Development of an integrated peri-urban wetland degradation assessment approach for the Chatra Wetland in eastern India

The loss of peri-urban wetlands is a major side effect of urbanization in India in recent days. Timely and proper assessment of wetland area change is essential for the conservation of wetlands. This study follows the integrated way of the peri-urban wetland degradation assessment in the case of medium and small-size urban agglomerations with a special focus on Chatra Wetland. Analysis of land-use and land cover (LULC) maps of the past 28 years shows a decrease of 60% area of the wetland including marshy land. This has reduced the ecosystem services value by about 71.90% over the period 1991–2018. From this end, The Land Change Modeler of IDRISI TerrSet using the combination of MLPNN and Markov Chain has been used to predict the LULC map of this region. The scenario-based modeling following the LULC conversion and nine explanatory variables suggests the complete loss of this wetland by 2045. However, the authors have also tried to present a future LULC pattern of this region based on an environmental perspective. This proposed map suggests possible areas for built-up expansion on the western side of the city without significantly affecting the environment.

1. Assessing the quantitative degradation of wetland areas through LULC change analysis, landscape metrics, and the construction of an urban water index. 2. The prediction of future LULC change in response to urbanization and consequent loss of ecosystem services values for the Chatra Wetland located in the fringe areas of English Bazar Urban Agglomeration far away from Kolkata metropolitan area.
The novel contribution of this study is that it provides a comprehensive detailed structure to assess the degradation of urban/peri-urban wetlands by integrating different quantitative approaches through the application of geospatial techniques. This study is based on the hypothesis that assessment of quantitative degradation requires an integrated method such as LULC analysis with future projection, estimation of ESVs, and proposing alternative scenarios as part of policy management. Following the hypothesis, this study will address some important questions regarding the sustainable management of Chatra Wetlands. The research question includes-(1) How has the LULC of Chatra Wetland changed over the past 28 years? (2) Is the wetland still in a continuous stretch or has it become fragmented? (3) What will be the future scenario of this wetland? (4) How much ecosystem services value has been lost due to degradation? (5) Is there any alternative way to save this wetland? The prime objective is to assess the quantitative degradation of Chatra Wetland with the help of geospatial techniques by highlighting the spatial-temporal evolution of Chatra Wetland. Lastly, this study has also proposed an alternative LULC plan for urbanization without deteriorating this wetland.

Materials and methodology
The entire methodological section for this purpose is divided into sub-sections to achieve the various objectives. Starting with a brief description of the test site and data collected, the rest of the methodological framework for this section is broken up into five strategies, outlined in Fig. 1, and further elaborated on in Table 1. These strategies represent both the present situation and also a direction to the future.
The site. English Bazar City (Fig. 2) is an important city of West Bengal, which is situated around 327 km away from Kolkata (the capital of West Bengal) and in the middle of West Bengal, India. It has a long colonial history 38 and currently, it serves as Malda District's headquarters. Shaw and Das 39 have noticed that the city has expanded around 30% in the last 28 years and around 3,13,681 persons reside in this urban agglomeration 40 .
The Chatra Wetland (Fig. 2), situated in the English Bazar city, is an important region known as the 'kidney of the English Bazar' 41 . It is mainly a rain-fed wetland, but due to its vast extent, water remains in it throughout the year. The site is located on the plains region of Gangetic plain with elevation ranges from 37 to 7 m. This region www.nature.com/scientificreports/ is made with the newer alluvium of Ganga and Mahananda, which is known as 'Diara' . It supports good drainage and fertile soil which is also a reason for the high population density in this region 42 . Chatra Wetland provides a variety of ecosystem services to the city dwellers of English Bazar urban agglomeration such as recharging of groundwater, cooling effect, flood control, biodiversity, food, recreation, etc. On-field observation has revealed that the people of adjacent areas of this wetland are directly dependent on it for fishing, drinking water, agriculture, etc. This wetland is very rich with diversified living organisms. Delineation of the actual boundary of this wetland is a very difficult task due to its rain-fed character, so this study has followed past articles 41 to demarcate the boundary of this wetland. It stretches over an area of 8.47 km 2 . Unlike other wetlands, this wetland consists of the marshy tract and the areas with visual surface water. Marshy tracts were demarcated with the help of on-field observation and Google Earth Imageries. LULC is prepared LULC analysis. The following two subsections discuss various indicators and metrics that were used to evaluate the historical LULC change of Chatra Wetland.

