Unravelling the Role of Vegetation Dynamics in the Execution of ArcSWAT Hydrological Modeling for Cumulative Streamflow of a Tibetan Watershed

Abstract

Monitoring vegetation change and hydrological variation is crucial as they are useful means of appraising the ecological environment and managing water resources in water-resource-sensitive regions. The leaf area index (LAI) describes water consumption in hydrologic processes and is an important vegetation variable for water budgeting in catchments. As part of the Soil and Water Assessment Tool (SWAT), LAI is a significant parameter, which links vegetation dynamics with the hydrological cycle. In the current study, we have aimed to describe the Lhasa River (LR) cumulative streamflow based on simulation scenarios obtained with the SWAT model. After dispensing a heterogeneous LAI time series developed by MODIS NDVI as a source of data at the HRU level (SWAT-synthetic LAI scenario), the study has produced a better representation of LR cumulative streamflow in terms of the selected evaluation criteria, encompassing the SWAT-baseline (SWAT-B scenario)-simulated and SWAT-built-in LAI-influenced (SWAT-LAI scenario) LR cumulative streamflow. The study has revealed a close relationship between the observed and the SWAT-SLAI-scenario-generated LR streamflow, with a similar MK trend for the study time span. The LAI has been found to share a close relationship with LR streamflow, as both the LAI and LR streamflow are found to be influenced by the rainfall received in the Lhasa River Basin (LRB). The study is instrumental in understanding the association between LR streamflow, vegetation change, and the climatic conditions of the Lhasa River Basin (LRB).

1. Introduction

Various prior studies have identified intensified and enhanced hydrological changes in the Tibetan Plateau (TP) under current climate change conditions, which have also prompted changes in the water balance [1,2,3,4,5]. Additionally, the hydrological influence on the behavior of vegetation is obvious [6]. TP is the most widespread alpine ecosystem in the world, where water is the principal means that confines vegetation activity; furthermore, both climate change and hydrological variations have intense and substantial impacts on vegetation changes in the TP [4,5,7,8,9,10]. Hence, research on hydrological variations and their effects on vegetation and vice versa has been crucial; moreover, this evidence is useful not only for understanding how the hydrological cycle responds to the climate change but also for appraising the ecological environment and managing water resources in water-resource-sensitive regions. For this reason, an improved representation of vegetation dynamics in hydrological processes may aid in developing a better understanding of the hydrological phenomena in regional studies. For hydrological models that use land cover information as a basis for parameter sensitivity and definition, land-cover-specific vegetation parameters (e.g., leaf area index) may be the most suitable in improving their predictive capability. The leaf area index (LAI) describes the ratio of the total one-sided green leaf area per unit of ground [11,12]. LAI describes water consumption in terms hydrologic processes, which control the amount of water intercepted by leaf area. This vegetation variable plays an important role in hydrologic processes and water budgeting in catchments [13,14,15,16,17,18]. Concerning the influence of LAI on eco-hydrology processes, investigating its changes on a catchment scale is vitally important [19] as it enables hydrologists to estimate accurate water budgets under climate change scenarios.

Developing methodologies for the remote estimation of the LAI [20,21] for various purposes is an area of concern. The growth of remote sensing techniques has helped in the retrieval of LAI on a regional scale. The current methods of estimating LAI from remote sensing data are classified into three categories: (1) the statistical relationship between LAI and vegetation indices [22,23,24,25]; (2) inversion of canopy reflectance models [26,27]; and (3) hybrid inversion methods combining the two methods mentioned above [28,29,30,31]. Global LAI products, such as MODIS, CYCLOPES, ECOCLIMAP, and GLOBCARBON, are generated from remote sensing observations using these methods [32]. However, remote sensing data are prejudiced by atmospheric conditions; this established fact may result in errors and temporal incoherence in the inversion results. Current global LAI products often face this problem, and this restricts their further application in many fields [33,34]. Empirical techniques are one of the maximum customary procedures used to evaluate the LAI, and they commonly involve developing the correspondences between the LAI and some vegetation indices (VIs). VIs are broadly used in remote sensing, primarily owing to their feasible source and application [20]. They are particular arrangements of several spectral bands that enable the assessment of plant status from images. Since vegetation displays a resilient absorption in the red spectral range (subject to plant chlorophyll) and a significant reflectance in the near-infrared bandwidth (liable to the intercellular structure of the leaves mesophyll) [35], VIs, merging these spectral reactions may offer a gauge of vegetation “greenness” and, therefore, act as a proxy of the LAI and chlorophyll content [36]. Consequently, a simple ratio (SR) [37], normalized difference vegetation index (NDVI) [38], and soil-adjusted vegetation index (SAVI) [39] are among the most preferred VIs for estimating LAIs [40].

In a development that has enhanced the significance of the LAI parameterization approach by guaranteeing its physical realism in many widely used watershed models (e.g., the Soil and Water Assessment Tool and Variable Infiltration Capacity), remotely sensed LAI data have recently been employed in traditional watershed modeling approaches. In this study, with MODIS NDVI data as the vegetation catalog and the Beer–Lambert law [41] as the basic theory, the remote sensing inversion model for estimating grassland LAI in southern Chinese TP catchments has been employed. The use of remote sensing for the LAI assessment of grasslands has been previously reported [29,34,42,43,44], but few studies on grassland in China have been done [45,46]. A remote sensing appraisal of grassland LAI in southern China has not been undertaken. China has vast grassland assets, with approximately 400 million ha of grassland, of which >60 million ha are on mountains and slopes. It is imperative to investigate the state of the southern grassland ecosystem exclusively when northern grassland has receded in recent times. Thus, when studying the underlying forces of change in southern grasslands, the LAI is of great importance to the management of the grasslands of China [47]. Estimated LAIs from NDVI are incorporated into the SWAT model by replacing the SWAT-simulated LAI for the Lhasa River Basin (LRB)—a vital southern Chinese TP catchment for portraying its influence on the cumulative streamflow simulation process. This study aims to develop a better physical understanding of the hydrological phenomena of LRB by incorporating vegetation features into the streamflow simulation process.

2. Materials and Methods

2.1. Perspective of the Study

The Soil and Water Assessment Tool (SWAT), a fundamental component of the United States Department of Agriculture Conservation Effects Assessment Program, is one of the most widely applied, semi-distributed, catchment-scale, eco-hydrological models [48,49,50]. Still, it has been found to have some deficiencies in modeling plant growth in grasslands with diverse vegetation characteristics in arid and alpine regions:

(a)

Land use and land cover data are imperative input parameters for SWAT to generate its fundamental calculation unit—the hydrologic response unit (HRU)—which are, however, commonly categorized without consideration of the vegetation coverage [51]. The second level classification of land use/land cover type can reveal vegetation status, for example, high compensation, medium coverage, and low grass cover [52]. So far, SWAT plant and land cover database does not incorporate these land cover categories [53].

