Multi-dimensional scaling for space-time transformation to achieve sustainable planning and management of water resource under changing land use pattern

Abstract

The land use transition plays an important role for terrestrial environmental services, which had a mixed impact of positive and negative on the groundwater and terrestrial water resource. The health of ecological systems and groundwater depends on the mapping and management of land use. The Ganga basin is one of the most densely populated and agriculture-intensive river systems in the South Asia and the world. The multi-temporal spatial database includes land use (ESA-CCI), satellite-based gravity anomaly (GRACE/GRACE-FO), and well log (CGWB) adopted in this study for assessment of the impact of land use transition on groundwater depth, groundwater drought, and terrestrial water storage. The methodology includes the computation of land use transition, trend magnitude by Sen’s slope, Innovative Trend Analysis (ITA) for graphical visualization, clustering techniques employ to identify pattern & structure, and finally space-time transformation was assessed based on multi-dimensional scaling using Alternating Least Squares Scaling (ALSCAL). The land use transition over two decades shows an increase in forest (2.23%), wetland (2.2%), settlement (208.4%), bare area (3.18%), water (5.18%), and a decrease in agriculture (-1.16%), grassland (-4.5%), & vegetation (-2.8%). The non-parametric climatological trend of groundwater depth, drought, and terrestrial water loss was maximally observed during the post-monsoon season in the Ganga basin. The seasonal climatological trend statistics shows that, the upper Ganga and northern (left) of the Ganga basin shows an alarming rate of groundwater depletion, with increased in the severity of groundwater drought in near future with the loss in terrestrial water storage. The ITA shows the monotonic decreasing trend depicting loss of groundwater and terrestrial water resources. Bi-dimensional regression, ALSCAL shows that the model is efficient based on the input data having stress value and RSQ (proportion of variance) of 0.09 and 0.97 with excellent linear fit. The impact assessment of land use transition was obtained in low dimensional space showing that the conversion from sparse vegetation, agriculture, grassland, wetland and forest to settlement has the maximum impact on groundwater and TWSA loss, although the persistent settlement area is also responsible. The results are extremely useful for the policymakers, scientists, concern Govt. section, and local communities must work together to manage groundwater sustainably. Water resource management can also help to lessen the effects of climate change on groundwater and terrestrial water loss by focusing on the environmental, economic, social, and institutional dimensions of UN-SDG.

Introduction

Terrestrial water resources are crucial for the planet’s climate, ecology, and biogeochemistry, making them an essential climate variable^1–6. The global water resources are under pressure to meet future demands as a result of population growth and climate change, which may alter the spatial and temporal distribution of freshwater availability around the world^7–11. On average, 10% of the global population lives in countries with high or critical water stress which has a significant impact on water access and availability for personal needs¹². However, with rapid global economic growth and population, ensuring a sustainable supply of water resources is becoming increasingly difficult^13,14. Modeling land use at river basin scale can provide essential information for appropriate decision-making for sustainable development targeting agenda 2030 of UN-SDG. Human intervention directly affects water resources through water use, particularly irrigation, and indirectly through land-use change, such as agricultural expansion and urbanization¹⁵. Several studies show that human-caused effects on the terrestrial hydrological cycle have been far greater than climate change and will outweigh the effects of moderate climate change in parts of Asia and the United States^16,17. Terrestrial ecosystem services (ES) are vulnerable to changes in land use and land cover (LULC)¹⁸. In the last few decades, the Ganga River basin (GRB), has accomplished good economic growth, which was reflected in an increase in the urban areas at the cost of the conversion of various land use types like forest, agricultural, grassland, sparse vegetation, and barren land¹⁹. Spatial simulation of land use land cover (LULC) changes is very important to determine and quantify the interaction between LULC dynamics and potential drivers causing changes of groundwater depth, drought, and terrestrial water resources in the Ganga River basin for decision-making.

Groundwater depletion on a global scale contributes to sea level rise and has increased significantly since the mid-20th century. However, its effects on water resources are more visible at the regional level, as in agriculturally important areas of India, China, and the United States¹¹. Worldwide, agriculture that relies on groundwater for irrigation uses about ~ 70% of the total groundwater extracted and covers approximately ~ 38% of the total irrigated land area^19–21. With recent rapid changes in climate and land use, the global water cycle is experiencing high levels of spatial and temporal variability, resulting in a plethora of water-related issues that jeopardize human water security²². As a result, improving understanding of the hydrological cycle and water resources has become a top priority for environmental and natural resource research¹².

The world’s most vulnerable regions to groundwater depletion are Asia as a continent and India as a nation. The majority of water-stressed nations and basins are found to be in Southeast Asia²³. The study indicates that the 21st century poses a threat to India’s food and freshwater security due to the significant interconnection between groundwater and land use/land cover across the three basins (Indus, Ganga, Brahmaputra) and how land cover change affects groundwater storage²⁴. Groundwater exploitation is primarily driven by the rapid expansion of agricultural and urban projects to meet the demands of an expanding population^9,25–27. The reduction in precipitation and the rise in temperature lead to significant alterations in the hydrological cycle of the Ganga basin^28,29. Various studies based on the well log as well as gravity anomaly (GRACE) show the alarming condition of groundwater depletion in the Ganga basin^{3,11,19,30,31}. Unsustainable groundwater withdrawal is exemplified by the Ganga basin. Approximately 57% of irrigated lands in the Ganga basin use groundwater irrigation, making it one of the most irrigated regions in the world. The Ganga basin contains 0.6% of the world’s total land area and 8% of its population. Furthermore, 10.3% of the world’s irrigated land is located in the basin^19,20,32.