Urban Water Index (UWI), Urban Land Index (UI), and Urban Expansion Index (UI x ). Normalized Difference
Water Index (NDWI) is a popular tool to acquire the area of the water body from satellite images. However, for this study, one particular index is required that can relate the changes in the water area to changes in land use. In this regard, the Urban Water Index (UWI) has been devised following Haas et al. 58 (Eq. 1). This index helps to quantify the development of Chatra Wetland as compared to simultaneous urban development. Two more indexes (Eqs. 2, 3) have been used to represent the urban development within the area of Chatra Wetland. These are the Urban Land Index (UI) and Urban Expansion Index (UI x ) 58 . UI has been defined as the ratio between existing urban land and total land. The UI x depicts the changes in urban land as devised by Hu et al. 59 . In this study, the negative effect of urbanization on the Chatra Wetland was assessed by using these three indices i.e. UI, UI x, and UWI.     www.nature.com/scientificreports/ Accuracy estimation. In this study, two MLPNN classification models were tested-one using the 1991-2000 LULC combinations and the other using the 2000-2009 LULC combinations. The best performing model would then be used in predicting the future scenarios. More information relating to the development of the models themselves is found in "Land Change Modeler" section below. The models were validated by using the root-mean-square-error (RMSE), confusion matrix, and receiver operating characteristics (ROC) curve. The RMSE is well known for determining error of predictions, and the confusion matrix was discussed previously in the image classification section. The ROC is an estimator of goodness of fit of the model, with a value of 0.5 indicating complete randomness, and a value of 1 indicating an exact model without error 65,66 . As per the recommendation of Olofsson 67 , the reference map of validation should be more accurate than the classification map. Therefore, the 1991-2000 LULC model was used to predict a 2009 map, and the 2000-2009 LULC model was used to predict a 2018 map. The validation was performed by comparing the predicted 2009 and 2018 LULC maps with the respective satellite imagery of the area of interest (Acquired from the Google Earth Pro version).
In this regard, 300 points (pixels) were selected randomly to carry out the validation by RMSE, confusion matrix, and ROC curve. A good video tutorial about the implementation of the confusion matrix in R-Studio statistical software is available on YouTube 68 and the ROC curve is prepared in SPSS statistical software 25.0 69 . RMSE was estimated following Eq. (4) 70 .
where Î i is the predicted value for sample i; I i is the observed value of sample i; N represents the total number of observations (samples).