(b)

In SWAT, LAI is the significant parameter that associates vegetation dynamics with the hydrological cycle [53,54]. It can unveil the plant growth condition, plant density, and vegetation. However, in SWAT, the LAI is established by the average plant density in HRUs [55]. Furthermore, LAI accumulation in SWAT is governed by a heat-based integrated model of idealized leaf growth, which ignores other factors such as precipitation and topography [53,56].

To highlight these shortcomings, a number of studies have emphasized on enhancing LAI assessment. Referring to [57], they included user-defined minimum LAI to the SWAT model to simulate perennial vegetation configuration in the tropics. Referring to [53], they have inserted a direct soil moisture index to improve the vegetation growth module of SWAT for simulating LAI in tropical forests. The study of [55] also included average forest density in the forest growth module of SWAT to analyze eco-hydrological processes of forests with diverse vegetation inclusion in the Meijiang River Basin, China. These alterations added more parameters to the plant development module and eventually moved forward with the execution of SWAT in demonstrating eco-hydrological processes involving distinctive plant densities and cover. In any case, an increment in parameters makes the process more complex. Therefore, in the current study, we claim, “what if the default plant module of the SWAT model is constrained with the physical vegetation dynamics in the form of daily LAI to simulate the cumulative or aggregate streamflow of a TP watershed?”

At the regional scale, remotely sensed LAI has significant exposition over LAI values measured within the field. In the study by [56], proposing a strategy to enhance the illustration of vegetation elements of evergreen forests, the Moderate Resolution Imaging Spectroradiometer (MODIS) LAI product was utilized, which incremented the pertinence of SWAT application in tropical or subtropical zones. The study of [58] integrated MODIS LAI into the SWAT to advance crop yield predictions in homogenous row crops. Assessments shown in the mentioned hydrologic studies have been typically restricted to engage LAI incorporation into the SWAT model for vertical water flux and storage simulations (e.g., evapotranspiration and soil moisture), irrigation and cropping patterns, plant density, and forest cover, with very little emphasis on the watershed’s vegetation characteristics’ response to cumulative downstream waters (i.e., streamflow) [56,59].

2.2. Study Site—Lhasa River Basin

The Lhasa River Basin (LRB), ranging from 29°19′ to 31°15′ N and from 90°60′ to 93°20′ E, is the economical and authoritative center of the autonomous TP. The Lhasa River (LR) is the stretched tributary of the Yarlung Tsangpo River; LRB covers a ~32,321 km² basin area (ArcSWAT-estimated area by the digital elevation model used in the current study), covering 13.5% of the total area of the Yarlung Tsangpo basin [60]. The LRB exhibits distinctive semi-arid monsoonal climate characteristics, where the major amount of received rainfall (~80%) is allotted to the summer season from June to September with the coinciding generation of peak LR discharge during the same time. The study of [61] showed that rainfall in summer is a dominant component in generating summer stream flow in the Lhasa River basin. As a result, the rainfall disparity poses an undeviating effect on the rainfall-dependent runoff generation process in the basin. The hydrological and meteorological archives for the LRB are retained at the Pondo, Tanggya, and Lhasa hydrometric stations and the Damxung, Maizhokunggar, and Lhasa meteorological stations, respectively, presented in Figure 1.

The LR has been experiencing some key hydraulic interferences in the form of reservoir development and impediment during the preceding and current decades. The chief hydraulic amendments in the study area are the setting up of Zhikong and Pangduo hydropower stations (Figure 1) over the LR. Having a typical temperate steppe ecological setting, the semi-arid monsoon climate is exposed to an annual average temperature of 7.7 °C owing to the high elevation and a mean annual precipitation of 440 mm. LRB is well vegetated with alpine meadow, alpine steppe, marsh, shrub, forest, etc. [62]. The source area of Lhasa River (LR) belongs to pastoral area, while the middle and downstream areas are farming–pastoral areas, respectively [63].

2.3. Scenario-Based SWAT Modeling of LRB Streamflow

In order to elucidate the role of physical vegetative attributes of LRB in the hydrological simulation process, the current study has been designed to collate and contrast the performance of SWAT hydrological model in apprehending the cumulative streamflow received at the LRB outlet by developing three SWAT scenarios for the cumulative daily LR streamflow simulation for the years 2008–2016, with 2008 as the model warm-up year, followed by the model calibration for the years 2009–2012 and, ultimately, the model validation for the years 2013–2016. The simulation scenarios are summarized step by step as follows:

i.

The basic SWAT model for the hydrological simulation of LR streamflow collected at the basin’s outlet is represented as the SWAT-baseline (SWAT-B) scenario.

ii.

The simulation scenario is built where the vegetation parameters are included and tuned for the SWAT model LR simulation. The simulation scenario simulated the LAI for LRB using the default plant growth module including the associated parameters for LAI estimation. As the particular scenario established the simulated LAI values for LRB, the scenario has been represented as SWAT-leaf area index (SWAT-LAI) scenario.

iii.

The scenario for the simulation of LR cumulative streamflow, where the LAI time series developed by using remotely sensed data products is dispensed into the SWAT model, replacing the default LAI values estimated by SWAT for LRB. As the LAI time series was prepared from the remotely sensed vegetation data of LRB, this particular scenario is represented as the SWAT-synthetic leaf area index (SWAT-SLAI) scenario.

The framework of the methodology adopted for the three scenarios developed in the current study is presented in Figure 2.

The motivation behind carrying out the scenario-based SWAT modeling of LR streamflow in the current study is the depiction of the influence on the simulation process by the addition of vegetation data into the hydrological simulation process as this information can be a vital input for the improved hydrological representation of LR streamflow obtained at the basin’s outlet. The detail of the scenarios developed for the current study is as follows.

2.3.1. SWAT-B Scenario

The SWAT model has proven to be one of the most applied eco-hydrological models worldwide [64,65,66,67,68,69]. Agreeing to the operational principles of the SWAT model, a watershed is firstly distributed into sub-basins and, individually, each sub-basin is partitioned into hydrologic response units (HRUs) established on the land use, topography, soil, and slope inputs. The hydrologic phase for each HRU depicts the water balance, including precipitation, interception, surface runoff, evapotranspiration, percolation, lateral flow from the soil profile, and return flow from shallow aquifers. SWAT-B scenario has been developed as the baseline for the daily simulation of LR cumulative streamflow for the desired study period. The necessary input data for the basic model simulation includes geospatial and hydro-meteorological records of the study area. In the current study, the three GIS layers (geospatial input data) for LRB as the topography, soil properties, and land use were fed into the model in the form of:

(a)

Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) “www.usgs.gov (accessed on 31 August 2022)”, with a resolution of 30 m (1 arc-second) (represented as Figure 1).

(b)

Land use raster with 30 m (1 arc-second) resolution developed from cloud-free Landsat Operational Land Imager (OLI-8) images “www.earthdata.nasa.gov (accessed on 13 September 2022)” by using the ENVI Feature Extraction Module (ENVI Fx) and engaging the K nearest neighbor (KNN) supervised image classification method (represented as Figure 3a).

(c)

FAO-UNESCO Harmonized World Soil Database version 1.2 (HWSD v1.2) “https://www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmonized-world-soil-database-v12/en/ (accessed on 29 September 2022)” derived soil raster, having a 30 arc-second resolution (represented as Figure 3b), respectively.