Several attempts have been made to assessing the climate change and anthropogenic-induced impacts on groundwater in different region of the world^19,33–36. But very limited studies had been done to assess the land use impact on groundwater resources at the basin scale, as LULC directly related to the runoff and infiltration characteristics of the groundwater^37–41. However, the balance of groundwater resources is vulnerable to various environmental factors, including changes in land use and land cover patterns^26,42,43. Therefore, assessing their past & present situation of land use pattern is crucial and pressing for successful water resource management and ecological restoration, which is lacking.

The research highlights the significant gap, based on the lack of detailed assessment due to changing land use patterns on the groundwater and terrestrial water storage anomaly in the study area. These changes triggered the terrestrial ecosystem services, which have both positive and negative impacts over the sustainable management of the water resource. The novelty of the research work lies in the compressive assessment of multi-temporal land use impact on seasonal groundwater depth, groundwater drought, and Terrestrial Water Storage Anomaly (TWSA) at the pixel level. The Ganga River basin serves as one of the most populous and agriculturally intensive regions of the world, having diverse hydrogeological settings, climatic conditions, and morphometric characteristics.

The objective of the present study helps to evaluate the driver of land use transition class on depletion/loss and recharge/restoration using well log data and satellite-based gravity mass anomaly (GRACE/GRACE-FO) in the Ganga basin. This study uniquely highlights the interaction among the three Essential Climate Variables (ECV) i.e., land use, groundwater, and terrestrial water, integrated on the Geographic Information System (GIS) platform. The assessment helps us to quantify the magnitude of the trend responsible for the groundwater depth, drought, and terrestrial water storage anomaly for the long-term water resource sustainability under changing land use patterns. Therefore, the work focuses on the comprehensive assessment based on the multi-temporal time series analysis using non-parametric tests, graphical representation using Innovative Trend Analysis (ITA), two-way clustering technique, and Multi-Dimensional Scaling (MDS) using Alternating Least Squares Scaling (ALSCAL) to understand the causal relationship among various land use transition class impacts on groundwater, drought, and GRACE-TWSA. The findings are extremely useful for the planning and management of water resource in changing physical cover of the Earth’s surface. The present study provides scientific information for water resources management and decision-making, highlighting the need for protection and governance of groundwater resources.

Study area

In the South Asian countries, the Ganges River System, also referred to as the Ganga is a major river. The region extends from 31°28’ to 21°32’ North latitude and from 73°24’ to 89° 5’ East longitude, encircling the tropical and sub-tropical climatic zone that is spread across multiple states in India (Fig. 1). The diverse morphometric characteristics shows the mean elevation of 386.12 ± 685.25 m, and area ~ 0.84 million km². For modelling the hydrological characteristics of the basin, information on the soil profile is also necessary. The high resolution gridded Hydrological Soil Group (HSG) data having ~ 250 m spatial resolution were prepared by integrating groundwater, bedrock depth and soil texture, which helps to determine the runoff potential⁴⁴.

Fig. 1.

Spatial location of the Ganga basin showing hydrological soil group.

Soils with low, moderately low, moderately high, and high runoff potential are categorised into four standard classes-A (sand), B (Sandy loam, Loamy sand), C (Clay loam, Silty clay loam, Sandy clay loam, Loam, Silty loam, Silt), and D (Clay, Silty clay, Sandy clay). The depth to groundwater table dataset was used to assign dual HSGs to pedons with shallow water tables (less than 60 cm below the surface). The dual HSG as A/D (sand), B/D (Sandy loam, Loamy sand), C/D (Clay loam, Silty clay loam, Sandy clay loam, Loam, Silty loam, Silt), and D/D (Clay, Silty clay, Sandy clay) in the dataset were indicated by the column pixel values^44–46. The hydrological soil group, shows that the Ganga basin have mixed impact by moderately to high runoff as well as high to low infiltration potential (Fig. 1).

Data and methodology

For the space time transformation assessment of land use impact on the groundwater resource required a various dataset such as groundwater depth data, multi sensor-satellite based gravity anomaly dataset (GRACE/GRACE-FO), and the land use. For the evaluation of land use impact on groundwater resource management in the Ganga basin. The Inverse Distance Weightage (IDW) analysis was performed along with the importation of all the data into GIS software in order to estimate the attribute values for each location. Land use dynamics have significant impact on the groundwater level depth trend. The land use datasets were geoprocessed from the ESA-CCI having spatial resolution of 300 m (C3S,2019). The European Space Agency Climate Change Initiative Land Cover data (ESA CCI-LC) has an accuracy rate of approximately 71.1% worldwide among different observational land cover datasets^47,48.

According to^23,47, the weighted overall accuracy of the ESA CCI land cover map is 73.14% with a certain sample dataset of 3167, and 79.25% with 2115 certain and homogeneous sample points. Both land use and groundwater depth trend datasets were configured for the same time period and spatial resolution from 1996 to 2016.

The assessment of groundwater depth trend, and drought were very well explained, along with the data processing by⁴⁹. The groundwater depth database in three distinct seasons—summer (pre-monsoon), monsoon, and post-monsoon was gathered from the Central Groundwater Board, Ministry of Jal Shakti, Department of Water Resources, River Development, and Ganga Rejuvenation, Government of India. A Geographic Information System (GIS) platform was integrated with the point data that was geo-processed for the pre-monsoon, monsoon, and post-monsoon seasons from the various well locations, total 52,699. The utility of Terrestrial Water Storage (TWS) in hydrological research saw a significant enhancement post-2002, following the advent of dependable data from the Gravity Recovery and Climate Experiment (GRACE) satellite, a joint venture by NASA and the German Aerospace Center (DLR)^50,51. GRACE offers a superior capability compared to conventional remote sensing methods and measurements in ungauged watersheds by capturing a more detailed view of the entire water storage, ranging from surface levels down to the deepest groundwater aquifers⁴⁹.