Future LULC projection.
The following subsections relate to determining future scenarios of Chatra Wetland and include explanatory variables for the transition potential; transition potential map calculation and simulation; and accuracy assessment for determination of the ideal combination of variables. These future scenarios were predicted using the best model as determined from the previous subsection (i.e. either the 1991-2000 LULC combination or the 2000-2009 LULC combination).
Drivers. Prediction of future Land Use and Land Cover (LULC) enables the researchers to foresee the future LULC pattern and helps to estimate the demand for land use in the future 71 . Furthermore, it provides a guide to take up effective policies for the effective management of natural resources. The selection of a proper set of input variables is essential to produce optimum future projection. Previous studies have pointed out a few influential groups of drivers behind LULC change i.e. physical, Proximity, socio-economic factors, neighborhood factors, etc. [32][33][34][35][36][72][73][74] . However, there are no definite universal explanatory variables in this regard.
The dynamics of urban growth and peri-urban growth in this region was analyzed by Dutta and Das 75 and Shaw and Das 39 . From these findings, it can be concluded that driving factors like accessibility to public services, economic opportunity, political influence, and population growth, are the main driving forces behind the existing LULC transformation in this region. However, it is essential to mention that some of the driving variables (like political situation) are of non-spatial characteristics and it is difficult to incorporate in the prediction model. Besides this, formal consultation with the planning expert of English Bazar Municipality and careful analysis of four LULC transformation stages (1991-2000, 2000-2009, and 2009-2018) have identified a set of driving variables that have provided stimulus behind the LULC transformation in the study area.
Finally, nine driving variables having spatial characteristics were taken into consideration for this study. These factors are: the distance from National Highways, distance from the railway station (Malda Town railway station), distance from the city centre, population density, and proportion of main worker (workers engaged in work for more than 180 days in a year) in different mouzas (villages) as per census 2011, elevation, slope, neighborhood built-up, and DAF cells in LULC. The details of these variables are presented in Table 7. The raster layers of these variables have been generated in ArcGIS 10.2.1.
It is important to mention that all the parameters do not influence the rate of urbanization in the same way. For example, a high population density of a region will create more pressure for the conversion of natural resources. Contrary to this, the rate of urbanization will be high in those areas which lie near the city centre. Therefore, a standardization procedure was followed to remove the contrasting character of the dataset and to make all the layers unidirectional with an equal scale. Some of the popular standardization techniques are 10 point scale 76 , Analytic Hierarchy Process (AHP) 77 , fuzzy standardization 17 , etc.
The main advantage of fuzzy standardisation membership functions over other methods is that it can work in an uncertain system where theory seems inadequate but expert judgment alone can make sense 78,79 . Fuzzy membership transforms all the layers in a continuous crisp membership form which ranges from 0 to 1. The value of '1' denotes membership of the set and '0' denotes non-membership. Following McBratney and Odeh 80 , a sample of the fuzzy set can be represented as Eq. (5).
where µ A = membership of x in the fuzzy set A; µ A = 0 if x does not belong to the set A; µ A = 1 if x totally belongs to the set A; µ A = 0 < µ A (x) < 1 if x belongs to a certain degree in set A.
In this study, fuzzy sigmoid membership functions like 'Large' or 'Small' or 'MS Large' have been applied to driving factors ( Figure S1. See supplementary section). This step was performed to make all the driving factors unidirectional and to give all the factors an equal comparable scale 17 . A description regarding these functions is www.nature.com/scientificreports/ provided in Table 6 and the allocated fuzzy membership and logic are shown in Table 7. After fuzzy standardization, these variables were imported into the IDRISI environment 1 .
Land Change Modeler. Prediction of future LULC includes the creation of a transition potential map to represent the scenarios of future LULC. However, it is important to mention that English Bazar is not a planned city and there is no proposed LULC plan for this city in the upcoming 25 years. So, it is quite expected that the expansion of this city will occur in the same way in the near future as it is today. Hence, scenarios based on "as usual growth" was taken into consideration for the projection of the condition of Chatra Wetland in the future following the work of Thapa and Murayam 71 . The entire operation of future LULC prediction was performed by the IDRISI TerrSet software 1 with the help of its Land Change Modeler (LCM) extension. The simulation procedure in Land Change Modeler has been designed in a combination of Multi-Layer Perception Neural Network (MLPNN) and Markov framework. MLPNN 86 classifiers can be described by Eqs. (6) and (7) 70 .
where net j represents the input that a single node j receives; W ij refers to the weights between node i and j; I i represents the output of node i of an input or hidden layer. Furthermore, output (O j ) from the j node was calculated as follows: where function 'f ' represents a non-linear sigmoidal function. The main advantage of MLPNN is that no prior assumptions are required while modeling [87][88][89] . MLPNN performs the integration of all the explanatory variables, whereas, Markov model determines the transition potential where f 1 = spread of membership value from 0 to 1; f 2 = mid-point having membership value as 0.5

Small
Sigmoid shape with small values having larger memberships 81 where f 1 = spread of membership value from 0 to 1; f 2 = mid-point having membership value as 0.5

MS large
Sigmoid shape wherein mid-point and spread of membership values are defined by the mean and standard deviation of the data. Here, larger values represent larger membership 82 where m represents mean; s represents standard deviation; a and b are the multiplier Table 7. Description of selected variables and Fuzzy standardization process.

