Introduction

Wetlands are among the most threatened habitats of the world (Vörösmarty et al. 2010). Wetlands in India, as elsewhere, are increasingly facing several anthropogenic pressures (Ozesmi and Bauer 2002; Dhillon and Mishra 2014). The rapidly expanding human population (Keddy 2010), large-scale changes in land use and land cover (Jenerette and Wu 2001), unplanned infrastructure development and encroachment (Romshoo and Rashid, 2014), landscape degradation (Joshi et al. 2002), and water pollution (Najar and Khan 2012) affect the health of wetlands. These anthropogenic pressures have accelerated the wetland degradation by impairing their ecohydrological and socioeconomic functionalities (Bourgeau-Chavez et al. 2001).

Different wetland types, due to their location, are affected by multiple drivers of change and therefore require tailored approaches to address human impacts (Zedler and Kercher 2005). Any alteration in the hydrological and sedimentation regimes results in wetland degradation. Urbanization increases run-off peak flows and total flow volumes, and deteriorates the otherwise pristine water besides affecting the esthetic setting through conversion of wetland area into impervious surfaces (Azous and Horner 2000; Romshoo et al. 2012). The intensities of these anthropogenic disturbances vary along different spatial and temporal scales (Chopra et al. 2001; Lindegarth and Hoskin 2001; Rashid and Romshoo 2013). These anthropogenic impacts lead to increased sediment loads (Badar et al. 2013a; Jain et al. 2001; Prachansri 2007; Romshoo and Muslim 2011; Zhou et al. 2008), water quality deterioration (Rashid et al. 2017b), and changes in aquatic biodiversity (Khan et al. 2004) thereby affecting wetland health. In addition, the rising temperatures in Kashmir have adversely affected the snow cover distribution and associated hydrological regimes that have a direct bearing on wetland hydrology (Fowler and Archer 2006; Rashid et al. 2017c; Rashid and Majeed 2018). There is no change in the annual precipitation (Rashid et al. 2015); however, owing to the rising temperatures, the form of winter precipitation is speculated to have changed (Romshoo et al. 2015). Due to human-induced modification, the natural wetland landscapes also are under acute pressures (Rashid et al. 2013). Wetland areas across Kashmir have been gradually squeezing affecting the buffering capacity of these ecosystems to withhold flood waters and storm water runoff (Rashid and Naseem 2007). This was clearly manifested during the 2014 devastating flood when residential area of the outskirts of capital city Srinagar, which used to be traditional floodplain, was inundated for more than 3 weeks in flood waters (Romshoo et al. 2017). The problem of wetland degradation in the region is further exacerbated by ill-conceived policy mechanism and lenient legal framework.

Keeping this in context, our study assessed the land system changes in and around Narkara wetland over 52 years between 1965 and 2016 by using high-resolution earth observation data. To our knowledge, it is for the first time that high-resolution satellite data has been used for assessing historical changes in land use land cover of a wetland in Kashmir region. The present study covers important aspects of monitoring the wetland health by assessing the catchment-scale land system changes and impact of such changes on soil erosion. The present is significant keeping in view the brunt of reckless unplanned urbanization (goo.gl/t4oScU) that Narkara wetland is facing and also that no major scientific work has reported historical changes in the land system of this semi-urban wetland.

Material and methods

Study area

Narkara wetland is located in the south western outskirts of Srinagar, the capital of north most state of India, Jammu and Kashmir (Fig. 1). The wetland catchment, spread over an area of 40.19 km2, is located between geographical coordinates 34°03′03″–33°02′30″ N latitude and 74°46′10″–74°45′54″ E longitude with altitude range from 1558 to 1678 m amsl. It receives the waters from Sukhnag nallah, a left bank tributary of Jhelum River. The catchment of Narkara is predominantly a semi-urban setting with a lot of area under settlements, agricultural fields and table lands locally called karewas which are barren denuded landscapes. The wetland is a breeding ground for water fowl species that migrate from Russia and Central Asia especially during winters. Over the last few decades, the massive unplanned urbanization and encroachments have resulted into a substantial area under marshland getting converted into concrete built-up areas.