Daily time scale meteorological data for climate variables, including precipitation, maximum and minimum air temperature, relative humidity, wind speed, and solar radiation, were obtained for a long-term period from Damxung, Maizhokunggar, and Lhasa meteorological stations (the location and elevation of the three meteorological stations are presented in Figure 1) and fed into the model for the years 2008–2016. The daily river discharge data series used in this study was acquired from Lhasa gauging station located at the basin’s outlet (Figure 1) on daily time steps for the years 2008–2016. The hydro-meteorological datasets used in the current study are the ground observations for LRB. The current study has utilized the ground-observed hydro-meteorological long-term record to ensure the closest and reliable hydrological simulation process and the outcomes to be practically feasible for future water resource management and planning strategies. LRB happens to be a large TP catchment with vital authoritative importance. Hence, the study intended to produce physically relatable hydrological simulation outcomes. Secondly, as the study is taking into account the total volume of LR streamflow collected at the basin’s outlet, therefore, the hydrological observations recorded at the Lhasa hydrological station, being located near the LRB’s outlet, are utilized for assessing the hydrological simulation capability of ArcSWAT for LR. Additionally, being a human-intervened Tibetan watershed, it is important to analyze the total flow volume received at the basin’s outlet. For this purpose, the LR discharge series maintained at the Lhasa flow gauging station has been used as the hydrological input data series for the SWAT modeling.

All the data on hydro-meteorological variables were prepared as per the requirement of the SWAT model. ArcSWAT2012, set up in ArcGIS 10.2 platform, was employed for the purpose of watershed delineation and sub-basin discretization. The prepared ArcSWAT model for LRB produced 21 sub-basins that were further classified into 149 HRUs given with the 10%–15%–10% land use–soil–slope threshold values, respectively. The identification and sensitivity of the parameters controlling the hydrological phenomena of LRB followed by ArcSWAT model calibration and validation was performed in the SWAT-CUP (calibration and uncertainty program) embedded Sequential Uncertainty Fitting version 2 (SUFI-2) module. This algorithm is proficient in representing all uncertainties (parameter, inputs, conceptual model, etc.) in terms of parameter ranges by trying to enfold most of the measured data inside the 95% prediction uncertainty (95PPU) band, which is calculated at the 2.5% and 97.5% intervals of the cumulative dissemination of all simulated outputs. The desired hydrological parameters were ranked in accordance to their sensitivity by engaging a global sensitivity analysis method.

2.3.2. SWAT-LAI Scenario

SWAT includes a basic erosion productivity impact calculator model to evaluate plant growth [70]. The plant growth module of SWAT model is comprised of two parts: biomass accumulation and LAI accumulation. LAI can reveal plant growth status and plant patterns [55,70], where its development model included in SWAT model is described below.

Before the LAI reaches its maximum value, the LAI on day “i” is calculated as follows [71]:

ΔLAI_i = (frLAI_max,i − frLAI_max,i−1) × LAI_max × {1 − e ^{[5 × (LAI}_i−1 ^{− LAI}_max^)]}

(1)

where ΔLAI_i is the new LAI on day i; frLAI_max,i and frLAI_max,i−1 are the maximum LAI calculated based on heat on days i and i − 1, respectively; LAI_max is the maximum LAI for a plant; and LAI_i−1 is the LAI on day i − 1.

The LAI does not change on subsequently attaining its maximum value. However, after leaf senescence exceeds leaf growth, the LAI is calculated as follows [70]:

LAI = LAI_max × 1 − fr_PHU/1 − fr_PHU,sen (fr_PHU > fr_PHU,sen)

(2)

where LAI is the LAI on a day; LAI_max is the maximum LAI for a plant; fr_PHU is the accumulated potential heat unit fraction on a day; and fr_PHU,sen is the fraction of days where leaf senescence exceeds leaf growth in the entire plant growth season. SWAT model was run by utilizing the built-in default LAI estimation model embedded in the plant growth module (Equations (1) and (2)) to produce LAI values for LRB. Keeping all the other input (geospatial and hydro-meteorological) datasets similar for ensuring uniformity between the SWAT-B and SWAT-LAI scenario, the additional sensitive parameters influencing the plant growth pattern and the vegetation features along with the hydrological sensitive parameters (identified in SWAT-B) for LRB were identified, tuned, and ranked in the SWAT-LAI scenario. This was conducted by using SWAT-CUP (calibration and uncertainty program) embedded Sequential Uncertainty Fitting version 2 (SUFI-2) module through global sensitivity analysis during the calibration, followed by the validation of SWAT model for a similar time period as for the SWAT-B scenario.

2.3.3. SWAT-SLAI Scenario

SWAT simulation in SWAT-B scenario identified the sensitive hydrological parameters for LR cumulative streamflow simulation. Proceeding forward with the study, SWAT simulation in the SWAT-LAI scenario managed to adjust the vegetation parameters in addition to the hydrological sensitive parameters. SWAT-SLAI scenario has been constructed to introduce a new vegetation input dataset into the default plant growth module of the SWAT model. For this purpose, the current study had a two-step design for the establishment of the vegetation time series for LRB:

i.

Remote sensing big data acquisition and treatment;

ii.

Development of daily LAI time series.

Remote Sensing Big Data Acquisition and Treatment

The NDVI was utilized as the input constituent from the MODIS data. In this study, the NDVI remote sensing products of MOD13Q1 v006 with a fine spatial resolution of 250 m and temporal resolution of 16 days were downloaded from the NASA USGS Earth Explorer platform for the time period of 1 January 2008 to 31 December 2016. A cloud cover filter of 10% was selected for the images to be used in the study. In the ArcGIS 10.2 environment, the MOD13Q1 downloaded tiles for a certain date were first mosaicked to a uniform raster to ensure uniformity and credibility in the NDVI estimation for LRB. The 30 m SRTM DEM was then used as a mask to extract the study area from the mosaicked raster of 250 m. The extracted study area NDVI raster had the same spatial resolution of 30 m as the DEM (mask for extraction). ArcGIS rendered a mean NDVI value to the extracted NDVI raster for a particular day. The 16-day NDVI obtained from the MOD13Q1 remote sensing products was extrapolated by using linear regression, resulting in a heterogeneous daily NDVI time series for LRB.

Development of Daily LAI Time Series

After the study of [47], the relationship between MODIS NDVI and the grassland LAI was established according to the Beer–Lambert law. In their study, the Beer–Lambert law was modified and transformed into a model specifically for grassland LAI estimation of southern Chinese catchments. The model using a logarithmic equation is represented as [47]:

LAI = ln (I/I₀)/K

(3)

Equation (3) describes the modified Beer–Lambert law for LAI estimation in the study by [48] where I and I₀ is the light transmittance as per the Beer–Lambert law and K is the canopy extinction coefficient. By exploring the relationship between I/I₀, K, and NDVI, the model for grassland LAI was established as below (for the details of model development, see [47]):

LAI = ln (a₁ × NDVI + b₁)/(a₂ × NDVI + b₂)

(4)

where a₁, b₁, a₂, and b₂ are regulation coefficients for the model. The parameter values in the model presented in

Table 1. Parameter values for LAI calculation using modified Beer–Lambert law represented as Equation (3).