Category Variable Description
Fuzzy standardization

Membership function
The reason behind the allocation of membership  57 to generate the overall sample size. Equation (9) was used to determine the number of hidden layers.
where N represents the number of units (pixels) in the area of interest; W i represents the mapped proportion of class I; S i represents the standard deviation of stratum I and S( ⌢ O ) represents the standard error of the estimated overall accuracy which is likely to achieve.
where N h represents the number of hidden layer nodes; h represents the hidden layer; N i represents the number of input layer nodes; i represents the input layer; N o represents the number of output layer nodes; o represents the output layer.
Finally, all the nine driving variables were incorporated in the data sub-model structure and the transition potential maps were generated by MLPNN with a 1025 sample size. Equation (9) suggests that the number of hidden layers will be six. However, after several cross-validations, 5 hidden layers were selected to run the model due to better output (results). In LCM, MLPNN automatically performs the training process by generating training cells from the transition of land for each LULC type. It creates a multivariate function that helps to predict the transition probability of any cell based on the values of explanatory variables. In LCM, more than 80% of classification accuracy is generally recognized as good to carry out the prediction analysis 71,90,91 . In this study, the MLPNN model has displayed the classification accuracy rate of 84.20% and 86.42% after 10,000 iterations for the first and second combinations respectively.
ESVs estimation. The term ecosystem services refer to all the material and non-material benefits that the living organisms enjoy from a particular environmental unit 92 . Compact natural cover with no significant loss in natural cover will deliver steady ESVs over time. Previous studies showed that the conversion of the natural cover into anthropogenic land uses led to negative changes in the overall ecosystem service values (ESVs) 28,93 .
The pattern of changes (positive/negative) in the ESVs helps to understand the loss of natural cover and helps to undertake necessary strategies for its management. Therefore, temporal estimation of ecosystem service value serves an important role to assess the rate of degradation of the environment 73,94,95 . To show the impact of urban growth on the ecosystem, calculation of ecosystem service value of each LULC (Waterbody, Vegetation, Urban, Agricultural land, sand bar, etc.) was necessary 29 .
Costanza et al. 92 issued a list of coefficient values for different LULC to estimate the ecosystem service values based on a simple benefit transfer method. However, subsequent assessments based on a larger number of case studies have updated the ESVs estimation 37 . Modification in the value of coefficients has been done based on the report of de Groot et al. 96 . This report consists of aggregated values of 22 ecosystem services, which were prepared based on 665 value estimations from over 300 case studies in different parts of the world. This study used the modified global coefficient values of ESVs estimation following Costanza 37 . A previous study by Das and Das 29 , showed the successful application of these modified coefficient values to estimate the ESVs in the tropical areas of West Bengal.
Following this, the ecosystem service values were calculated using Eqs. (11) and (12).
where ESV = Ecosystem service value of a particular year; A k = Area in hectare; VC k = Value coefficient ($/ha/ year) for land class 'k'; ESV f = Ecosystem service value of function 'f '; VC fk = Value coefficient for function 'f ' ($/ ha/year) for land class 'k' .