Fig. 1
figure 1

Location of study area

Full size image

Data sets

A repository of high-resolution earth observation data, meteorological data, soil data, ground truth, and field photographs have been analyzed in GIS to accomplish this study. The details of these datasets are provided in Table 1.

Table 1 Datasets used for accomplishing the study
Full size table

Image geometric correction

The high-resolution IKONOS data (0.5 m) of 2016 comes up as georeferenced basemap freely accessible from ESRI ArcMap 10.1, while the declassified Corona data (KH-4A and KH-9/16) comes up as raw non-georeferenced product freely available from https://earthexplorer.usgs.gov/. Owing to the geometrical distortion (Schmidt and Nüsser 2012), Corona images were co-registered using projective transformation in ERDAS Imagine 9.0 (Jensen 2009; Rashid et al. 2017a). The images were coregistered based on 23 control points (Fig. 2) whose coordinates were taken from IKONOS imagery available as ESRI Basemap in Arc Map 10.1. Prominent road-stream junctions, airport runway, old building, and bridges were used for GCPs assuming that no changes occurred for these points on the ground during the observation period. Additionally, 20 other common features were carefully identified in each of the Corona images and IKONOS data. The horizontal shift (rot mean square error) in the two Corona images was ± 4.6 m and ± 3.8 m with respect to the IKONOS image that served as reference data.

Fig. 2
figure 2

Location of control points and tie points for georeferencing Corona satellite data

Full size image

Quantifying the land system changes

High-resolution earth observation data of 1965, 1980, and 2016 were used to produce multi-date land use land cover (LULC) maps of Narkara catchment. The study area was extracted using extract by mask tool in Arc Map 10.1, whereas 2016 satellite data was used by accessing high-resolution images from Arc Map 10.1. Owing to high resolution of earth observation data, LULC types were delineated using onscreen digitization at the scale of 1:5000 in Arc Map 10.1. Visual image interpretation was preferred over digital classification approach as it is advantageous for delineating LULC in mountainous areas primarily as it incorporates cognitive inputs from image analyst (Rashid et al. 2010, 2013, 2016). Total area under each land cover category was calculated for 1965, 1980, and 2016 which was followed up by change detection analysis. Survey of India topographical sheets, surveyed in 1962, were used for accuracy assessment of 1965 Corona generated land cover data. Since no reference data was available for 1980s, we could not carry out the accuracy assessment for 1980 land cover data. Accuracy assessment of land cover types delineated from 2016 satellite data is based on 117 well distributed field samples collected using Trimble Juno SB GPS with a positional accuracy of 5 m (Fig. 3). Kappa coefficient (Eq. 1) was used for assessing the accuracy of delineated land cover types of 1965 and 2016 based on 117 samples each from topographical map and ground survey as:

$$ k=\frac{N\sum \limits_{i=1}^r{X}_{ii}-\sum \limits_{i=1}^r\left({X}_{1+}.{X}_{+i}\right)}{N^2-\sum \limits_{i=1}^r\left({X}_{i+}.{X}_{+i}\right)} $$
(1)

where r is the number of rows in error matrix, Xii is the number of observations in row i column i (on the major diagonal), Xi+ is the total of observations in row i (shown as marginal total to right of the matrix), X+i is the total of observations in column i (shown as marginal total at bottom of the matrix), and N is the total number of observations included in the matrix.

Fig. 3
figure 3

Location of field points for validating land use land cover maps of Narkara catchment

Full size image

Additionally, user’s accuracy, producer’s accuracy, errors of omission, and commission were also computed to assess the accuracy of the land use and land cover data of 1965 and 2016. Kappa is lower than overall accuracy. The difference between kappa and overall accuracy results due to the fact that each index uses different information from the error matrix. While overall accuracy only includes the data along the major diagonal, kappa incorporates the non-diagonal elements of error matrix as a product of row and column marginal (Romshoo and Rashid 2014).

