Quantifying the isolated and integrated impacts of land use (LU) and climate change on streamflow is challenging as well as crucial to optimally manage water resources in river basins. This paper presents a simple hydrologic modeling-based approach to segregate the impacts of land use and climate change on the streamflow of a river basin. The upper Ganga basin (UGB) in India is selected as the case study to carry out the analysis. Streamflow in the river basin is modeled using a calibrated variable infiltration capacity (VIC) hydrologic model. The approach involves development of three scenarios to understand the influence of land use and climate on streamflow. The first scenario assesses the sensitivity of streamflow to land use changes under invariant climate. The second scenario determines the change in streamflow due to change in climate assuming constant land use. The third scenario estimates the combined effect of changing land use and climate over the streamflow of the basin. Based on the results obtained from the three scenarios, quantification of isolated impacts of land use and climate change on streamflow is addressed. Future projections of climate are obtained from dynamically downscaled simulations of six general circulation models (GCMs) available from the Coordinated Regional Downscaling Experiment (CORDEX) project. Uncertainties associated with the GCMs and emission scenarios are quantified in the analysis. Results for the case study indicate that streamflow is highly sensitive to change in urban areas and moderately sensitive to change in cropland areas. However, variations in streamflow generally reproduce the variations in precipitation. The combined effect of land use and climate on streamflow is observed to be more pronounced compared to their individual impacts in the basin. It is observed from the isolated effects of land use and climate change that climate has a more dominant impact on streamflow in the region. The approach proposed in this paper is applicable to any river basin to isolate the impacts of land use change and climate change on the streamflow.
Land use (LU) and climate are the drivers of hydrologic processes in a river basin (Vörösmarty et al., 2000; Nijssen et al., 2001; Oki and Kanae, 2006; Wada et al., 2011). Change in LU is observed to influence the hydrological cycle and the availability of water resources by altering interception, infiltration rate, albedo and evapotranspiration (ET) (Rose and Peters, 2001; Scanlon et al., 2007; Rientjes et al., 2011). Climate in contrast affects the basic components of hydrologic cycle such as precipitation, soil moisture, evaporation and atmospheric water content (Gleick, 1986; Wang et al., 2008). Therefore, understanding the hydrologic response of a river basin to changes in LU and climate forms a critical step towards water resources planning and management (Vörösmarty et al., 2000). Moreover, with increase in scarcity of water resources, hydrologic impacts of LU and climate change have drawn significant attention from the hydrologic community (Scanlon et al., 2007). In this regard, several studies have been carried out that focus on understanding exclusive impacts of either of the two drivers (Hamlet and Lettenmaier, 1999; Christensen and Lettenmaier, 2007; Beyene et al., 2010; Wagner et al., 2013; Islam et al., 2014). However, optimum management of water resources in a river basin needs an in-depth understanding of the isolated and integrated effects of LU and climate on streamflow. Due to complex response of streamflow to combined effects of LU and climate change (Fu et al., 2007; Guo et al., 2008), very few studies have been carried out on this aspect (Mango et al., 2011; Guo et al., 2008; Cuo et al., 2013; Wang et al., 2013). Segregating the individual contribution of LU and climate to streamflow has recently become the focus of scientific work (Wang and Hejazi, 2011; Wang et al., 2013; Renner et al., 2012, 2014).