Results and analysis
The following results are covered in two sections, those relating to the historical degradation of Chatra Wetland, and those relating to future scenarios and ecosystem services.
The quantitative degradation of Chatra Wetland. This section is further subdivided into consideration of LULC changes, UI, UIx, and UWI changes, and an analysis of landscape metrics.
Analysis of LULC changes from 1991 to 2018. Analysis of LULC (1991-2018) of the study area shows a gradual decrease in the net area of Chatra Wetland and it represents the transformation of Chatra Wetland from a continuous stretch to a discontinuous stretch (Fig. 3). In 1991, around 660.6 hectares was under the marshy land www.nature.com/scientificreports/ and wetland in the study area. However, the classified LULC map of 2000 (Fig. 3) displays a significant increase in the area of Chatra Wetland and a registered 1089.18 hectares under the wetland and swamp area. This was due to devastating flooding in English Bazar city and its surrounding region in 1998 and 1999 97 . The vegetated areas were converted into swamp areas due to waterlogging at this time 98 . This was the only exception year that shows an increase in the net area of marshy land in the study area. After nine years (2000-2009), the area under wetland and swamp has reduced significantly (Table 8)   However, it is mentioned earlier that Chatra Wetland is mainly a rain-fed wetland. Therefore, analysis of rainfall trend from 1991 to 2018 is essential to find out the role of rainfall in the shrinking character of Chatra Wetland. In this regard, one existing study 99 showed that there was no significant deviation in the rainfall trend of the Malda District during this period. Therefore, the period of 2000-2018 has registered a significant deterioration of environmental units in this region due to mainly the LULC conversion, and the Chatra Wetland has suffered the most.
UI, UI x , and UWI changes. The application of these indices was carried out only in case of the area of Chatra Wetland (Fig. 4) to understand the encroachment by urbanization. UI shows that the presence of urban land was only 0.52% (0.027 km 2 ) in 1991 within the demarcated area of Chatra Wetland. However, this value has increased to 4.49% (0.23 km 2 ) in 2000 despite the overall increase of marshy land in the study area. UI x indicates a 765.56% increase in urban land within the boundary of Chatra land at this time.
However, the share of DAF land has decreased by 1.94% (around 0.1 km 2 ). UWI also indicates an increase (4.108%) ( Table 9) in wetland cover. Few portions of DAF land emerged out as a marshy land in 2000. However, the share of wetland with visual surface water cover has decreased from 27.42 to 26.74% during 1991-2000. A positive increase in UWI was the result of an increase in marshy land due to the waterlogging event. Encroachment in Chatra Wetland by urbanization has further increased during the subsequent periods.
However, the next phase (2000-2009) has registered the lowest rate of urban expansion within the periphery of Chatra Wetland. The urban land has increased by 28.32% and its share has reached 5.76% in 2009. This time DAF land has remained the main force behind the shrinkage of the wetland area with a 15.78% share. Consequently, a significant decline has been observed in UWI (− 1.76% as compared to urban land and DAF) and the share of wetland with visual surface water and marshy land has been reduced to 22.75% and 26.74% respectively (Fig. 4).
The phase of 2009-2018 again registered an increasing urbanization effect on the wetland. The proportion of urban land and DAF land has increased to 7.90% and 25.89% respectively. The rate of urban land increase (UI x ) during this time was about 37.08%. As against this, the share of the wetland has reduced by 1.39% and only 29.70% area with the recognized Chatra Wetland area has remained as a true wetland.
Overall, urbanization has caused 1.02% shrinkage in the area of Chatra Wetland from 1991 to 2018. Besides this, assessment of the relationship between the loss of Chatra Wetland and the expansion of agricultural land has also revealed a significant role (Pearson's correlation value 0.92; R 2 = 0.66 with p < 0.001) of this factor behind the destruction of this wetland in the western margin.
Analysis of landscape metrics. Analysis of the past 28 years' (1991-2018) LULC has revealed significant shrinkage in the area of Chatra Wetland. The LULC of the study area in 1991 shows an almost continuous stretch of Chatra Wetland. However, the visual representation of subsequent LULC shows fragmentation in the landscape of Chatra Wetland. This section deals with the identification of possible landscape fragmentation of Chatra Wetland by using five landscape metrics.
The decreasing percentage of the landscape of Chatra Wetland with an increase in patch density (PD) ( Table 10) indicates a greater number of small patches in a comparatively lesser area. It also confirms the shrinkage of Chatra Wetland over time. The declining trend of LPI and AMPS suggests disaggregation among the patches of Chatra Wetland as the larger patches are disintegrating into smaller patches. This fact is supported by a relative decrease in the connectedness among patches as the degree of cohesion among the patches has decreased from 94.54% in 1991 to 87.71% in 2018. The relatively lower value of CONTAG also indicates small and dispersed patches of wetland. www.nature.com/scientificreports/  www.nature.com/scientificreports/