Parameter	a1	b1	a2	b2
Value	4.79	1.91	−1.46	1.73

are calculated as [48]:

Using Equation (3), the NDVI values were transformed into daily synthetic LAI values for LRB. We integrated these spatially disseminated and temporally constant LAI data into our basic model arrangement using a new SWAT database. In each event of simulation (days), our new database employed the “direct insertion approach” after [56,59,71,72,73] to substitute the simulated LAI across all HRU’s and all simulation time steps with the corresponding synthetic LAI data.

2.4. Model Performance Evaluation Criteria and Statistical Investigations

During the calibration and validation periods for all three scenarios, the simulated daily streamflow was matched with the observed data from the Lhasa hydrometric station by the use of the Nash–Sutcliffe coefficient (NSE) [74], the coefficient of determination (R²) [75], and the percent bias (PBIAS, %) [76]. Moreover, the observed and simulated discharge records were statistically verified for the association between them by Pearson correlation [77], Spearman’s correlation [77], and Kendall’s rank correlation [78] for the reliability authentication of SWAT modeling under all the scenarios for LR streamflow simulation. Additionally, the synthetic LAI was correlated with the LR discharge using the same correlation tests. The appraisal of a trend in a time series of hydro-meteorological phenomena and the vegetation behavior of LRB has been conducted using the non-parametric Mann–Kendall test (MK) [79,80,81] in the MS Excel software (2013) supported by XLSTAT 2014 macro. This test is extensively used and can handle the missing and distant data. The test has two parameters that are noteworthy for trend detection: a significance level (p) that symbolizes the command of the test and a slope magnitude estimate (MK-S) that denotes the direction and volume of the trend. The trends in time series were accomplished by calculation of the Kendall coefficient “τ” [82,83,84]. In the current study, the MK trend at a significance level of 5% (p < 0.05) was applied to the hydro-meteorological, NDVI, and synthetic LAI time series.

3. Results

3.1. NDVI as a Primary Variable for LAI Estimation

The vitality of satellite imagery as the most important resource for observing vegetation growth at large scales has long been declared [85,86,87]. The NDVI is, considerably, the most preferred and established VI for calculating the LAI [88]; however, NDVI’s relationship with LAI is principally nonlinear [89], presenting great receptiveness to variations in the plant canopy at initial growth stages (low LAI) and saturating when the plant canopy grows thick [90,91]. Therefore, novel approaches have been acclaimed to deal with this restriction by considering, for instance, the green- and red-edge spectral fragments [92,93,94]. As data from the red-edge spectral section are not accessible always, and the green band is frequently accessible at a coarser spatial resolution than other bands [89], as a result, the development of the LAI from the NDVI is still extensively practical as supported by the literature. In the current study, the precursor NDVI values have shown a very close direct relationship with the synthetic LAI product values (Figure 3). The greater the NDVI, the higher the corresponding logarithmic LAI value, as shown in Figure 4. This direct relationship indicates the presence of a thriving vegetation cover during the hot and wet summer months as represented by a higher NDVI value for the reason that the major proportion of rainfall is received during these months (May–September) and vice versa. The peculiarity of deriving LAI data time series from independent NDVI values using the improved Beer–Lambert equation (logarithmic equation) specifically developed for southern Chinese grasslands is adding individuality to the findings of the current study. Referring to [95], it reports that, when merely short-term NDVI spot data are available, the linear equations are usually used, reproducing single instants during the season instead of the whole growing process. On the other hand, when a large amount of NDVI seasonal observations are at hand, the exponential relationship is frequently used, due to being proficient at replicating nonlinearities and the NDVI saturation at canopy close phases [95]. However, the logarithmic scale is yet to be explored for determining the possible NDVI-derived LAI data time series in a study where LAI can be a defining parameter for the purpose of depicting the hydrological phenomena in a certain region. The study by [95] investigated the use of NDVI-LAI conversion equations based on mathematical form and found an application rate of 20% for logarithmic equations for the purpose as compared to an application rate of 26% and 48% for linear and exponential equations, respectively. In order to accomplish a better understanding of the NDVI–LAI affiliation in the current study, the three well-known and certified correlation tests were carried out on the two data sets, indicating a strong significant direct correlation with each other, as presented in

Table 2. The three correlation matrices for the NDVI and NDVI-derived synthetic LAI time series from 2008 to 2016. Bold values are significant at p < 0.05.

No.	Correlation Coefficient	Value
1.	Pearson’s	0.99
2.	Spearman’s	1
3.	Kendall’s rank	1

. The closer the value is to 1, the stronger the relationship.

The vegetation dynamics and their feedback to climatic circumstances and anthropogenic activities from a watershed ecology outlook continue to be poorly anticipated [96]. The NDVI is a key indicator agreed upon for describing vegetation subsidized organic matter and green vegetation coverage and have been far and wide engaged in the field of vegetation variability, both locally and globally [97,98,99]. Thus, the current study also aims to understand the vegetation dynamics in the vital and most heavily populated LRB located in TP, in which the ecosystem facility is diligently linked not only to the welfare of inhabitants in the lower reaches and feasible development of the basin but the entire TP [100,101].

The MK test has revealed a surprising trend in vegetation and meteorological parameters of LRB for the daily time scale during the years of 2008–2016, as presented in

Table 3. MK-S trend on synthetic LAI time series, precursor MODIS NDVI time series, and average recorded temperature (°C) and rainfall (mm) from three meteorological station data time series, respectively, for the years 2008–2016 at p < 0.05. MK-τ represents the MK trend where (−) sign indicates a decrease with a magnitude as estimated by Sen’s slope (S).

	MODIS NDVI	NDVI Derived Synthetic LAI	Avg. T (3 Met. Stations)	Avg. P (3 Met. Stations)
MK-τ S	−0.009	−0.009	0.013	0.010
	0 day−1	0 day−1	0.0001 °C day−1	0 mm day−1

. The baseline NDVI and the daughter LAI time series is experiencing a nonsignificant decrease with a change equal to zero per day (as per Sen’s estimator). The daily average temperature and rainfall has been experiencing a nonsignificant increase in the area. The study referred to as [102] has investigated the NDVI variation in LRB by applying a zoning technique where the LRB has been spatially divided into zones based on human interferences and climatic changes in the area. Their study has concluded that the deviations in NDVI compelled by climatic elements merely engaged 16.19% of the total river basin and has identified a greater impact of P than that of T on vegetation growth in LRB. Additionally, for approximately 70% of the area in LRB, vegetation growth was not subjective to climatic factors, hence, specifying the more recognizable human impacts. Land use alteration from cropland/grassland into urbanization is designated as the cause for the decrease in NDVI [102].