Methods used to assess the impacts of LU and climate on streamflow can be broadly classified into four categories: (i) experimental paired catchment approach, (ii) statistical techniques such as Mann–Kendall test, (iii) empirical or conceptual models and (iv) distributed physically based hydrologic models. Among these techniques, the paired catchment approach is most difficult but often considered as the best approach for smaller catchments. However, applicability of the paired catchment approach over large catchments may not be possible (Lørup et al., 1998) since it requires years of continuous monitoring to gather sufficient data for the analysis. Statistical trend detection tests have been proved to be very useful in qualitatively determining the presence of a significant trend in the time series along with direction and rate of change (Zhang et al., 2008; Li et al., 2009). But these techniques cannot be used for quantifying the change and attributing it to a particular cause due to a lack of a physical mechanism (Li et al., 2009). Empirical or conceptual models are simple hydrologic models that require only a few parameters to simulate a catchment. However, a major drawback with these models is that the parameters may not be directly related to the physical conditions of the catchment, and thus may lack the ability to correctly represent a catchment. Therefore, one is left with the option of using distributed physically based hydrologic models, which are by far the most appealing tools to carry out impact assessment studies (Ott and Uhlenbrook, 2004; Mango et al., 2011; Wang et al., 2012). These models operate within a distributed framework to take physical and meteorological conditions of the basin into account (Refsgaard and Knudsen, 1996). Physically distributed models include both fully distributed and semi-distributed models. Owing to their extensive parameterization, fully distributed models are difficult to employ at a large catchment scale which make comparatively less data-intensive semi-distributed models a practical alternative. This paper presents a simple hydrologic modeling-based approach to isolate the impacts of land use and climate on streamflow. For this purpose, a physically based macroscale variable infiltration capacity (VIC) hydrologic model (Liang et al., 1994) has been employed for the analysis.
Location map and details of the UGB.
In the present paper, Ganga River basin in India is selected as the case
study to perform the analysis. Few studies have been reported in literature
(Nijssen et al., 2001; Arora and Boer, 2001; Nohara et al., 2006) wherein
Ganga basin is studied alongside other major river basins of the world (to
assess the effect of changing climate on flow regime); however, there is
a dearth of studies that comprehensively examine the effects of LU and climate
change on streamflow exclusively in this basin. Originating from the
Himalayas, the Ganga River traverses a stretch of 2525 km covering a
catchment area of around 800 000 km
In order to obtain the isolated impacts of LU and climate change on streamflow, the following objectives are addressed in the current work: (i) assessing sensitivity of the streamflow to changes in different LU categories, (ii) examining impacts of climate change on the streamflow and (iii) analyzing integrated impacts of LU and climate change on the streamflow. The three objectives are translated into three scenarios wherein the first two scenarios quantify the independent effects of LU and climate on streamflow under their invariant counterparts; i.e., climate and LU respectively are kept constant. The third scenario deals with concurrent changes in LU and climate. Results from the three scenarios are further used to segregate the hydrologic impacts of LU and climate change. The aforementioned objectives are investigated over the UGB as a case study by employing a calibrated and validated VIC model to simulate streamflows. To assess the impact of future climate on streamflow in the basin, dynamically downscaled climate simulations for six general circulation models (GCMs) obtained from the Coordinated Regional Downscaling Experiment (CORDEX) project are used. Climate change related analyses are carried out under the uncertainty framework to address two issues: (1) climate model-based uncertainties and (2) emission-scenario-based uncertainties.
The UGB (25
In this study, the UGB is divided into three regions, upstream, midstream
and downstream (Fig. 1), based on altitude, topography and land use
characteristics. The upstream region is highly mountainous, characterized by
glaciers and dense forests, with elevations from 297 to 7796 m. From
upstream to midstream regions, there is a transition from hills to plains.
The midstream region is dominated by forests and croplands with elevations
ranging from 75 to 3079 m. The downstream region is mostly covered by croplands with consistent elevations of around 100 m. In addition to the
varying land use characteristics, these three regions have different
climatology as well. From 1971 to 2005, upstream, midstream and downstream
regions recorded an average annual precipitation of 1294, 1009 and 826 mm,
respectively. Most of the precipitation was concentrated during the monsoon
months from June to September (JJAS). Average annual temperatures across the
three regions during the same period were 20, 23 and
26
The current study employs physically based VIC hydrologic model for the analysis. The VIC model is a semi-distributed soil–vegetation–atmosphere transfer model that solves coupled water and energy balance equations in a grid to calculate different hydrologic components (Liang et al., 1994). Within a grid the VIC model considers sub-grid heterogeneity by dividing each grid cell into a number of tiles which in turn depend on different land use types present in the grid. Each tile generates different responses to precipitation in the form of infiltration, soil moisture storage, runoff and evaporation, owing to a difference in land surface properties. When VIC concludes the computation of energy and water balance calculations for each grid within the watershed, a streamflow routing model developed by Lohmann et al. (1998) is activated, which transports the surface runoff generated within a grid along with the baseflow to the outlet of the grid cell which is further routed through the river channel to the watershed outlet.