Future scenarios and ecosystem services of Chatra Wetland. This section is further divided into an
analysis of the future scenarios, and an assessment of the loss of ecosystem services.
Analysis of the future scenarios of Chatra Wetland. After identification of the on-going shrinking and fragmenting character of Chatra Wetland, it is very much essential to project the future scenario of this wetland. Validation procedures like confusion matrix and ROC have displayed more than 80% accuracy (Fig. 5) in predicting future LULC, which suggests good functioning of the model 100 . Besides this, the RMSE was also very low in both cases (RMSE = 12.7% for the first combination and RMSE = 11.8% for the second combination). However   www.nature.com/scientificreports/ The predicted map of 2027 displays that the area of Chatra Wetland will further decrease and some parts of it will be transformed into built-up and DAF land ( Table 11). The area of Chatra Wetland will register a decrease of 6.87% in comparison to 2018 and only 244.62 hectares area will remain as a wetland area. The next phase of the predicted LULC map i.e. 2036 also suggests a similar trend and predicts more areas of Chatra Wetland will be converted into residential and commercial purposes due to the uncontrolled growth of the built-up area. This time the area under Chatra Wetland will remain only 208.26 hectares and the wetland will almost lose its existence by 2045.  www.nature.com/scientificreports/ city mainly in the southern direction. On average, the wetland with visual surface water area and swamp area together in 1991 had provided ecosystem services that were worth 61.18 million United States Dollar (USD) / hectare/year. This value increased up to 64.14 million USD/hectare/years in 2000 due to an increase in marshy land. Since then, the environmental services of this region have started to deteriorate due to the conversion of this wetland into other LULC. As a result, the benefit from different environmental functions such as water supply, food production, water regulation, nutrient cycling, etc. has also been disrupted significantly. This has resulted in comparatively low ESVs in 2009. The total value of ecosystem services, obtained from Chatra Wetland in 2009, was about 18.96 million USD/hectare/year. It shows a 70.44% decrease from the year 2000. However, the area of Chatra Wetland has reduced down further during 2009-2018. During this time, the rate of conversion of Chatra Wetland was due to residential and commercial purposes. As a result, the net value of ecosystem services has also decreased further. The amount of ESVs has registered a reduction of about 7.47% and it came down to 17.55 million USD/ hectare/year in 2018.
Ecosystem service values of different environment units for different environmental functions have been given in detail in Appendix 1-10 (see supplementary section). As per the future scenario, the ecosystem services values of this wetland will come down further and it will be only 17.65 million USD/hectare/year in 2027 (Table 12). With further shrinkage of the area of Chatra Wetland, the net value of ESVs from this wetland will remain only 13.36 million USD/hectare/year in 2036 (Table 12) and 6.78 million USD/hectare/year in 2045 (Table 12). It shows a decrease of 88.92% as compared to the ESVs of 1991. The entire area of this wetland will be converted  www.nature.com/scientificreports/ into other LULC as per the prediction of the IDRISI TerrSet software 1 . This will result in the loss of one major environmental unit in this region and ultimately will cause serious environmental deterioration in this region.