3.2. SWAT Modeling Scenarios for LR Basin Streamflow Simulation

For the current study, the three scenarios put up for the cumulative streamflow simulation of LR recorded at the watershed outlet have been able to clearly attest the assimilation of synthetic LAI developed from remotely sensed NDVI data products as a significant addition into basic SWAT model for a better representation of the hydrological modeling of LR streamflow at a daily time step, increasing the trustworthiness of the model simulations [64,103,104]. All the SWAT model scenarios were developed with a preliminary 1-year warm up of 2008 and then calibrated for years 2009–2012, followed by validation for years 2013–2016 with a basic set of nine identified sensitive parameters [105,106] using SWAT-CUP embedded SUFI-2 algorithm are explained in

Table 4. The SUFI-2 identified sensitive parameters for LR cumulative streamflow simulation process in all the developed SWAT scenarios for the study.

No.	Parameter	Parameter Description	Method Chosen
1.	CN2	Initial SCS curve number for soil condition II	Relative
2.	GW_DELAY	Ground water delay (days)	Replace
3.	GW_REVAP	Ground water “revap” coefficient	Replace
4.	ESCO	Soil evaporation compensation factor	Replace
5.	EPCO	Plant uptake compensation factor	Replace
6.	SOL_BD	Soil bulk density (mg/m3)	Relative
7.	SOL_K	Saturated hydraulic conductivity (mm/h)	Relative
8.	SOL_AWC	Available water capacity of soil layer (mm H2O/mm soil)	Relative
9.	OV_N	Manning’s “n” value for overland flow	Relative
10.	BLAI	Maximum potential leaf area index	-
11.	PHU	Potential heat unit	-
12.	LAI_INIT	Initial leaf area index	-

The parameters differentiated as “blue” in the table are the SUFI-2-idenified sensitive parameters for the hydrological simulation of LR streamflow in SWAT-B scenario. The following parameters differentiated as “green” are the SUFI-2-identified sensitive vegetation and plant growth parameters added into the hydrological simulation process in the SWAT-LAI and SWAT-SLAI scenarios, taking into account the vegetation dynamics for the LR cumulative streamflow simulation process.

(differentiated in blue) at a daily time step. However, additional parameters included in the vegetation and plant database of SWAT modeling were subsequently introduced into the cumulative streamflow simulation process at a daily time step, as elaborated in Table 4 (differentiated in green). The parameters presented in Table 4 were tuned to fit the best SWAT model simulation in all the three scenarios.

3.2.1. SWAT-B (SWAT-Baseline) Scenario Outcomes

The SWAT-B scenario represents the original SWAT daily streamflow simulation in comparison to the cumulative daily streamflow collected at the LR watershed outlet. The scenario has been developed as a simulation standard where the identified sensitive parameters differentiated as blue in Table 4 were influencing the LR streamflow simulation process. The model has ranked SOL_BD, CN2, SOL_K, and SOL_AWC (after the sensitivity analysis carried out in SUFI-2) as the powerful controlling parameters in LR cumulative streamflow simulation (presented in

Table 5. SUFI-2-calibrated parameters and their ranks in LR daily streamflow simulation under the simulation scenarios for the study.

		SWAT-B Scenario			SWAT-LAI Scenario			SWAT-SLAI Scenario
No.	Parameter	Range ·(Min–Max)	Fitted Value	Rank	Range ·(Min–Max)	Fitted Value	Rank	Range ·(Min–Max)	Fitted Value	Rank
1.	r__SOL_BD	−1–1	0.22	1	−1–1	0.22	1	−1–1	0.38	1
2.	r__CN2	−0.25–0.01	−0.1	2	−0.02–0.01	−0.1	3	−0.02–0.01	−0.1	5
3.	r__SOL_K	−1–1	−0.5	3	−1–1	−0.5	4	−1–1	−0.5	6
4.	r__SOL_AWC	−1–1	0.49	4	−1–1	0.5	6	−1–1	0.5	7
5.	v__GW_DELAY	150–500	235	5	150–500	235	8	150–500	235	8
6.	v__GW_REVAP	0.02–0.1	0.02	6	0.02–0.1	0.02	9	0.02–0.1	0.02	9
7.	r__OV_N	−1–1	−0.66	7	−1–1	−0.54	10	−1–1	−0.78	10
8.	v__EPCO	0–1	0.78	8	0–1	0.65	11	0–1	0.44	11
9.	v__ESCO	0.01–1	0.64	9	0.01–1	0.69	12	0.01–1	0.64	12
10.	BLAI	-	-	-	0–2	0.08	5	0–5	0.9	4
11.	PHU	-	-	-	0–1802	1543	7	0–1854	1607	2
12.	LAI_INIT	-	-	-	0–3	0.62	2	0–2	0.37	3

), which is fairly in correspondence to our previous work [105], again highlighting the importance of soil physical characteristics, soil water movement, and the land use and management practices adopted in the LR watershed in the hydrological simulation process. The calibration and validation years for SWAT-B scenario are presented in Figure 5 (panel a). We see a close correspondence between both the curves representing the gauged and simulated flow during the calibration time span, which is further confirmed through the evaluation standards of high R² and NSE values (>0.5) with a fairly lower PBIAS (<25%), as presented in

Table 6. SWAT model performance in simulating the daily LR streamflow in the three scenarios developed in the current study.

No.	Parameter	SWAT-B Scenario		SWAT-LAI Scenario		SWAT-SLAI Scenario
		Calibration	Validation	Calibration	Validation	Calibration	Validation
1.	R2	0.86	0.40	0.88	0.58	0.93	0.90
2.	NSE	0.82	0.38	0.81	0.49	0.89	0.83
3.	PBIAS	18%	24%	16.8%	19%	8%	11%

. The lower flows are captured very well by the model during the calibration years; however, high peaks have been underestimated during the calibration years. The study of [107] has also testified an underestimation of LR discharge by SWAT model. Similar behavior of the SWAT model has been reported in the studies carried out for PoKo [108], Cong [109], and Nam Kim [110] watersheds in Vietnam, where the model fulfilled the overall streamflow simulation criteria, however, could not capture some peaks precisely. For the validation years, the two curves are not well in compliance to the standard evaluation measures as the R² and NSE values are seen to be lower (<0.5), yet with a stable PBIAS of SWAT simulated values of LR streamflow from the gauged daily streamflow, as presented in Table 6. One of the reason for incompatible simulation curves could be the rainfall variability witnessed during the validation years from 2013 to 2016. This rainfall variability has already been described in our study [106] for LRB. Additionally, the LR streamflow has been subjected to multiple operating factors, resulting in a peculiar behavior over the years, as reported by [105,106].