Additionally, observations from GRACE have shown significant promise in forecasting the decline of groundwater levels and in examining the interplay between climate and human activities^49,52–55. The separation between the paired satellites fluctuates due to the dynamic nature of Earth’s gravitational field. These fluctuations are instrumental in tracking changes in terrestrial water storage, allowing for monthly monitoring of hydrological variations^56,57. This research utilizes the latest GRACE monthly mass data sets (RL 06), which were processed at the Jet Propulsion Laboratory (JPL)⁵⁸. The monthly data on Terrestrial Water Storage Anomalies (TWSA) spanning from 2002 to 2023, with a 1° × 1° spatial resolutions, undergo seasonal processing and analysis. A key benefit of the RL06 data set is the reduction of signal leakage errors along coastlines through the use of the Coastal Resolution Improvement (CRI) filter, enhancing the precision of the GRACE TWSA measurements⁵⁹. The grid points were analysed to check the significance trend and magnitude of TWSA in the Ganga basin using non-parametric test.

The adopted methodology divided into three section such as understanding the graphical trend of dataset using Innovative Trend Analysis (ITA), two way clustering using Unweighted Pair-Group Method using Arithmetic Averages (UPGMA), and the last is the multi-dimensional scaling for the space-time transformation assessment.

Innovative trend analysis

According to the Intergovernmental Panel on Climate Change (IPCC), recent trends are present in hydro meteorological time series as a result to see the impact of climate change. The novel Innovative Trend Analysis (ITA) approach is valid regardless of the sample size, time series serial correlation structure, and non-normal probability distribution functions (PDFs)^60,61. Additionally, the segmentation of the 1:1 scatter line help into the low, medium, and high clusters provides in-depth details on the internal trend structure of the time series under consideration. The trend magnitude weakens the closer the cluster of data is to the 1:1 line (slope). If the 1:1 line plots follow a straight line parallel to the 1:1 line, then the time series exhibits a monotonic trend; otherwise, it is composed of either trend-free portions or other types of trends^60,61. In the square area enclosed by the variation domain of the relevant variable, increasing (decreasing) trends are located in the higher (lower) triangular regions via a series of Monte Carlo simulations that account for independent and dependent processes, the validity of this novel technique is demonstrated and free from all the assumptions which was used in other non-parametric test like Mann-Kendall test⁶¹. The methodology presented in this study is based on subsection time series plots that are generated from a given time series using a cartesian coordinate system.

Unweighted pair-group method using arithmetic averages (UPGMA)

The Unweighted Pair-Group Method using Arithmetic Averages (UPGMA) is a hierarchical clustering method used to construct a dendrogram (tree diagram) based on the similarity or dissimilarity between data points. It is a bottom-up approach where each data point starts as its own cluster, and pairs of clusters are merged step by step based on the average distance between all pairs of data points in the clusters. UPGMA is a valuable tool in water resource management, providing a systematic approach to classify and analyse water quality data, identify pollution sources, and support ecological assessments. Its ability to group similar data points into clusters makes it easier to interpret complex datasets and make informed decisions for sustainable water resource management⁶².

Here’s an overview of how the UPGMA algorithm works;

1. Input;

   Distance Matrix (D); This matrix contains the pairwise distances or similarities between the entities.

2. Initialization;

   Each entity is initially treated as a cluster.

3. Iteration;

   Step 1; Finding the Closest Pair of Clusters; The pair of clusters with the distance between their members is identified. The average distance’s computed using arithmetic mean.

   Step 2; Creating a New Cluster; The two closest clusters are merged into a cluster.

   Step 3; Updating the Distance Matrix; The distances between the cluster and the remaining clusters are recalculated using mean. Specifically for each existing cluster the distance between it and the new cluster is determined by averaging the distances between each member of both clusters.

   This algorithm iteratively applies these steps to build up a dendrogram representing relationships, among entities based on their pairwise similarities or distances.

4. Keep repeating Steps 1–3 until there is one cluster left, which represents all the entities.

5. Result; The result is a tree (called a dendrogram) that visually shows the connections, between entities. The height of the branches, on the tree indicates the distance at which clusters were merged. The formula for updating the distance between clusters and during each iteration is:

   1

   Where: is the updated distance between clusters and ,

   are the number of entities in clusters and , respectively,

   is the distance between entities .

   UPGMA is advantageous because it is relatively straightforward and easy to understand.

Multi dimensional scaling

Multidimensional Scaling (MDS) is a statistical technique used for visualizing the level of similarity or dissimilarity of data. It is widely employed in various fields to transform complex data into a spatial representation that is easier to interpret. In many research areas, understanding the relationships between different items is crucial. These relationships are often represented as distances or dissimilarities. MDS provides a method for visualizing these distances in a lower-dimensional space, making it easier to identify patterns and insights that are not immediately obvious from the raw data^63–65.

The main concept behind MDS is to represent items or objects in a dataset as points in a space with dimensions. This representation aims to ensure that the distances between these points in the space closely approximate the dissimilarities between the corresponding objects, in the original higher dimensional space.

Applications of MDS in water resources

• Groundwater Assessment:

  Visualization of Water Quality Data: MDS can help visualize complex water quality data by reducing the dimensionality of the dataset. Parameters such as pH, dissolved oxygen, nutrient concentrations, and contaminant levels can be represented in a lower-dimensional space, revealing patterns and trends.

• Watershed Management:

  Land Use Impact Analysis: MDS can be used to visualize the impact of different land use types (e.g., urban, agricultural, forested) on water quality within a watershed. This helps in understanding the relationship between land use practices and water quality.

