) represents the number of ties for the \(p\mathrm{th}\) values, and q is the number of tied values. Positive test results sign of the upward trend of the variable and value of negative Z indicate that the variable trend is downward. If the Z is equal to zero, it represents there is no trend of the variable.

Sen’s slope test

This test is performed for majoring the magnitude of the trend at given time series data (Yaseen et al., 2013). The following equation is used for calculating Sen’s slope test.

$${T}_{i}= \frac{{X}_{j }- {X}_{k}}{j-k}$$
(7)

\({X}_{j}\) and \({X}_{k}\) indicate the values of data at j and k time, respectively. Also,

$$Q_i=\left\{\begin{array}{c}T_\frac{\left(N+1\right)}2if\;N\;is\;odd\\\frac{\left(T_{\left(N/2\right)}\right)+\left(T_{\left(N+2\right)/2}\right)}2if\;N\;is\;even\end{array}\right.$$
(8)

The positive Qi values indicate an increasing trend. On the other hand, the negative Qi value indicates a declining trend over time.

Results

Change of vegetation cover and NDVI evaluation

The NDVI was used to calculate the development of green vegetation and to identify changes in vegetation. It is often used to quantify the relationship between atmospheric variation and the amount of vegetation status. Following the classification between low and high vegetation, a pixel value with a greater NDVI value implies a high concentration of vegetation. In comparison, a pixel with a lower NDVI value usually indicates less concentration of vegetation (Fig. 2). Due to annual variation in the growing season, NDVI for different vegetation types is given in Fig. 2. Calculated NDVI values of the year 2018 (A), 2011 (B), 1999 (C), and 1972 (D) of the study locations range from − 0.16 to 0.41, − 0.39 to 0.67, − 0.47 to 0.64, and − 0.83 to 0.66, respectively. The maximum NDVI value was noticed in 2011 at about 0.67, and on the other hand, the lowest positive NDVI value was noticed in 2018 at about 0.41 (Table 2). On behalf of NDVI result outcomes, it has been noticed that the overall positive value of NDVI decreased from 0.66 in 1972 to 0.41 in 2018.

Fig. 2
figure 2

NDVI result of the years (A) 2018, (B) 2011, (C) 1999, and (D) 1972 of Pindari watershed of upper Himalaya Kumaun, Uttarakhand, India

Full size image
Table 2 Normalized difference vegetation index values for the study period of 1972, 1999, 2011, and 2018
Full size table

The variability of the NDVI corresponded well with the seasonal precipitation in the region. The amount of precipitation was comparatively higher in 1972 and 2011, with values of 0.66 and 0.67, respectively, and lower in 1999 and 2018, with values of 0.66 and 0.67.

Forest cover change during the period (1972–2018)

Forest cover has been decreased by 24 sq. km from 1972 to 2018. The estimated forest cover was 110.80 sq. km, 95.68 sq. km, 92.81 sq. km, and 86.06 sq. km for the years 1972, 1999, 2011, and 2018, respectively (Table 3). Minimum elevation of treeline remains the same from the year 1972 to 2018; on the other hand, maximum elevation of treeline increased histrionically from 3233 to 4779 m, and the total increases in maximum height of treeline were 1546 m while the continuous fall of the area of forest is clear from Table 3 and Fig. 3A to D. The total areal decrease of the forest is 24.74 sq. km (Table 3).

Table 3 Spatiotemporal changes of the forest, grassland, and scrubland cover from 1972 to 2018 of the Pindari watershed
Full size table
Fig. 3
figure 3

The forest cover of the Pindari watershed for the years (A) 1972, (B) 1999, (C) 2011, and (D) 2018. Grassland of the Pindari watershed for the years (E) 1972, (F) 1999, (G) 2011, and (H) 2018. Scrubland of the Pindari watershed for the years (I) 1972, (J) 1999, (K) 2011, and (L) 2018

Full size image

Grassland change during the period (1972–2018)

High altitude natural alpine grasslands are a vital part of the Montane grasslands and shrublands biomes. Melting of snow and upward moving snow lines provide suitable habitation to nurture these grasslands in the Himalayas. In the Upper Pindari region of Uttarakhand, lower and upper, both limits of grasslands are increasing continuously from the year 1972 to 2018 (Fig. 3E to H). It has been noticed that the grassland showed a positive change in land cover. Grassland got increased from 28.8 sq. km in 1972 to 46.64 sq. km in 2018. The geographical area cover of grassland was about 28.8 sq. km, 42.26 sq. km, 45.78 sq. km, and 46.64 sq. km for the years 1972, 1999, 2011, and 2018 respectively (Table 3). The minimum elevation of grasslands in 1972 was 2009 m which increased to 2058 m in the year 2018, while maximum elevation increased from 5208 to 5983 m in the same span. In the same period, the areal extent of grasslands in the upper Pindari glacier region increased from 28.80 to 46.64 sq. km.

Scrubland change during the period (1972–2018)

Scrubland is a mature vegetation type that remains permanent over time in a specific area or a transient population that exists temporarily due to a disturbance, such as fire. The grasslands form an important ecoregion in the Himalayas and are an essential component of biodiversity. Scrublands, dominated mainly by rhododendrons, predominate at lower elevations close to the actual treeline. They are a transitional plant community between grassland and trees. In the upper glacier region of Pindari, the minimum elevation of scrublands does not show any remarkable increase throughout the whole study span, while the maximum elevation of scrublands increased continuously: 4019 to 5568 m in the year 1972 to 2018. The area of scrubland also increased but not much noticeable as from 31.19 sq. km (in 1972) to 38.71 (in 2011) that further decreased to 34.32 sq. km in 2018 (Table 3 and Fig. 3I to L). Figure 4 shows the graphical representation of the geographical area of forest cover, grassland, and scrubland feature class.