Erosion assessment

Among soil loss estimation models, universal soil loss equation (USLE) is considered one of the robust models and is being used worldwide for estimation of soil erosion (Wischmeier and Smith 1978). A revised version of this model, Revised USLE (RUSLE), has enhanced soil loss predication capabilities (Renard et al. 1997). RUSLE predicts the long-term average annual rate of soil erosion (DAIS 2008) in variety of environments such as agriculture, forest, rangeland, mining sites, construction sites, etc. (Stone and Hilborn 2000) which means that it could be very well used for estimating soil erosion at catchment scale (Millward and Mersey 1999; Teng et al. 2018; Fang et al. 2019). RUSLE computes soil erosion using the following equation:

$$ A=R\times K\times LS\times C\times P $$
(2)

where A is the average annual predicted soil loss from sheet and rill erosion (tons/ha/year), R is the rainfall/runoff erosivity (MJ mm/ha/h/year), K is the soil erodibility (Mg h/MJ/mm), LS is the slope length and steepness/topographic factor (dimensionless), C is the crop management factor (dimensionless), and P is the support practice (dimensionless).

Using Eq. 2 above, we assessed soil loss in the catchment of Narkara wetland for 1965, 1980, and 2016. The generation of input datasets for RUSLE was generated in Arc Map 10.1. The steps involved in preparation of input datasets for RUSLE have been described in the following sections:

Rainfall erosivity (R-factor) depends on the intensity and duration of rainfall. Precipitation data with a spatial resolution of 1 km2 for the study area was extracted from WorldClim version1 (Accessible at http://worldclim.org/version1). Meteorological data from Worldclim has been extensively used for mapping and spatial modeling (Hijmans et al. 2005). Data of mean annual precipitation from 1960 to 1990 was used for the calculation of R factor keeping in view that no significant changes in precipitation have been reported in the region (Murtaza and Romshoo 2017). R factor was derived from Eq. 3 by Singh et al. (1981) as:

$$ {R}_{fac}=79+0.363R $$
(3)

where Rfac is the rainfall erosivity and R is the annual average rainfall in millimteter. Although it was prudent to use meteorological observations, however, owing to the scanty network of weather stations, we used gridded precipitation data from WorldClim which showed good agreement with rainfall observation station located at Rambagh, Srinagar. It must be noted that there is no weather station in the catchment of Narkara wetland. The entire mountainous Kashmir Valley spreading over ~ 15,500 km2 has just 6 meteorological observatories, so the data are scanty in terms of spatial distribution, because of the terrain complexity of the region (Romshoo et al. 2018). In order to overcome this impediment, it was prudent to use gridded WordClim data that matched well (showing annual precipitation of 711 mm) with the nearest observatory located at Srinagar (with an annual precipitation of 723 mm). The annual precipitation at Srinagar station maintained by India Meteorological Department is 723 mm against 711 mm from the WorldClim data indicating an underestimation of 12 mm.

Soil erodibility (K-factor) is a measure of susceptibility of soil particles to detachment and transport by runoff. An important prerequisite for estimating soil erodibility is the soil texture. Soil texture information at a spatial resolution of 100 m was compiled from data provided by Sheri Kashmir University of Agricultural Science and Technology (SKUAST). The soil erodibility map was finally prepared by assigning K-value (Table 2) to the respective soil texture types from Stone (2012).

Table 2 Soil types and their K-values as per Stone (2012)
Full size table

Topographic (LS) factor represents a combined effect of length and steepness of slope (Roose 1977). The gentler and shorter the slope, the lower is the risk of erosion and vice versa. Using SRTM DEM with a spatial resolution of 30 m, LS factor was computed using the method provided by Mitasova et al. (2001) as:

$$ LS={\left(l/72.6\right)}^m\left(65.41{\sin}^2\beta +45.56\sin \beta +0.065\right) $$
(4)

where l is the cumulative slope length in meters and β is the downhill slope angle, and m is the slope contingent variable (0.5 if the slope angle is greater than 2.86°, 0.4 on slopes of 1.72° to 2.86°; 0.3 on slopes of 0.57° to 1.72°; 0.2 on slopes less than 0.57°).