  Temporal changes: By applying MDS to time-series data, researchers can study changes in water quality over time, helping to identify trends, seasonal variations, and the impact of management practices or climatic events.

Input:

1. Data Matrix (D): A matrix, with dimensions that includes the differences, between every pair of objects.

2. Initialization: Start by setting up a configuration matrix (B), with either predefined value. This matrix essentially represents the coordinates of the points in a space.

   Y (configuration matrix) This matrix displays the distances or dissimilarities, between items in the space. Each element represents the distance between item and item in the representation, with reduced dimensions.

3. Iteration:

   • Step 1: Calculate Residuals (E): Calculate the residuals, which show the differences, between the observed dissimilarities (D) and the distances between points, in the space as determined by the configuration matrix B.

     
   • Step 2: Update Configuration Matrix (B): To minimize the disparity, between the dissimilarities we observe and the distances, in a given space we can modify the configuration matrix using an alternating squares technique.
     min_B∣∣E∣∣².
     Usually, the process of minimizing involves making updates to either the rows or columns of the configuration matrix. This is accomplished by alternating between fixing a set and optimizing the one.
   • Step 3: Calculate Stress: Determine the stress measure, which assesses how well the distances, in the space reflect the observed differences.

Convergence

Continue to repeat Steps 1 to 3 until you feel an increase, in stress or have completed the desired number of iterations.

Output

Understanding ALSCAL relies heavily on the matrix B, which serves as a representation of object positions, within a reduced space.

The primary goal of ALSCAL is to discover a configuration matrix (B) that reduces stress indicating a connection, between dissimilarities (similarity), and distances in the space. The asymmetric travel time case requires the Alternating Least Squares SCALing (ALSCAL) technique is steepest descent method. MDS models vary depending on the level of data measurement; the major types being ordinal, interval and ratio. Since travel times are ratio data, the ratio model appears at first glance to be the most suitable for constructing time spaces. Comparing two or more two-dimensional surfaces, like geographical maps or spatial images, is possible using the bi-dimensional regression technique.

The method was not well recognised until the publication of the methodology in the 1990s, despite having been created in 1978 ^65,66. When two values for each variable indicate a position in two dimensions, bi-dimensional regression is an extension of linear regression. By using a measure known as bi-dimensional correlation, bi-dimensional regression quantitatively assesses how similar two-dimensional surfaces are to one another^66–68.

2

where (x, y) is the location of a point in the dependent plane, and are predicted values of xi and yi, andand are averages of xi and yi (the centroid of the original map).When this index is the only tool used to compare maps, caution should be exercised because it reduces a lot of information into a single number. Perhaps something pictorial would reveal more.

Results and discussion

Grasping the patterns of land use in relation to climate, along with a numerical assessment of groundwater depth, Groundwater Drought Index (GWDI), and anomalies in terrestrial water storage (TWSAs), is crucial for guiding sustainable water resource management strategies in one of the world’s most densely populated basins.

Seasonal trend analysis

Terrestrial water storage (TWS) encompasses all forms of water found above and beneath the surface of the Earth, including moisture in the soil, snow, ice, water stored in vegetation canopies, groundwater, among others⁶⁹.The non-parametric trend statistics of TWSA were carried out using RL06 mascon data from the GRACE and GRACE-FO satellite missions over Ganga basin to explore seasonal changes in terrestrial water storage anomaly, encompassing an area of ~ 0.84 million km² with 79 grid points, from 2002 to 2023.

The TWS is a crucial element in both terrestrial and global water cycles, significantly impacting the flows of water, energy, and biogeochemical elements, thereby playing a pivotal role in the climatic system of the Earth⁷⁰. The work accomplishes the three major season such summer, monsoon, and post-monsoon (Kharif). The long period average statistics for the summer season varies from − 0.23 m to − 0.06 m having mean of −  0.14 m (1σ = 0.03), shows the stress condition of the Ganga basin as per the negative TWSA. The non-parametric test based on the grid analysis shows that only 3 grid points having increasing trend (2 = SIT,1 = NSIT). Out of 79 grid points 76 grid shows the decreasing trend of TWSA in the Ganga basin during summer season (70 = SDT,6 = NSDT). The magnitude of the trend using Sen’s slope for the TWSA varies from − 0.013 (m/year) to 0.004 (m/year) having mean of -0.014 (m/year) (1σ = 0.007) as shown in Fig. 2a.

Fig. 2.

Geovisualization of seasonal trend magnitude of Sen’s slope in mm/year a summer TWSA b monsoon TWSA c post-monsoon TWSA.

The monsoonal changes in the gravity anomaly of the TWSA over the long period time average ranges from − 0.05 m to 0.19 m having mean of 0.08 m (1σ = 0.06). The non-parametric test based on the grid analysis shows that only 3 grid points having non-significant increasing trend and two grid points shows no trend. Out of 79 grid points 74 grid shows the decreasing trend of TWSA in the Ganga basin during summer season (68 = SDT,6 = NSDT). The magnitude of the trend using Sen’s slope for the TWSA varies from − 0.003 (m/year) to 0.004 (m/year) having mean of -0.014 (m/year) (1σ = 0.008) as shown in Fig. 2b during monsoon season observed.

The post-monsoon seasonal TWSA long period average varies from − 0.01 m to 0.09 m having mean of 0.01 m (1σ = 0.05), as deviation shows maximum values suggest that there was huge spatial variability of TWSA. The non-parametric test based on the grid analysis shows that only 8 grid points having increasing trend (3 = SIT,5 = NSIT). Out of 79 grid points 71 grid shows the decreasing trend of TWSA in the Ganga basin during summer season (66 = SDT,5 = NSDT).