Taylor diagram for
Hydrologic models in general require topographic, soil, hydro-meteorological
and LU data which can be procured from various sources. In the present work,
topographic information is obtained from the ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) DEM (digital elevation model) available
at 30 m spatial resolution. The digital soil map for the region is procured from
the National Bureau of Soil Survey and Land Use Planning, India, at a scale of
1 : 250 000. Meteorological data (rainfall, maximum temperature, minimum
temperature and wind speed) for the period 1971–2005 at daily timescales are
procured from two sources: the Indian Meteorological Department (IMD) (Rajeevan
et al., 2006) and Princeton University (PU) (Sheffield et al., 2006).
Meteorological data from both sources are brought to a common grid
resolution of 0.5
For LU data, landsat imageries for the years 1973, 1980, 2000 and 2011 are selected and then classified to determine the LU change in the basin over 4 decades. A field study is carried out to collect the training sites for image classification. The accuracy of classified images is obtained to be 89, 83, 88 and 79% for 1973, 1980, 2000 and 2011 images, respectively, which is seen to be generally good. Thus, the classified images can be used as LU maps of the UGB for the corresponding time periods. Results of classification and change in LU are presented in Sect. 3.1.
To carry out hydrologic impact studies related to climate change, one needs
data on future climate variables, such as rainfall (
GCMs from the CORDEX project used in the present study.
GCMs climatology compared with observed climatology for
monthly
It is now well known that large-scale pattern of climate variables simulated
by GCMs may be realistic, but when downscaled to a regional level, they may
exhibit significant bias compared to the observed data (Maurer and Hidalgo,
2008; Ghosh and Mujumdar, 2009). This can have a significant effect on
hydrological impact studies which necessitates the need of performing bias
correction on the climate variables obtained. In the current work, climate
variables obtained from the GCMs are bias corrected with IMD gridded data
(which are considered as observed data) at a daily scale using the technique
developed by Wood et al. (2002). A distribution function is fit to the
observed daily data and individual GCM data.
Statistics of GCM simulated (post-bias correction) and observed climate
variables for the upstream region are presented in a Taylor diagram (Fig. 2). It
can be observed that all the models are clustered together, which could be
due to the fact that all the GCM outputs are from the same modeling center,
and the clusters in the case of
Structure of the VIC model obtained for upstream and midstream regions along with the performance measures during calibration and validation phase.
In addition to the correlation coefficient, climatology of variables for
different GCMs is compared with the climatology of the observed variable
from 1971 to 2005 at a monthly scale. These results are presented in Fig. 3 for
one of the grid cells within the UGB. The observed and GCM climatology at
a monthly scale for the time period 1971–2005 is represented following Wood et
al. (2002). It can be seen in Fig. 3 that the GCMs successfully reproduce
the mean and variance of the rainfall climatology for most of the months.
However, for the post-monsoon period (i.e., October, November and December), GCMs
overestimate rainfall compared to the observed rainfall. For
In addition to the meteorological data and LU information, VIC requires
explicit information about the vegetation type in the study region. In the
study area, it is observed from the agricultural statistics
(
For the model calibration in the present work, three parameters as suggested
by Lohmann et al. (1998) are calibrated to obtain an optimum combination
such that the error between observed and simulated streamflow is at a minimum.
The three parameters considered are (i)
The VIC model is established independently for upstream, midstream and
downstream regions but model calibration is only possible for upstream
and midstream regions since
To perform model calibration, initially the sensitivity of the simulated
discharge to each of the three parameters is tested and their rough estimate
of range for both upstream and midstream regions are obtained. Within this
range, several candidate models for upstream and midstream regions are
created based on several plausible combinations of these three parameters.