Discussion
LULC map of the study area defines six different types of LULC for the years of 2000, 2009, and 2018. The post-monsoon month (November) was taken for the LULC as it will enable the net area loss after the monsoon season 17 . Lean period (dry season) was avoided in this study as some portion of Chatra Wetland gets covered by plantation at this time and it makes it very difficult to assess the actual extent. Analysis of spatial-temporal LULC change displays that expansion of built-up and DAF land within the periphery of Chatra Wetland was the main reason behind its net area loss. It is also a major reason for wetland loss throughout India 101 . The open space of Chatra Wetland in the southern and western sides during the lean period becomes the ground of cultivation. It was observed that the recurrence of cultivation without any restriction has converted the wetland into a permanent plot of agriculture and has led to the loss of wetland in West Bengal 17,102 . A similar trend was also observed in the case of Chatra Wetland (Pearson's correlation = 0.92; R 2 = 0.66 with p < 0.001 between net area loss of wetland and expansion of agriculture). The extension of Chatra Wetland was continuous in 2000 and the entire adjoining part of southwestern English Bazar city was under the swamp and wetland with visual surface water cover. However, with the expansion of English Bazar city, the area under Chatra Wetland has started to squeeze during the successive period 103 .
Conversion of wetland into built-up areas is very much profitable for the housing sector as the land is free and the process is cheap [104][105][106] . Chatra Wetland as the peri-urban wetland of English Bazar city has also paid the price of urbanization, which has affected the eastern part of this wetland. Further analysis by the landscape metrics also indicates the gradual fragmentation of this wetland over the period.
Several past studies have shown that large patches are more suitable for bigger and greater biodiversity 107,108 ) and some studies have shown the importance of connectivity among the habitat patches within the urban landscape 108 . However, the result of landscape metrics has shown a simultaneous decrease both in the patch size and connectivity. It indicates that larger patches are disaggregating into smaller patches with greater dispersion 58 . This will ultimately harm ecosystem services 26,84,109 and biodiversity as it will influence the species dispersal due to a decrease in habitat connectivity and patch size.
The twenty-first century has registered a huge number of rural to urban migration in India in search of a better living standard 110 . The same situation can also be observed in the case of English Bazar city. However, most of the newly migrated rural people were unable to purchase land within the city due to the high cost of the land value and it forced them to settle in the vicinity of the city. In terms of accessibility, the adjoining areas of the western and southwestern parts of this city are much more suitable to settle down in due to the proximity of National Highway 34 and Malda Town railway station. It acts as the driving force to settle down in this region 39,75,103 . This has ultimately resulted in the conversion of wetland into residential and commercial use.
The transition probability scenarios based on 2000 and 2009 LULC illustrates relatively better suitability in prediction. The model suggests that the conversion process of the natural cover of Chatra Wetland is of a centripetal type wherein the peripheral region is first converted into DAF land and later on built-up has emerged in this place. This process gradually proceeds towards the inner core of Chatra Wetland and the model suggests that the existing LULC conversion rate will consume almost the entire Chatra Wetland by 2045.
As a consequence of this, it will bring several negative environmental effects and some of them are already visible 103 . Besides this, Chatra Wetland acts as a central drainage system in this region. If the wetland area has been converted into other LULC, then waterlogging conditions will possibly arise in the western and southwestern part of English Bazar city after heavy rainfall. This situation has already started to take place in this region. An interview of 100 residents by the authors in this region has revealed that the frequency and duration of waterlogging conditions after heavy rainfall has increased around 19% over the last five years as per their opinion. However, still, there is no appropriate action plan in this regard to tackling this emerging problem.
Finally, It is a known fact that wetland provides many environmental functions to urban systems such as flood control, filtration of polluted elements, nutrient recycling, accretion of sediment, recharge of groundwater, erosion rate control, preservation of biotic elements, etc. So, the loss of wetland will result in the loss of ecosystem services 31,111,112 . A decrease in the net area of Chatra Wetland in the study area has also resulted in a decrease of 71.90% ecosystem services value from 1991 to 2018. As per the projected LULC, a further 61.35% decrease in ESVs will occur by 2045 if the current LULC dynamics continue. So, the quantitative analysis of Chatra Wetland degradation indicates significant deterioration of this wetland and raises concern for an alternate LULC plan to save this wetland.