The magnitude of the trend using Sen’s slope for the TWSA varies from − 0.004 (m/year) to 0.007 (m/year) having mean of -0.015 (m/year) (1σ = 0.009) as shown in Fig. 2c during post-monsoon season observed. Similar findings from a recent study that estimated the water budget in the Ganga River system under changing climatic conditions were also based on the Noah-Land Surface Model. All 19 of the Ganga basin’s sub-basins had declining trends in their water budgets, according to the results of a non-parametric trend test that was performed on the data³¹.

The groundwater loss based on Sen’s slope in the Ganga basin for the pre-monsoon, monsoon, and post-monsoon are found 0.09 ± 0.22 m/year, 0.09 ± 0.21 m/year, and 0.10 ± 0.22 m/year respectively⁴⁹. The multi-temporal well log data for the groundwater depth trend of three different seasons, that was pre-monsoon (summer; April-May), monsoon (rainy; August), post-monsoon (kharif; November) from 1996 to 2016 were assess. All the point based well log data was imported into GIS software in order to estimate the attribute values for each location using geostatistical approach using the Inverse Distance Weightage (IDW).

In this work, surface raster for the pre-, monsoon, and post-monsoon periods are constructed using multi-temporal groundwater depth data acquired through a geostatistical IDW interpolation approach. These raster are then utilised to produce spatial maps. The multi-point statistics at pixel level using non-parametric test based on Sen’s slope applied to the seasonal groundwater depth for the computation of trend magnitude, which helps to interpret the depletion, and restoration of groundwater resource.

Growing groundwater loss in the Ganga basin has created a concerning situation, according to the future trend based on the non-parametric Sen’s slope analysis. Understanding groundwater drought is crucial to understand the dynamics of climate change and other human-caused activities. Based on the multipoint statistics, the surface raster was used to analyse the seasonal groundwater drought. The pixel based multi temporal studies carried out to determine the nature and magnitude of the Groundwater Drought Index (GWDI) in the Ganga basin during pre-monsoon, monsoon, and post-monsoon season.

For the pre-monsoon non-parametric Sen’s slope used to identify future trend of pre-monsoon GWDI, which ranges from − 0.05/year to 0.04/year with a surface raster mean of -0.006/year (1σ = 0.009). In the same way for monsoon season, the descriptive statistics of Sen’s slope have a mean of -0.006/year (1σ = 0.01) and range from − 0.04/year to 0.04/year. The post-monsoon having a mean of -0.007/year (1σ = 0.01), the descriptive statistics of Sen’s slope based on multipoint statistics vary from − 0.05/year to 0.03/year. The pre-, monsoon, and post-monsoon area percentages of the groundwater drought propagation were found to follow a seasonal trend in the Ganga basin of 74.86%, 74.17%, and 78.58%, respectively³⁷.

Land use transition

The changes in land use and land cover (LULC) categories over time are displayed in Table 1 (Fig. 3). The LULC classes at two distinct dates are compared from 1996 to 2016, and the area of each class that either stayed constant or changed to a different class are shown. The matrix’s rows correspond to the starting LULC classes (1996), while its columns correspond to the finishing LULC classes (2016). The diagonal of the transition represents the constant area, while each class’s changing class area is represented by the off-diagonal elements. The total area of each class at the beginning or end of the term is equal to the sum of each row and each column.

Table 1.

Land use transition (in km²) between1996 and 2016.

1996–2016	Agriculture	Forest	Grassland	Wetland	Settlement	Sparse Vegetation	Bare Area	Water	Snow	Total 1996	% Change
Agriculture	677875.6	3139.13	210.39	0	5383.31	7.37	83.81	755.3	0	687,455	-1.16
Forest	1061.12	99210.6	92.64	1.73	70.89	29.03	0	20.2	0	100,486	2.23
Grassland	376.36	147.23	14806.53	0	297.02	0.36	11.47	245.8	0	15,885	-4.5
Wetland	0	0.27	0	18.75	0.09	0	0	1.55	0	20.66	2.2
Settlement	0	0	0	0	2800.08	0	0	0	0	2800.1	208.4
Sparse Vegetation	67.16	192	0.82	0	37.67	11311.69	0	11.19	0	11,621	-2.28
Bare Area	27.48	0	10.46	0	3.18	1.91	2248	1.27	0	2292.7	3.1
Water	66.7	35.76	48.96	0.64	42.4	5.73	20.02	15,519	0	15,739	5.18
Snow	0	0	0	0	0	0	0	0	3918.6	3918.6	0
Total 2016	679474.4	102,725	15169.79	21.11	8634.64	11356.1	2364	16,554	3918.6	840,217

Fig. 3.

Land use transition map between 1996 to 2016; (a) land use of 1996, (b) land use of 2016, (c) land use transition from 1996 to 2016.

The Tables 1 and 2, shows that the agriculture land decreased from 687454.91 Km² in 1996 to 679474.4 Km² in 2016, with net percentage change of -1.16%. Most of the agriculture loss was due to the conversion to forest (3139.13 Km²), settlement (5383.31 Km²), water (755.28 Km²), grassland (210.39 Km²), bare area (83.81 Km²), and sparse vegetation (7.37 Km²). The forest class increased from 100486.19 Km²in 1996 to 102,725 Km² in 2016, with net increase percentage of 2.23%. Most of the forest conversion came from the conversion of agriculture (3139.13 Km²), grassland (147.23 Km²), sparse vegetation (192 Km²), water (35.76 Km²), and wetland (0.27 Km²).

Table 2.

Showing cross class conversion of land use.