Fig. 4
figure 4

Geographical area of the feature classes (forest cover, grassland, and scrubland)

Full size image

Trend analysis of climatic variables

The Mann–Kendall test showed that the trend of the annual mean maximum and minimum temperature and perception over the study timeline from 1981 to 2018 was statistically insignificant (Table 4). Here, in the test, we found that the annual mean minimum temperature and precipitation variables showed an upward trend. However, the annual mean maximum temperature showed a downward trend over the 37-year study period. The positive slope of the annual mean minimum temperature and precipitation data explains the nominal value increment through the chosen timeline review throughout the study site. Figure 5 illustrates a graphical representation of the annual mean maximum (A) and minimum (B) temperature and precipitation from 1981 to 2018. The Mann–Kendall test represents that the p values are − 0.96, 4.42, and 3.27 for annual mean maximum and minimum temperature and precipitation, respectively (Table 4). The test of Sen’s slope shows an upward shift trend in annual minimum temperature and precipitation data by 0.03 ℃ and 8.79 mm, respectively, and downward for the annual maximum temperature dataset by − 0.0009 ℃ (Table 4).

Table 4 Mann–Kendall statistical trend test of mean, maximum, and minimum temperature (a) and annual precipitation (b) from 1981 to 2018
Full size table
Fig. 5
figure 5

Graphical representation of the annual mean maximum (A) and minimum (B) temperature and precipitation (C) from 1981 to 2018

Full size image

Discussions

The spatiotemporal study of the Pindari valley vegetation cover reported valid changes over the 46 years from 1972 to 2018. There were noticed both negative and positive shifts. For the full timeframe, the forest region showed a decline in a broad geographical area. Simultaneously, there was a positive shift in some classes. There are three types of forests in the study area, viz., dense, scattered, and scrubland. The dense forest in the research area is being replaced by a sparse forest.

The growing population and building environment in this area has greatly impacted the natural environment, resulting in the depletion of natural forests and a shift in the region’s overall ecosystem. Over the past 46 years, it has been noticed that the forest cover area has decreased rapidly and grassland cover increased in the Pindari watershed of the upper Himalaya, Kumaun region of Uttarakhand, India (Table 5 and Fig. 6). The change in the geographical area of grassland feature class was about 13.46 sq. km, 3.52 sq. km, and 0.86 sq. km, for scrubland about 4.63 sq. km, 2.89 sq. km, and − 4.39 sq. km, over about − 15.12 sq. km, − 2.87 sq. km, and − 6.75 during the time of 1972–1999, 1999–2011, and 2011–2018, respectively.

Table 5 The net change in the feature class of forest cover, grassland, and scrubland from 1972 to 2018 (46 years) in the Pindari watershed
Full size table
Fig. 6
figure 6

Net geographical area changes of forest cover, grassland, and scrubland from 1972 to 2018 in the study location

Full size image

It has been noticed that the grassland and scrubland features showed an overall positive change from 1972 to 2018, about 17.84 sq. km for grassland and 3.13 sq. km for scrubland, and high negative change occurred in the forest cover over the study region, which was about − 24.74 sq. km. The overall geographical area and overall percent area that cover the study site are shown in Fig. 7 and Table 6.

Fig. 7
figure 7

Geographical area of different vegetation classes of study site from 1972 to 2018

Full size image
Table 6 Net area cover of forest cover, grassland, and scrubland from the different periods of the study site
Full size table

The highest geographical area was noticed 46.64 sq. km in 2018, 38.71 sq. km in 2011, and 110.8 sq. km in 1972 for the feature of grassland, scrubland, and forest cover, respectively.

Mann–Kendall and Sen’s slope mathematical analyses, the optimistic upward annual pattern of annual mean minimum temperature and precipitation, and on the other hand, the negative downward trend for annual mean maximum temperature from 1981 to 2018 of the study sites are shown in Table 6. It has also been noticed that the maximum positive values of NDVI by 0.66 in 1972 to 0.41 in 2018 (Table 2).

Land-use changes occur as a result of many interconnected causes, including changes in vegetation cover. With dramatic increases in temperature and precipitation, the climate in the study area was discovered to be the most dominant factor affecting land-use changes.

Conclusion

In this present study, the NDVI different values were included for extracting the different vegetation cover classes of the Pindari watershed of upper Himalaya in the Kumaun region of Uttarakhand, India. Threshold-based outcome results illustrated that the total area coverage under forest cover was 28.3%, 24.11%, 23.71%, and 21.98% during 1972, 1999, 2011, and 2018, respectively, which indicates the overall negative charge of − 24.74 sq. km from 1972 to 2018 in the land cover. Despite its insignificance, the study location has seen an increase in positive annual mean precipitation from 1981 to 2018 by 8 mm. The Mann–Kendall test was also used during the research, which revealed the upward trend in precipitation, but the annual mean maximum temperature was not statistically significant; however, the annual mean minimum temperature showed a positive upward trend of temperature during the timeline of the dataset over the study site. For the grassland, the area coverage was 7.36% in 1972, 10.8% in 1999, 11.69% in 2011, and 11.91% in 2018, indicating a positive change of about 17.84 sq. km, respectively. The leading causes behind decreasing forest cover and increasing grassland cover are anthropogenic activities; increases in population directly raise the demand for extra land for cultivation and land for the establishment of settlement over in a given study region. These requirements are fulfilled by removing the forest cover, especially the natural forest cover in the area.