The VIC model is executed for all the combinations and the one that has
the maximum predictive power in terms of coefficient of determination
(
Calibration results of
Validation results of
For the current work, the periods of 1987–1999 and 1977–1995 in upstream and
midstream regions, respectively, are considered for calibration. Figure 4
provides the plots of corresponding observed and calibrated VIC simulated
monthly streamflow series for the two regions. It can be observed from Fig. 4
that simulations during the calibration period captured the observed
pattern and magnitude of hydrograph very well. In particular, rising and
recession limbs of hydrographs are simulated accurately for both the
regions. Shortcomings in the VIC simulations for both the regions include
a mismatch of peak flows, which could be due to errors in modeling extreme
precipitation by the model. Since we are not dealing with extremes in the
present case study, this error is not of much concern. In addition, it may
also be observed that at the end of each recession limb, there is a sharp
drop, which is below the level of
The calibrated models are validated from 2000 to 2005 and from 1996 to 2005 for
upstream and midstream regions, respectively (presented in Fig. 5).
The streamflow pattern and magnitude of runoff are well simulated during
the validation. Table 2 presents optimum set of parameters for the two regions
along with their performance measures during calibration and validation.
Based on the performance measures, it is seen that the model is able to predict
Overview of the work.
LU maps for
Despite strong correlation between the model simulated and observed climate
variables (Fig. 2), it is noticed that the magnitude of uncertainty across
different models is quite large with respect to observed
In order to infer the confidence level in terms of climatology, the
classification considered by Maurer (2007) is used whereby a
confidence level (i)
Section 3.1 and 3.2 provide analysis pertaining to the quantification of changes observed in LU and climate. In Sect. 3.3, these results are used to quantify streamflow variations within the uncertainty framework.
Classification of landsat imageries resulted in LU maps for the UGB that are presented in Fig. 7. It can be observed that the UGB exhibits wide variations in LU wherein upstream parts are snow covered and downstream parts are cropland. The dominant LU type in the UGB is cropland that covers about 56 % of the entire basin (45, 53, 64 and 66 % for 1973, 1980, 2000 and 2011, respectively). Upon visual examination of figures, it is evident that from 1973 to 2011, area under forest in the upstream region has diminished significantly. The percentage of total basin area under different LU categories in the UGB for different time periods is provided in Table 3.
LU analysis of UGB for years 1973, 1980, 2000 and 2011.
Change in the ensemble mean of rainfall (top panel),
It should be noted that for the present study, detailed snow cover mapping
is not performed. Thus, the percentage area observed under the snow category in
Table 3 should not be considered as a trend in the snow cover of the region.
The urban category is observed to occupy far less area in the basin (
Observed rainfall obtained from IMD and projections of rainfall (
T2 in case of RCP4.5 emission scenario is observed to exhibit maximum change for all the three regions along with high uncertainties. High confidence level associated with T2 imply probable impacts in hydrologic response associated with this time slice. The RCP8.5 emission scenario, for most of the time slices, exhibits moderately significant change which may result in less probable impacts.
Upon assessing the monthly variability in
On analyzing the trend in observed and projected annual mean
Upon evaluating the emission-scenario-based uncertainty, it is found that
there is no significant difference between the two scenarios RCP4.5 and
RCP8.5,
which indicates that the scenario-based uncertainty will be minimum.
Impacts of changes in
To evaluate the effects of LU and climate change on the hydrology of the study area, three scenarios are considered. The first two scenarios are based on the single factor approach (Li et al., 2009); i.e., one driving factor is changed at an instant keeping the other constant. In the first scenario, climate is considered invariant while LU is varied with time, whereas in the second scenario, LU is considered invariant while climate is varied with time. These two scenarios are constructed to understand how streamflow would respond if only one of the driving forces is changed with time thereby assisting in quantifying the influence of individual factors on streamflow. In reality, both LU and climate change simultaneously with time and the hydrologic response is generated based on their integrated effect which is addressed by the third scenario. Finally, from the integrated response, contributions of LU and climate on the streamflow variability are segregated using results from the other two scenarios. In-depth analysis in the first two scenarios is carried out due to a lack of detailed studies that examine the effects of LU and climate change on streamflow in the UGB.