SL.No.	LU_2016	LU_1996	From To	LUT_Abbreviation	Area_Km2
1	1	1	Agriculture (Persistent)	A	677875.6
2	1	2	Forest to Agriculture	F to A	1061.119
3	2	1	Agriculture to Forest	A to F	3139.133
4	1	3	Grassland to Agriculture	G to A	376.3647
5	2	2	Forest (Persistent)	F	99210.59
6	3	1	Agriculture to Grassland	A to G	210.3857
7	2	3	Grassland to Forest	G to F	147.2336
8	3	2	Forest to Grassland	F to G	92.63522
9	2	4	Wetland to Forest	WL to F	0.272992
10	3	3	Grassland (Persistent)	G	14806.53
11	4	2	Forest to Wetland	F to WL	1.728948
12	5	1	Agriculture to Settlement	A to S	5383.308
13	1	6	Sparse Veg to Agriculture	SV to A	67.15599
14	5	2	Forest to Settlement	F to S	70.88687
15	6	1	Agriculture to Sparse Veg	A to SV	7.370779
16	1	7	Bare Area to Agriculture	B to A	27.48118
17	2	6	Sparse Veg to Forest	SV to F	192.0042
18	4	4	Wetland (Persistent)	WL	18.74544
19	5	3	Grassland to Settlement	G to S	297.0151
20	6	2	Forest to Sparse Vegetation	F to SV	29.02813
21	7	1	Agriculture to Bare Area	A to B	83.80849
22	1	8	Water to Agriculture	W to A	66.701
23	3	6	Sparse Vegetation to Grassland	SV to G	0.818975
24	5	4	Wetland to Settlement	WL to S	0.090997
25	6	3	Grassland to Sparse Vegetation	G to SV	0.363989
26	8	1	Agriculture to Water	A to W	755.2774
27	2	8	Water to Forest	W to F	35.76193
28	3	7	Bare Area to Grassland	B to G	10.46469
29	5	5	Settlement (Persistent)	S	2800.077
30	7	3	Grassland to Bare Area	G to B	11.46566
31	8	2	Forest to Water	F to W	20.20139
32	3	8	Water to Grassland	W to G	48.95653
33	5	6	Sparse Veg to Settlement	SV to S	37.67287
34	8	3	Grassland to Water	G to W	245.7836
35	4	8	Water to Wetland	W to WL	0.636981
36	5	7	Bare Area to Settlement	B to S	3.184905
37	6	6	Sparse Veg (Persistent)	SV	11311.69
38	8	4	Wetland to Water	WL to W	1.546954
39	5	8	Water to Settlement	W to S	42.40473
40	6	7	Bare Area to Sparse Vegetation	B to SV	1.910943
41	6	8	Water to Sparse Vegetation	W to SV	5.732828
42	7	7	Bare Area (Persistent)	B	2248.361
43	8	6	Sparse Veg to Water	SV to W	11.19266
44	7	8	Water to Bare Area	W to B	20.0194
45	8	7	Bare Area to Water	B to W	1.273962
46	8	8	Water (Persistent)	W	15518.67
47	9	9	Snow (Persistent)	Snw	3918.616

The maximum net negative percentage change was observed in the grassland, as it decreases from 15884.76 Km² in 1996 to 15169.79 Km² in 2016. The maximum conversion of grassland was observed into the agriculture (376.36 Km²), settlement (297.02 Km²), water (245.78 Km²), and forest (147.23 Km²). The land uses areal conversion wetland from 20.66 Km² to 21.11 Km² were observed from 1996 to 2016, with net percentage increase of 2.2%. Forest (1.73 Km²), and water (0.64 Km²), made the total conversion to the wetland.

The urbanization and peri-urban development’s takes place in most of the region of the Ganga basin, with an effective increase of 208.37% from 2800.08 in 1996 to 8634.64 in 2016. Most of the areal conversion to the settlement came from agriculture (5383.31 Km²), grassland (297.02 Km²), forest (70.89 Km²), water (42.4 Km²), sparse vegetation (37.67 Km²), bare area (3.18 Km²), and wetland (0.09 Km²). Sparse vegetation shows the negative net percentage change of -2.28%, from 11620.53 Km² in 1996 to 11356.1 Km² in 2016. Most of the sparse vegetation conversion takes place into forest (192 Km²), agriculture (67.16 Km²), settlement (37.67 Km²), water (11.19 Km²), and grassland (0.82 Km²). The bare area increased from 2292.68 Km²in 1996 to 2363.65 Km² in 2016, with net positive percentage increase of 3.1%. The bare area contribution came from agriculture (83.81 Km²), water (20.02 Km²), and grassland (11.47 Km²). The surface water body plays a crucial role in maintaining eco-hydrological process, the increased in water body area from 15738.89 Km² in 1996 to 16553.95 Km² in 2016, with a net positive increase of 5.18%. Most of the land use conversion into water takes place from agriculture (755.28 Km²), grassland (245.78 Km²), sparse vegetation (11.19 Km²), forest (20.2 Km²), wetland (1.55 Km²), and bare area (1.27 Km²). The permanent snow remains constant and no conversion takes place into it. Figure 4 shows the land use transition impact from different class on the trend magnitude of groundwater depth, drought, and TWSA.

Fig. 4.

Land use transition impact on seasonal groundwater depth, drought, and TWSA using Sen’s slope trend magnitude (cm/year).

Innovative trend analysis (ITA): graphical techniques for time series data

To optimise the use of water resources, it is crucial to analyse different seasonal groundwater depth, GWDI, TWSA, and land use dynamics properties and pattern. Identification of groundwater depth, land use, and TWSA variations is crucial for water resource, agriculture planning and management in the world’s most densely populated Ganga River basin due to its morphometric, and diverse climatic regions shows more vulnerability towards the hydro-meteorological extremes.