Support practice factor (P) reflects the effect of practices that reduce the runoff rate and check the soil loss (Stone and Hilborn 2000). The P-factor was computed from the slope and LULC as suggested in Kumar et al. (2014).

Crop management factor (C) represents the relative effectiveness of soil and crop management systems in terms of LULC practice with respect to preventing soil loss (Roose 1977). Among various RUSLE factors, the C factor has been considered as the most important as it is a conservation related factor that controls soil loss at a specific site (Teh 2011). To prepare the C-factor map, different LULC which were identified in 1965, 1980, and 2016 data were assigned C-value after Patil et al. (2015). LULC-wise C and P values are shown in Table 3.

Table 3 LULC-wise C- and P values after (Kumar et al. 2014; Patil et al. 2015)
Full size table

All the above derived input parameters for RUSLE (R, K, LS, C, and P) were generated for the estimation of soil loss in the study area at different points in time.

Results and discussion

LULC dynamics

Six land use cover classes, which include Agriculture (AG), Aquatic vegetation (AV), Bare land (BL), Built-up (BU), Plantation (PL), and Water (WT), were delineated (Table 4) using on-screen digitization of high-resolution earth observation data of year 1965 (Fig. 4a), 1980 (Fig. 4b), and 2016 (Fig. 4c). From the analysis of 2016 land cover data, the catchment of Narkara wetland is predominantly an urban setting with built-up areas covering 37.12% of the total catchment area followed by Bare land (26.45%), Agriculture (12.47%), Plantation (12.22%), Aquatic vegetation (11.45%), and Water (0.30%). Built-up areas are mainly residential; however, Srinagar airport, associated military installations, and few government offices fall within the catchment of wetland. There are a lot of open spaces categorized as Bare land within the study area mostly under airport limits. Agriculture land is mainly composed of rice paddies and vegetable gardens. Plantation comprises of willows, poplars, and sporadic apple orchards. Aquatic vegetation restricted to Narkara catchment is mainly composed of watermoss, water lily, lotus, and invasive water fern, Azolla. However, the shallow waters on the periphery are dominated with cattail and common reed. Water is represented by a feeding Sukhnag tributary and a flood spill channel constructed after 1960s. Our analysis indicated substantial land system changes over the 52-year observation period between 1965 and 2016. Agriculture decreased from 22.63 km2 (in 1965) to 5.01 km2 (in 2016) indicating a 77.86% decrease; however, the decrease in agriculture area was more pronounced (53.02%) during 1980–2016 period. Massive increase (~ 2663%) in built-up areas were manifested between 1965 and 2016 owing to reckless urbanization in Narkara. The pace of urbanization exacerbated between 1980 and 2016 period as indicated by 2394% increase in built-up areas compared to 268% increase between 1965 and 1980 period. Similar urbanization patterns have been reported by Badar et al. (2013a) and Rashid et al. (2017a), (c) around wetlands and lakes in Kashmir. Area under plantation has increased by 55.87%, while bare lands, water, and aquatic vegetation are relatively stable between 1965 and 2016. A small increase in water area is attributed to construction of alternative flood spill channel after 1965 in the study area. Overall accuracy and class-wise accuracies for 1965 and 2016 data are provided in Tables 5 and 6. The accuracy of year 1965 land use classes was assessed using the Survey of India topographical map.

Table 4 LULC changes from 1965 to 2016 in the catchment of Narkara wetland
Full size table
Fig. 4
figure 4

High-resolution land use land cover maps of Narkara wetland catchment for a 1965, b 1980, and c 2016

Full size image
Table 5 Accuracy Assessment of 1965 LULC dataset
Full size table
Table 6 Accuracy Assessment of 2016 LULC dataset
Full size table

The overall accuracy of the 1965 land cover data was 87.18% (Table 5) with AV and WT having 100% accuracy followed by BU (86.49%), AG (85.71%), BL (85.71%), and PL (80%). The error of omission, i.e., probability of excluding a pixel that should have been included in the class was highest for AV (28.58%) followed by AG (25%), BL (20%), PL (4.76%), BU (3.03%), and WT (0%). Similarly, the error of commission which is the probability of including a pixel in a class when it should have been excluded was highest for PL (20%) followed by AG (14.29%), BL (14.29%), BU (13.51%), AV (0%), and WT (0%).