3.2.2. SWAT-LAI (SWAT-Leaf Area Index) Scenario Outcomes

The study progressed with the development of SWAT-LAI scenario where, along with the SUFI-2-identified nine sensitive parameters in SWAT-B scenario, three additional parameters highlighted in green in Table 4 were introduced into the streamflow simulation process to investigate the influence of plant and vegetation parameters in the LR Basin hydrological simulation phenomena. For the similar time span of 2008 as a warm-up, 2009–2012 for calibration, followed by a validation from 2013–2016, SWAT model involving the original LAI parameter embedded in the plant and vegetation growth module was engaged for the daily LR streamflow simulation. The model discretized the LR watershed in 21 sub-basins and 149 HRUs. The LAI was simulated by the model for all the developed HRUs using the PHU parameter as described in Equations (1) and (2). After the tuning of all the sensitive parameters in SUFI-2, the model ranked SOL_BD, LAI_INIT, CN2, and SOL_K as the most influential in LR streamflow simulation under the SWAT-LAI scenario presented in Table 5. These findings again confirm the importance of soil physical properties and soil water movement along with the land use practices in LRB in the hydrological simulation of the LR cumulative streamflow collected at the basin’s outlet. These results are objectively in line with our previous work on SWAT LR streamflow simulation [105,106]. However, in the current scenario, SWAT model has highlighted the simulated LAI as one of the influential parameters in the hydrological simulation process. This brings us to the assumption that calibration of the LAI parameter can serve as an alternative to identify the role of vegetation dynamics in the LR cumulative streamflow simulation using the SWAT model vegetation-related parameters. Here, it is important to mention that SWAT-simulated LAI was found to be homogenous for almost all the HRUs with a value of ~0. Similar findings have been reported by [111], where the original SWAT-simulated LAI at the HRU level was 0 for the barren land and slightly higher for the grassland vegetation type in Bayin River watershed, a northwestern QTP catchment located in China.

The presentation of SWAT model in simulating the LR streamflow received at the basin’s outlet, while included LAI and the associated parameter, is presented in Figure 5 (panel b) and Table 6. An evident improved calibration of LR streamflow has been shown by the model while working under SWAT-LAI scenario. The high peaks are captured quite decently during the simulation of LR streamflow. However, the lower flows are slightly overestimated by SWAT model and the bias in simulating the LR discharge is lowering as compared to the SWAT-B scenario. The R² and NSE values have also improved for the SWAT-LAI scenario in comparison to the SWAT-B scenario. For the validation of the SWAT model in the current scenario, similar behavior of capturing the majority of the high peaks in a decent manner has been displayed, despite large variation in the rainfall pattern recorded for LRB during the validation time span. Nevertheless, the lower flows are again overestimated by the model; still, the performance of the model is a clear improvement by engaging SWAT-LAI parameter in the current scenario in terms of R², NSE values, and bias of simulated flow from the observed flow under SWAT-LAI.

3.2.3. SWAT-SLAI (SWAT-Synthetic Leaf Area Index) Scenario Outcomes

SWAT-SLAI scenario was developed after dispensing the synthetic MODIS NDVI-derived daily LAI database in the 149 SWAT HRU files for the current study. The original SWAT-simulated LAI was replaced with the daily synthetic LAI database for LRB, where the LAI varied along with the seasonal cycle of LRB, i.e., lower NDVI values for a dry cold winter season generating lower LAI values, thus indicating lower vegetation and vice versa. As already discussed in Section 3.1, the vegetation in LRB is found to be more reliant on and influenced by P rather than T; therefore, during the hot wet summer months, higher NDVI values were achieved, which, in turn, generated higher LAI values for an established vegetation cover in LRB during that particular time. The synthetic LAI values were incorporated into SWAT model and all the sensitive parameters influencing the hydrological process (highlighted as blue in Table 4) along with the vegetation parameters (highlighted as green in Table 4) were tuned and settled for their fittest value in SUFI-2 module of SWAT-CUP, as represented in Table 5. The land use classes generated as the SWAT model input shown in Figure 3a were specifically identified while calibrating the PHU parameter. SWAT model, while utilizing the synthetic LAI as the refined and high-resolution vegetation database for LR, prioritized the influential parameters as SOL_BD, PHU, LAI_INIT, and BLAI as being the dominant parameters affecting the hydrological simulation process of LR cumulative streamflow. The SWAT model has performed extremely well in capturing the higher as well as lower flows for both the calibration and validation time period. The representation of SWAT-SLAI scenario results in Figure 5 (panel c) is a clear improvement in the SWAT model’s ability to simulate the cumulative LR streamflow. The findings of the SWAT modeling in the current scenario has emphasized on the interrelationship of vegetation dynamics, soil physical properties, and the hydrological simulation of LR streamflow. The synthetic remotely sensed LAI along with associated heat unit are a physical representation of the vegetation cover in the LRB. This brings us to the fact that physical realization of vegetation cover and dynamics is instrumental in understanding the hydrological behavior of the LRB and vice versa. Our findings are strongly in line with the findings of [71].

To showcase a relationship between the LR discharge collected at the basin’s outlet and the simulated LR discharge sets produced in all the three SWAT model scenarios, the statistical analyses carried out on all the hydrological data series revealed a significant and close association of simulated LR discharge (all three scenarios) with the cumulative LR discharge recorded at basin’s outlet. The correlation coefficients are seen to attain higher values (greater than 0.5) for all the simulated LR hydrological series. In particular, the highest and closest correspondence has been displayed by the SWAT-SLAI-simulated LR discharge with the observed discharge values. The results of the correlation analyses are presented in

Table 7. Statistical correlation analyses for LR daily observed and simulated streamflow under the three SWAT model scenarios. Bold values are significant at p < 0.05.

No.	Correlation Coefficient	SWAT-B Scenario	SWAT-LAI Scenario	SWAT-SLAI Scenario
1.	Pearson’s	0.82	0.83	0.95
2.	Spearman’s	0.84	0.77	0.93
3.	Kendall’s rank	0.65	0.59	0.79

.

3.3. Comprehending the Role of LAI in the LR Hydrological Behavior

Many studies have already agreed on the integration of vegetation-based parameters in the hydrological distributive and rainfall–runoff models. Studies show that vegetation processes have major effects on runoff produced by controlling vegetation evapotranspiration and influencing soil moisture conditions [112,113]. Coupling the SWAT model with the remotely sensed (MODIS) LAI product has been applied for enhanced modeling of tropical and subtropical green vegetation dynamics and crop arrangements in semi-arid regions [56,58]. The enhancement of SWAT in these studies was assured by the remotely sensed LAI, which could precisely apprehend a more genuine plant phenology and configuration at a higher resolution [111]. Even though remotely sensed LAI is an indirect measure, it is nevertheless considered a better estimate than standard LAI [114]. In agreement to the findings of these studies, in the current study, the original SWAT model was nurtured with high-resolution physically authenticated vegetation conditions of LRB and has resulted in a visibly improved depiction of LR cumulative streamflow at the basin’s outlet. The synthetic LAI has been recognized as the pivotal representative of vegetation features of LRB in the improved hydrological simulation process. On developing a relationship between synthetic LAI and the LR discharge recorded at the basin’s outlet, the SWAT-B scenario simulated, the SWAT-LAI scenario simulated, and the SWAT-SLAI scenario simulated presented in Figure 6, we see an almost similar conduct of synthetic LAI towards the different LR streamflow data, where a majority of the values are enclosed within the 95% prediction intervals.

To further affirm the association of remotely sensed synthetic LAI with the LR streamflow datasets presented in Figure 6, the correspondence between them was exposed to statistical tests. The results presented in

Table 8. The correlation matrices for synthetic LAI with the observed, SWAT-B scenario generated, SWAT-LAI scenario generated, and SWAT-SLAI generated LR streamflow. Bold values are significant at p < 0.05.