Most research from throughout the world uses the well-known techniques of linear regression, Sen’s slope, Spearman’s rho, and Mann-Kendall trend tests to attempt to discover trend changes. Recently, a method of “Sen-Innovative Trend Analysis” (ITA) was established that offers some benefits of visual-graphical depictions and the detection of trends, which is one of the main themes of our objective^71,72.

It is important to note that the forest, wetland, settlement, bare area and water shows the increasing trend with monotonic trend characteristics (Fig. 5b, d,e, g,h). The agriculture, grassland and sparse vegetation shows the monotonic decreasing trend with low and medium cluster (Fig. 5a, c, f). The pre-monsoon, monsoon, and post-monsoon GWL depth also shows the monotonic increasing trend having low, medium and high increasing trend of the groundwater level depth (Fig. 6a, b, c). Ultimately the TWSA, graphical representation of the Ganga basin using ITA depicts, there is monotonic decreasing trend, shows the future GRACE based drought (Fig. 6d, e, f).

Fig. 5.

Graphical pattern analysis of Innovative Trend Analysis (ITA) of temporal land use class a agriculture b forest c grassland d wetland e settlement f sparse vegetation g bare area h water.

Fig. 6.

Graphical representation of GWL Depth and TWSA a summer GWL b monsoon GWL c post-monsoon GWL d summer TWSA e monsoon TWSA f post-monsoon TWSA.

UPGMA analysis: a two-way approach

Unweighted Pair Group Method with Arithmetic Mean (UPGMA) is a simple and fast algorithm that works well with small datasets. It assumes that the rate of evolution is constant across all branches of the tree and that the distance between two clusters is proportional to the time since they diverged⁷³. The UPGMA work on the principle of sequence alignment over one another to obtain the lowest differences between the classes. The dendrogram prepared through the UPGMA clustering techniques of 47 land use transition class on the trend of groundwater depth (GWD), drought (GWDI), and TWSA of summer (s), monsoon (m), and post-monsoon (PM) using the Euclidean distance approach.

Based on the UPGMA analysis on the LUT transition from 1996 to 2016 corresponding to the significant time series analysis of groundwater depth, drought, and TWSA were carried. The LUT from the sparse vegetation to settlement were observed for the summer, monsoon, and post-monsoon, are 0.78 m/year (78.25 cm/year), 0.77 m/year (77.83 cm/year), and 0.80 m/year (80.09 cm/year) respectively. The overall maximum decreasing (Sen’s slope) of the trend of groundwater depth obtained during the LUT conversion from sparse vegetation to water. The LUT from the sparse vegetation to water were observed for the summer, monsoon, and post-monsoon, are − 0.03 m/year (-3.73 cm/year), -0.02 m/year (-2.71 cm/year), and − 0.01 m/year (-1.75 cm/year) respectively.

The land use transition impact on groundwater depth and fluctuations from 1996 to 2016 in the Ganga basin are shown in Fig. 7 using UPGMA algorithm. It is clearly found that the maximum increasing (Sen’s slope) trend of groundwater depth obtained in the sparse vegetation to the settlement, and declining trend were observed from sparse vegetation, agriculture, wetland to water body. For the groundwater depth trend, summer and post-monsoon groundwater depth shows almost similar pattern, as compared to the monsoon groundwater depth study.

Fig. 7.

Two-way dendrogram clustering based on the Euclidean distance shows the impact of LUT on the seasonal trend of groundwater depth (m/year).

For the assessment of LUT on groundwater drought using the UPGMA clustering techniques based on Euclidean distance. The two major dendogram formed based on the two way clustering analysis. The one major group (A) shows the maximum vulnerability caused to groundwater drought corresponds to five major land use class such as sparse vegetation to settlement (-0.021/year), wetland to settlement (-0.019/year), water to sparse vegetation (-0.015/year), water to forest (-0.018/year), and forest to settlement (-0.012/year).

The spatial regions under these land use transition of cluster A are under more groundwater drought compared to other land use transition class. The cluster B, have the sub-groups as C and D, which shows the recharge and moderate rate of influence to groundwater drought respectively. Interestingly the cluster C shows the groundwater recharge based on the surface land use characteristics. The five land use transition class such as sparse vegetation to water (0.002/year), bare area to grassland (0.002/year), wetland to water (0.001/year), grassland to water (0.0009/year), and agriculture to water (0.002/year). The monsoon and post-monsoon GWDI shows similar pattern, as compared to summer GWDI (Fig. 8). The GRACE/GRACE-FO based TWSA were analysed to evaluate the impact on terrestrial water storage under changing land use dynamics. The summer and monsoon season TWSA shows almost similar pattern as compared to post-monsoon TWSA. It is matter of concern that Ganga basin shows the declining trend of TWSA in summer, monsoon, and post-monsoon season, which shows the precursor to the various extreme events in near future. The most dominant land use class impact the declining rate of TWSA consists of cluster A (13 class), followed by cluster B (34 class). The conversion of sparse vegetation to settlement have the serious negative impact on the TWSA during summer (-0.025 m/year), monsoon (-0.025 m/year), and post-monsoon (-0.03 m/year) season in the Ganga basin. The summer, and monsoon TWSA shows similar pattern, as compared to post-monsoon TWSA (Fig. 9).

Fig. 8.

Two way dendrogram clustering based on the Euclidean distance shows the impact of LUT on the seasonal trend of groundwater drought index.

Fig. 9.

Two-way dendrogram clustering based on the Euclidean distance shows the impact of LUT on the seasonal trend of TWS.

LUT (Land use transition); S,M, PM (Summer, Monsoon, Post-Monsoon); TWSA (Terrestrial Water Storage Anomaly).