The accuracy of 2016 land cover data was assessed using field data of 117 ground sample points (Table 6). The overall accuracy of 2016 land cover data was 94.87% with AV and WT having 100% accuracy followed by BU (97.3%), PL (96.00%), AG (90.48%), and BL (85.71%). The higher accuracies of 2016 LULC data could be attributed to multispectral satellite data (Gao 1999; Khanna et al. 2018) and better spatial resolution (Hsieh et al. 2001) used for mapping in this study as compared to greyscale data of CORONA for 1965. The error of omission was highest for AG (9.52%) followed by AV (9.09%), BL (7.69%), PL (4%), BU (2.7%), and WT (0%). Similarly, the error of commission was highest for BL (12.49%) followed by AG (9.52%), PL (4%), BU (2.7%), AV (0%), and WT (0%). The kappa statistics for 1965 and 2016 generated land cover data are 0.84 and 0.93, respectively.

By analyzing the multi-date LULC maps, it is clear that the expansion of built-up areas has resulted into shrinkage of agriculture fields. Similar observations have been reported by Romshoo et al. (2011) around a neighboring Hokersar wetland. Lack of a comprehensive wetland conservation scheme has turned the former wetland areas into concrete residential jungle mainly due to encroachments by the land grabbers (Parihar et al. 2013). It is pertinent to mention that built-up areas increased 28 times in the catchment of this semi-urban wetland during the 52-year observation period which could have a direct impact on the nutrient flux, eutrophication, sedimentation, and plant invasions (Khuroo et al. 2007; Badar et al. 2013b; Amin et al. 2014). Increase in Orchard land could be attributed to the fact that farmers have shifted from traditional rice paddy culture to orchard cultivation in view of depleting stream flows and economic considerations (Showqi et al. 2014; Rather et al. 2016). Further, the symptoms of the wetland deterioration are attributed to nutrient enrichment resulting due to the reckless use of fertilizers and pesticides from agriculture and orchard lands that pave way for proliferation of invasive aquatic species in wetlands across Kashmir (Pandit and Kumar 2006; Mushtaq et al. 2015); however, it needs to be researched.

Erosion estimates

The input datasets, R, K, LS, C, and P factors (Figs. 5, 6, and 7), were generated using rainfall, soil, SRTM DEM, and land cover data. LS (Fig. 5a) and K (Fig. 5c) factors are also assumed to remain constant throughout the analysis period. The R factor (Fig. 5b) has been kept constant for all the three time periods assuming that there are no changes in annual precipitation (Romshoo et al. 2015). However, owing to the land system changes, C (Fig. 6) and P factors (Fig. 7) are dynamic as they depend on crop and support practices which are primarily determined by land use and land cover patterns in an area.