No.	Correlation Coefficient	Observed Q	SWAT-B Scenario Based Q	SWAT-LAI Scenario Based Q	SWAT-SLAI Scenario Based Q
1.	Pearson’s	0.74	0.61	0.65	0.73
2.	Spearman’s	0.78	0.65	0.64	0.75
3.	Kendall’s rank	0.60	0.58	0.51	0.57

The columns highlighted represent the closely similar values of correlation coefficients for the observed Lhasa River discharge and SWAT-SLAI-simulated Lhasa River discharge, revealing an almost similar association of observed and SWAT-SLAI-simulated Lhasa River discharge with the synthetic LAI time series.

reveal a close significant direct relationship of synthetic LAI with all the LR streamflow datasets with the correlation values > 0.5 for all the three correlation coefficients. Thus, synthetic LAI agrees with the LR streamflow seasonal variation in a parallel manner. The smaller the synthetic LAI value, the lower the vegetation cover and vice versa. Higher vegetation cover is reached during the high-rainfall hot summer months, which is also a high flow time in LRB. Furthermore, the Pearson’s, Spearman’s, and Kendall rank’s correlation coefficient values for synthetic LAI with the actual recorded LR discharge at the basin’s outlet and the SWAT-SLAI-simulated LR discharge are very close to each other (as highlighted in Table 8). This certainly confirms the success of establishment of the SWAT-SLAI scenario for daily LR streamflow simulation using high-resolution synthetic LAI time series in the process.

The LR discharge is susceptible to extensive influential agents and, thus, a trend analysis can aid in the development of better planning and management strategy in LRB to sustain the water requirement and availability in one of the highly populated QTP catchments [102]. Therefore, the current study has identified the trend on LR streamflow by applying the renowned nonparametric MK trend test on the gauged and all the hydrological datasets generated by the developed scenarios in the study and presented in

Table 9. MK-S trend on observed, SWAT-B scenario simulated, SWAT-LAI simulated, and SWAT-SLAI simulated daily LR streamflow for the years 2009–2016 (2008 is excluded being the warm-up year for all the simulation scenarios) at p = 0.05. MK-τ represents the MK trend with a magnitude as estimated by Sen’s slope (S). Bold values are significant at p value.

	Observed Q	SWAT-B Scenario Based Q	SWAT-LAI Scenario Based Q	SWAT-SLAI Scenario Based Q
MK-τ	0.05	0.12	0.17	0.06
S	0.006 m3 s−1 day−1	0.012 m3 s−1 day−1	0.045 m3 s−1 day−1	0.008 m3 s−1 day−1

The highlighted MK-S test results for the observed and SWAT-SLAI simulated Lhasa River discharge reveal an analogous trend and its measure.

. It is noteworthy that daily LR discharge experienced a significant increase for the time span of 2009–2016 (2008 has been excluded as being the warm-up year for all the SWAT model simulation scenarios) for the gauged as well as the simulated (all scenarios) LR streamflow.

Here, again, we see that all the LR streamflow datasets reveal the same hydrological pattern for LRB, but the most imperative deduction is the comparable trend (MK-τ value) and its estimate (as calculated by Sen’s slope, S) for the observed and the SWAT-SLAI scenario generated LR streamflow, as highlighted in Table 9. This brings us to a strong belief in the vitality of LAI estimation and then its accumulation for the enhanced and physically realistic LR streamflow simulation using the SWAT model. The findings of the study hold importance for similar watersheds where vegetation, streamflow, and climate variables are in close interaction with each other. The improved hydrological simulation process can aid water resource managers and hydrological modelers in better decision-making and strategy development for vital water resource management around the world.

4. Discussion

4.1. NDVI-LAI Relationship and Their Conduct in LRB

Owing to dissimilarities in weather pattern and grassland classification with other parts of China, the previous models estimating grassland LAI may not be suitable for southern grasslands [48]. Several studies have presented the extensively used spectral reflectance index NDVI as a good estimator of LAI and has been used to estimate LAI indirectly [115,116,117,118,119]. Since the NDVI values are attained by a downward-looking sensor, the method is beneficial for short-stature vegetation [119]. Recognizing the importance of the reliability of the synthetic LAI data series developed from the MODIS NDVI (MOD13Q1) satellite imagery, it is reported in the study by [120] that high-temporal-resolution satellites, such as MODIS, yet providing low-spatial-resolution data, are especially suitable for vegetation estimation over large areas. Moreover, contrary to high-resolution satellite missions such as Sentinel 2, MODIS operationally produces temporally aggregated images (i.e., composites), which reflect the most steadfast observations within a time window and, therefore, are slightly perturbed by cloud cover [121]. In the present study, the LAI data have been prepared by the remotely sensed MODIS NDVI data product, ensuring an improved spatial heterogeneity for LRB. The modified Beer−Lambert law delivered LAI time series on a logarithmic scale, where daughter LAI dataset has a close correspondence with parent NDVI values. The credibility of the adopted approach of using the remote-sensing-based model by incorporating a well-suited vegetation index (NDVI) for the southern Chinese catchment is supported by the study of [13]. In their study, appraisal of two different regression models, Caraux-Garson and Beer, was executed to develop the LAI dataset from NDVI. The application of the Beer’s law was verified to be intrinsically superior and well suited for environments characterized by low vegetation density and rapidly evolving cover types ranging from bare-soil conditions to high LAI values. Likewise, the other benefit is in the likelihood to estimate the Beer’s law coefficients from the available dataset attaining more realistic LAI values for the vegetation covers of the study site. Likewise, the modified Beer−Lambert law employed in the current study is developed particularly for the southern Chinese catchments where the coefficient values are estimated during the month of July for the thriving vegetation conditions. The similar approach for LAI estimation can be replicated in other parts of grassland ecosystem around the globe by developing the respective coefficient values for the desired study site.

NDVI, being the ancestor to the synthetic LAI, is facing a decreased trend despite an increased tendency of P and T. Congruently, the synthetic LAI is also experiencing a similar decrease over study time span in divergence to the P and T trend for the similar time span. A similar decrease in NDVI has been reported by [102], which is attributed to the increased human interference with the ecological situation of LRB, where the land use transformation in the form of obstructing and water reservoir building measures may obliterate the riparian vegetation area. The study of [122] recognized that the impoundment of the Three Gorges Project inundated enormous vegetation cover, and the region obscured by vegetation in the upper reaches declined by 3.31 km². The development of water conservancy ventures has led to major alterations in natural vegetation in the reservoir region [123]. In the Lhasa River Basin, Zhikong-Pangduo reservoir group structure conveyed extreme disruption to the vegetation in the central segment of the reservoir group. The influence range of reservoir construction on vegetation in the middle section is found to be contained by 6 km of the river, where the impairment of vegetation cover is seen to be the highest within 1 km. The dual water reservoir assembly development at the expense of subjugating the grassland has been the core reason of vegetation degeneration. Thus, a plethora of factors are posing their simultaneous effect on the eco-hydrological behavior of LRB. For their study, [124] applied the SWAT model to assess the potential impact of climate change on dam inflow by considering future vegetation cover conditions. While making use of MOD15A2 LAI data product for Chungju Dam watershed in North Korea, the authors found a nonlinear relationship of LAI with temperature in the summer high-vegetation months. However, the increased temperature can positively aid in the advancement of the phenological period of vegetation and proliferate the vegetation greenness; conversely, the intensified transpiration water losses by plants and soil interrelated with rising temperatures may lessen the positive effects of temperature and aggravate the demand for water. For this reason, vegetation growth may be inhibited due to drought.