Interestingly based on the seasonal trend analysis for groundwater depth, GWDI and TWSA, the land use transition from sparse vegetation to settlement plays pivotal role for the groundwater stress in the Ganga basin. This is an alarming situation for the proper land use planning and management requires for the sustainable natural resource management in future.

Multi-dimensional scaling

The work focus on the ALSCAL, in which the data is unrolled into a low-dimensional space, with a focus on perceptual maps that illustrate the proximity of items⁷⁴. Time-space transformation techniques can be realised to the fullest extent possible with the advent of geographic information systems (GIS), as well as the continuous development and expansion of transformation techniques like multidimensional scaling (MDS) and spatial analytical techniques like bi-dimensional regression⁶⁶. The stimulus coordinate using ALSCAL of the land use transition on impact on groundwater resource shows the good results as the stress value and RSQ (proportion of variance) obtained as 0.09 (< 0.2), and 0.97 (> 0.60) of the bi-dimensional regression analysis.

As per the results based on the bi-dimensional regression analysis, the land use transition impact (Fig. 10) on seasonal trend on groundwater depth, drought and TWSA in the Ganga basin. The first, second, and fourth quadrant shows the maximum land use transition into different classes but shows the relatively high to moderate mixed impact on the groundwater resource. The third quadrant shows the cluster in which most of the conversion takes place to settlement (S) such as sparse vegetation (SV to S), agriculture (A to S), grassland (G to S), forest (F to S), and bare land (B to S), which shows these five major land use transition class are highly impacted to the groundwater resource (Fig. 10).The bi-dimensional regression helps us to understand the similarity among the different land use transition on groundwater resources in low dimension over the space-time transformation. The ALSCAL MDS, the scatter plot of linear fit (Fig. 11) shows that the results are good fitted.

Fig. 10.

Perceptual mapping of land use transition impact on groundwater resource.

Fig. 11.

Shows the distance and disparities of ALSCAL MDS.

Conclusion

In light of the changing climate, long-term sustainability requires striking a balance between environmental preservation, and human needs. When addressing the effects of land use, adopt an integrated approach for managing water resources in the Ganga basin. Complex social, and biophysical processes lead to changes in land use transition. Groundwater depth, drought and terrestrial water storage are significantly affected by long-term changes in LULC.

The following conclusions have been drawn from this paper.

The results show, land use transition has mixed impact on groundwater resource for depletion/loss as well as restoration/recharge. Most of the conversion shows the negative impact on groundwater, and terrestrial water resource leads to the depletion. The cluster analysis shows that, both summer and post-monsoon season for groundwater depth and TWSA corresponds to almost similar characteristics based on the land use dynamics.

For the summer season the groundwater restoration was observed from conversion of sparse vegetation to water (3.73 cm/year), agriculture to water (3.43 cm/year), grassland to water (1.29 cm/year), the high depletion was observed in sparse vegetation to settlement (-78.25 cm/year), water to sparse vegetation (-41.27 cm/year), grassland to sparse vegetation (-33.76 cm/year), and forest to settlement (-32.11 cm/year). For the post-monsoon season three land use class shows the maximum loss of groundwater were observed in sparse vegetation to settlement (-80.09 cm/year), water to sparse vegetation (-38.2 cm/year), and forest to settlement (-36.32 cm/year).

The maximum loss of TWSA during summer season corresponds to sparse vegetation to settlement (-2.50 cm/year), persistent barren land (-2.10 cm/year), barren land to settlement (-2.13 cm/year), grassland to sparse vegetation (-2.35 cm/year), grassland to barren (-2.08 cm/year), sparse vegetation to agriculture (-2.35 cm/year), and forest to grassland (-2.17 cm/year). Interestingly the during monsoon season the gravity mass anomaly loss of TWSA were observed in permanent snow (-2.27 cm/year), barren area (-2.12 cm/year), sparse vegetation to settlement (-2.51 cm/year), sparse vegetation to agriculture (-2.50 cm/year), forest to grassland (-2.25 cm/year), and forest to agriculture (-2.09 cm/year).

The post-monsoon season shows maximum loss of terrestrial water resource corresponds to following land use dynamic such as sparse vegetation to settlement (-2.95 cm/year), grass to sparse vegetation (-2.64 cm/year), sparse vegetation to agriculture (-2.81 cm/year), forest to grassland (-2.57 cm/year), and snow loss (-2.58 cm/year). The loss of groundwater, and terrestrial water resource as an essential climate variable may further threaten the water-stressed region, affecting the environment and socio-economy and also resulting in land degradation. Extreme climatic vulnerability is increased in a region by unsustainable land use/cover transformation, loss of soil moisture, and both.

The findings provide benchmarks for adaptation progress to the UN 2030 Agenda for Sustainable Development and its 17 Sustainable Development Goals (SDGs), the UN Sendai Framework Disaster Risk Reduction (SFDRR), the UN Habitat New Urban Agenda, and the United Nations Framework Convention on Climate Change (UNFCCC). The results are very helpful for decision-making regarding climate risk management. Adaptation has advanced from the basin to sub-basin or watershed level, from the planning stage into implementation.

Author contributions

M. S. U. H.: conceptualization; methodology; investigation; validation; writing—original draft. A. K. R.: conceptualization; methodology; investigation; validation; writing—original draft; A. H. M.: conceptualization; methodology; investigation; validation; formal analysis; resources; writing—original draft; writing—review and editing. M. A. K.: conceptualization; methodology; investigation; validation; writing—original draft; writing—review and editing. F. M. A.: investigation; formal analysis. S. A.: conceptualization; methodology. O. J. A.: investigation; validation, A. M.: Analysis and writing manuscript. All authors reviewed the manuscript.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.