Proposed LULC scenarios following environmental perspective for sustainable development
Peri-urban wetlands are considered an essential component of the urban ecosystem and sustainable use of these units helps to meet the objectives of sustainable development goals 113 . These wetlands help to improve the environmental condition of an urban area by supplying several environmental benefits such as offsetting carbon emission 114 , air pollutants reduction 115 , regulation of microclimate 116 , amenities and recreation 117 . These ultimately help to maintain sustainable urban development [118][119][120][121] .
Despite all of these benefits, the degradation of peri-urban/urban wetlands across the world has remained a matter of serious concern. In India, the process of urbanization took place at a much faster rate after the implementation of economic reforms in 1991. As a result, all the existing urban centers have started to expand their area through the development of urban sprawl. It has generated immense pressure on the urban/peri-urban www.nature.com/scientificreports/ wetlands, which has ultimately resulted in the rapid infilling of these wetlands due to the growing demand for lands 41,122,123 .
Ongoing degradation of Chatra Wetland is also a result of this rapid urbanization. It is very much true that the survival of this important wetland is essential not only for the residents but also for the sustainability of this city. In this regard, the authors have tried to present environmental perspective-based LULC projections in the near future without affecting the environmental units. Besides this, the focus was also given to the accessibility of the newly proposed built-up areas while projecting future LULC. The environment based LULC scenarios was prepared using the Land Change Modeler extension of IDRISI TerrSet 1 . In this case, Normalized Difference Water Index (NDWI) ((G − NIR)/(G + NIR)), Normalized Difference Vegetation Index (NDVI) ((NIR − RED)/ www.nature.com/scientificreports/ (NIR + RED)) map was prepared in ArcGIS 10.2.1 for the study area additionally along with the previous layers. Fuzzy membership function 'small' was applied in the case of NDWI and NDVI, which represents the absence of these units are suitable for built-up expansion. English Bazar city is an unplanned city. Hence, NDWI and NDVI were not taken into consideration in the previous section of future LULC prediction. However, this section deals with the prediction of LULC by protecting the natural environment. So, NDWI and NDVI were taken into consideration. All other criteria remained the same for this modeling. Finally, three LULC maps for the years 2027, 2036, and 2045 respectively (Fig. 7) have been projected. All three LULC maps display suitable areas for built-up expansion in the future without deteriorating the environmental units. Table 11 represents the projected transformation of LULC in the future by this model.
The proposed environment based LULC plan resembles no significant further degradation in the area of Chatra Wetland. Besides this, it also highlights that the area along the state highway-10 towards the North-Western side of the city is much more suitable for future urban expansion both in terms of connectivity and from an environmental perspective. The expansion of urban areas will take place in this region at the expanse of DAF land (mainly fallow land). Table 12 represents the consequently estimated ecosystem service values for these proposed LULCs. It suggests no further significant decrease in the net ecosystem service values of the region. Hence, the proposed LULC model is quite acceptable to implement at the ground level.

Conclusion
The present study deals with the effects of LULC transformation on environmental units, especially on Chatra Wetland. The quantitative assessment displays over 60% degradation of Chatra Wetland in the last 28 years due to anthropogenic influence. As a result of this, total ESVs from Chatra Wetland has decreased from 61.18 million USD/hectare/year in 1991 to 17.55 million USD/hectare/year in 2018.
The prediction model based on the existing LULC transformation rate shows that there is a probability of complete loss of this wetland within the first half of the twenty-first century. Further assessment of ecosystem service values also suggests significant environmental deterioration in this region. In recent years, the Government of India has promoted the development of smart cities which are eco-friendly. But, the situation in the case of English Bazar city shows opposite condition where the peri-urban Chatra Wetland registered significant deterioration in the past 18 years. Besides this, still, there is no proposed LULC plan in the case of this city for the future. It indicates the possibility that the conversion of Chatra Wetland will continue in the future as well.
In this regard, the authors have tried to present possible areas for future built-up expansion in this region by protecting the environmental units. The proposed LULC map by the authors shows that the expansion of built-up in the northwest side of the city along the major highway will be suitable in the near future. However, a proper action plan from the local administrative unit is required to stop further encroachment in the area of Chatra Wetland.

Limitations of this study and future scope
Social and economic drivers such as political influence, economic growths, etc. are very important behind the LULC change of any region. However, this study has not incorporated these into LULC modeling due to the non-availability of such databases at the local level and its non-spatial characteristics. Besides this, the study has taken into consideration the MLPNN-Markov chain model for LULC modeling as it comes automatically with the Land Change Modeler. The suitability of other models such as the CA-Markov chain model etc. has not been explored. Finally, the ESVs of Chatra Wetland has been calculated based on global coefficients. However, the estimation of coefficients for local areas may provide slightly better results. www.nature.com/scientificreports/