Fig. 5
figure 5

RUSLE model input parameters. a LS factor. b R factor. c K factor

Full size image
Fig. 6
figure 6

Crop management (C factor) for a 1965, b 1980, and c 2016

Full size image
Fig. 7
figure 7

Support practice (P factor) for a 1965, b 1980, and c 2016

Full size image

Six risk levels of soil detachment—very low (up to 1 kg ha−1), low (1–2 kg ha−1), medium (2–3 kg ha−1), medium-high (3–400 kg ha−1), high (400–800 kg ha−1), and very high (800–1142.49 kg ha−1)—were identified (Fig. 8). The majority of the study area falls under very low and low risk levels. Very few areas fall in medium-high and high risk levels. Our analysis indicated that the soil erosion in the Narkara catchment decreased substantially from 106.33 × 103 kg in 1965 to 119.84 × 103 kg in 1980 to 61.90 × 103 kg in 2016 indicating a decrease of 41.78% during the assessment period at an annual average rate of − 0.87 × 103 kg. During the period from 1965 to 1980, soil erosion rates increased at an average annual rate of 0.9 × 103 kg per year. Such increase in soil erosion could be attributed to the increase in bare land area from 1965 to 1980. On contrary, soil erosion decreased by 44.83 × 103 kg with an annual average rate of 0.87 × 103 kg from 1980 to 2016. This could be attributed to the fact that area under barren land and agriculture was taken over by built-up areas owing to reckless urbanization from 1980 to 2016. Similar increase in the built-up has been reported by Rather et al. (2016) and Rashid et al. (2017b) in Kashmir Himalaya. Land system changes in the catchment over the 52-year observation period indicate massive increase (2622.96%) in the impervious surfaces (built-up areas) from 0.54 km2 in 1965 to 1.99 km2 in 1980 and 14.92 km2 in 2016, in the Narkara catchment that prevent the rainfall from destabilizing upper soil leading to reduction in sediment yield over the course of time. However, massive increase in the built-up areas has reduced the buffering capacity of Narkara wetland to hold flood waters during extreme precipitation events. This was experienced during the mega flood event of September 2014 when almost 20.4% of the Narkara catchment got completely inundated including 10.2% of actual Narkara wetland area (Fig. 9).

Fig. 8
figure 8

Soil loss (kg ha−1) over Narkara catchment for (a) 1965, (b) 1980, and (c) 2016

Full size image
Fig. 9
figure 9

Inundation of Narkara wetland catchment, as indicated by Yellow outline, during September 2014 flood. Cyan outline represents Narkara catchment, and green outline represents Narkara wetland

Full size image

Of the total soil erosion from the catchment, the highest contribution was from bare lands and agriculture land cover types which could be attributed to the loose soil particles having a capability to easily get detached as compared to other land cover types. High erosion rates from agriculture and bare lands together with sewage influx into Narkara wetland have a potential to cause sedimentation of the wetland bed and accelerate the process of eutrophication (Jain et al. 1999; Khan 2004; Vass et al. 2015); however, it needs to be further researched.

Conclusion

Keeping in view the reckless urbanization, land system changes in Narkara, a semi-urban wetland, were assessed using high-resolution satellite data between 1965 and 2016. The study assumes importance keeping in view the pace of urbanization that is deteriorating the health of this semi-urban wetland. Detailed ground truth was done to validate the land use land cover of Narkara catchment delineated from 2016 data. The land use land cover delineated from 1965 corona data was validated using Survey of India topographic map surveyed in 1962. Owing to the absence of any ancillary information, the land use land cover data of 1980 could not be validated; however, since corona data was also used for delineating land use land cover from 1965, it was presumed to have similar accuracy. Additionally, soil erosion was quantified for 1965, 1980, and 2016 to quantify the impact of land system changes on soil loss using RUSLE in a GIS environment. Massive land transformations were observed in the Narkara catchment. The area under built-up and plantation increased by 2663% and 55%, respectively, while agriculture area decreased by 78% between 1965 and 1980. However, the magnitude of these changes is more pronounced in 1980–2016 period owing to large-scale migration of people from rural Kashmir to urban Srinagar for better civic amenities. The soil erosion estimates from RUSLE suggest decrease in the soil detachment from 120.21 t ha−1 for 1980 to 62.16 t ha–1 for 2016 that could be attributed to the impervious urban surfaces taking over agricultural land especially during 1980–2016 period. The reckless urbanization in the Narkara affects the hydrological connectivity and ecology of this important semi-urban wetland which increases the vulnerability of people to flooding in the catchment of this Himalayan wetland. It is imperative that the development of infrastructure in the catchment could be allowed after carrying out a proper land suitability analysis of the wetland catchment. Additionally, robust scientific studies need to be taken up to assess the water quality, biota, and sediment characteristics for unraveling recent pattern of eutrophication of this semi-urban wetland. Owing to the fact that a large portion of Narkara wetland catchment falls under barren denuded landscapes, it becomes imperative for wetland managers to increase the green cover in the form of plantation to reduce the influx of sediment into the wetland. Since the wetland is very close to the Srinagar city center, it has a potential to be developed as an ecotourism site that may not only help build the economy but also inculcate a sense of environmentalism among the local population for safeguarding the pristine and ever shrinking wetlands of Kashmir Himalaya.