4.2. Benefit of LAI Incorporation in SWAT Model for LR Simulation Enhancement

We see that SWAT modeling was progressively improved as the vegetation factors were incorporated into the model in the form of LAI and the associated parameters. LAI is an important parameter of SWAT as it affects a series of processes, such as ET and infiltration; therefore, accurate simulations of the LAI are a key to accurate hydrological simulations. However, in the standard plant growth module of SWAT model, whereby the heat unit management option is selected (“PHU-potential heat unit”), the start and the end of the vegetation growth cycle occur at the default values of 0.15 and 1.2, respectively. With this management setting, the simulated LAI is zero at the beginning of each simulation year, as reported by [57,125,126]. In the current study, a similar homogenous LAI has been calculated for LRB along the year for all the 149 HRUs in the SWAT-LAI scenario, with a value of approximately zero. However, SWAT-LAI scenario yielded better results as compared to the SWAT-B scenario in terms of evaluation criteria. Further advancing with the study, on dispensing the heterogeneous synthetic LAI into the HRUs, SWAT model’s performance improved vividly. The synthetic LAI dataset imparted enhanced physical vegetation conditions of LRB into the model. This led to a better hydrological simulation of LR streamflow by SWAT model. In the studies by [55,71,111], a clear enhancement of hydrological simulation was reported by the SWAT model after utilizing remotely sensed heterogeneous LAI datasets in their studies.

Predominantly, for LRB, LR discharge is delivered by precipitation [107], and inconsistency in rainfall proclaims a direct impact on the hydrological events of the LRB. Temperature is, likewise, an imperative aspect in defining hydrological occurrences, for it affirms its effect through evapotranspiration and, thus, deviations in the temperature of the LRB may have a probable influence on the water assets in the region [106]. However, as per [102], the vegetation pattern and dynamics in LRB are more sensitive to precipitation as compared to the temperature. As an additional benefit, synthetic LAI also corresponds closely to the precipitation received during the time span for which the LAI has been developed, as represented in Figure 7. We can see a similar variation in synthetic LAI as the precipitation.

So, we can easily accomplish that the LR discharge is rainfall-dependent. Vegetation similarly is controlled by rainfall if considering the climatic factors only and neglecting the strongly influential human disturbances. Correspondingly, LAI is also influenced in accordance to the rainfall as per the findings of our study. Thus, there exists a relationship between LAI and LR discharge, as both are related to the same climate variable (P) and in a similar way. This interesting interrelationship is a vital code in identifying and understanding the importance of vegetation in the hydrological phenomena of LRB and vice versa. Additionally, LAI holds a special value in serving as a proxy to the vegetation dynamics in SWAT simulation of the hydrological process of LRB. This finding is opening doors to very instrumental research ideas in the future for the same or other similar watersheds around the globe. While applying physically based models such as MIKE SHE, the use of remotely sensed LAI seems recognizable, as the study by [114] demonstrates considerable improvement in the discharge simulation for Senegal River Basin as compared to using predefined LAI. The major improvement appears to stem from the year-to-year variation introduced by the remotely sensed LAI. Similarly, in the current study, the assimilation of remotely sensed LAI into the SWAT model in SWAT-SLAI scenario has proven to be a vital addition for the improved cumulative streamflow simulation as compared to the two previous scenarios where the model achieved the best R² and NSE and lowest PBIAS from the observed LR discharge. The verification of LAI assimilation in the hydrological process is further supported by [127], where their study has confirmed a better calibration and runoff prediction by using MODIS LAI incorporated into Xinanjiang model for 210 ungauged catchments in China. This verifies the importance of vegetation dynamics as an input for an improved hydrological modeling process. Furthermore, the statistical analyses carried out on all the three simulated LR discharge series has shown that SWAT-SLAI produced LR discharge corresponding very closely to the observed hydrological data. This verifies that the remotely sensed LAI has proved as a reliable alternative for the physical vegetation characteristics of LRB.

However, though we have achieved successful development of synthetic LAI from remotely sensed fine-resolution MODIS NDVI products, measured ground data unavailability for the validation of empirically estimated vegetation dynamics in LRB can be a concern regarding the actual behavior of vegetation dynamics in the area. Similarly, using LAI remote sensing data products can also impart uncertainty to the findings of remote-sensing-based studies for vegetation dynamics, where LAI can be a central parameter for vegetation dynamics and greenness change detection for multiple purposes [128].

LRB is a transboundary data-sensitive region with high social and political value. Physical ground data availability and sharing is one of the major concerns in the area. Thus, the current study has made use of the weather and hydrological ground-observed data record that could be accessed and was easily available. However, this could be a limitation of the study, but the findings of the current study are based on the real-time observations of the study area.

5. Conclusions

The current study emphasized on the depiction of human and climatic factors influencing LR streamflow under the assimilation of a remotely sensed vegetation input database into the well-known eco-hydrological SWAT model. While carrying out the current study, we found:

• Replacing the remotely sensed LAI products with a remotely sensed MODIS NDVI source, the synthetic LAI generated after applying the modified Beer−Lambert law particularly developed for southern Chinese grassland, the parent NDVI produced the log transformed daughter synthetic LAI values, where NDVI and synthetic LAI values very strongly correspond to each other.
• The MK trend on daily NDVI values is decreasing, which is attributed to human disturbances in the form of grassland transformation as a result of aggravated urbanization practices in the form of utilizing the land for water conservation and reservoir construction projects. Equally, the daily synthetic LAI is experiencing a decreased MK trend in contrast with the increasing trend in temperature and rainfall. There exists a close relationship between rainfall and LAI, as vegetation is affected more by rainfall received in the LRB as compared to the temperature. Additionally, the LR discharge is controlled by the rainfall received in LRB. Thus, it can be assuredly concluded that LAI is closely associated with the LR discharge, as both the variables reveal a close connection with the same climatic variable. This is further supported by the MK trend analysis on the observed and SWAT-SLAI-simulated LR discharge revealing an almost similar behavior for the study time span.
• The SWAT model, when equipped with the physically real-time vegetation database in the form of synthetic LAI (during SWAT-SLAI scenario), has performed tremendously well in capturing the cumulative streamflow simulation process on a daily time step. The performance of the model during SWAT-SLAI scenario is better than the other two developed scenarios. This brings us to the conclusion that dispensing the vegetation data in the hydrological simulation phenomena is a success for hydrologically and climatically sensitive regions in the world.