In order to investigate hydrological impacts of LU change, simulations are carried out keeping climate fixed at 1971, while LU is changed progressively from 1971 to 2011. LU in any region changes gradually over a period of time, and therefore starting and ending years may satisfactorily represent the change that has occurred in each LU class. Considering this, LU of the intermittent years can be obtained using rate of change in each LU class between the starting and ending years. It is to be noted that to obtain LU information for 1971 and 1972, rate of change between 1973 and 1980 is considered. LU obtained for each year is then used to drive the VIC model to obtain simulations under LU effect with invariant climate. Although simulations are carried out continuously from 1971 to 2011, for the sake of brevity, results corresponding to the starting year (1971) and the ending year (2011) for all the three regions are presented in Fig. 9.
Simulation results for 1971 and 2011 for
It can be observed in Fig. 9 that from 1971 to 2011 there is an increase in the magnitude of peak discharge for upstream and midstream regions. This observation is consistent with other studies reported in literature which state that LU change has a pronounced effect on peak flows due to alterations in the infiltration capacity of the surface (Fohrer et al., 2001; Naef et al., 2002; Tollan, 2002; McIntyre et al., 2014). No change in the discharge regime of the downstream region is noticed. LU and topography of the region is observed to have a conspicuous effect on the hydrologic response from the basin which is reflected in the hydrograph patterns for the three regions. The rising limb of the upstream region (Fig. 9a) begins during April while for midstream and downstream (Fig. 9b and c, respectively) it occurs during May–June. The early occurrence of a rising limb in the upstream region can be attributed to the snowmelt-runoff contribution to the streamflow. However, for midstream and downstream regions, a rising limb begins with the onset of monsoon. The recession limb of hydrograph for upstream region falls quickly owing to the steep slope of the region. For the midstream, a sharp drop is observed up to a certain level during October, indicating the termination of direct runoff contribution to streamflow. Following this, the contribution is predominantly through baseflow which in this case is observed to be higher than the baseflow before the monsoon months. The higher baseflow during post-monsoon period could be attributed to slow release of water stored by forests (dense and scrub) in the region aided by low elevation of the terrain in the region. Downstream region, though entirely a flat terrain, is dominated by cropland and urban areas that lack the capacity of holding the water, therefore limiting the contribution of baseflow to streamflow which leads to the observed sharp decline in the recession limb. Furthermore, long-term impacts of LU change are more evident in annual streamflow that is observed to increase by 12, 17 and 1 % from 1971 to 2011 for upstream, midstream and downstream regions, respectively.
Sensitivity of the region to different LU categories is assessed in separate
simulations. In this case, simulations considering each LU class are
performed and change in streamflow under each category is quantified. To
quantify the magnitude of change in streamflow caused by change in LU, the ratio
between streamflow and LU is computed. The ratio is referred to as the runoff / LU
ratio (RL) in the present study. The RL indicates the effect of 1 % change in
any LU category on streamflow and aids in identifying the significance of a
particular LU class in determining the hydrologic response. Based on the
ratios obtained, streamflow response (to a particular LU category) is
classified under three categories: (i) highly sensitive if RL is
It can be observed from Table 4 that in the upstream region, RL is maximum for the urban area implying that the hydrologic response in this region is highly sensitive to the changes in urban area. It can be inferred that 1 % change in the urban area results in 4 % increase in the streamflow from the upstream region. The upstream region has a significant portion of area under dense forest that has shown a minor increase in the last decade (2000 to 2011) (Table 3). The simulated streamflow is observed to be moderately sensitive to this increase, though the observed impact is in the opposite direction; i.e., an increase in forest results in a decrease in streamflow. Furthermore, streamflow simulated from the upstream region is moderately sensitive to croplands as well. The midstream region has cropland as the dominant LU type covering 53 % of the area during 1971 and 81 % of the area in 2011; streamflow is observed to be moderately sensitive to it. It is also observed that streamflow is moderately sensitive to urban area in this region. Though the downstream region is predominantly cultivated land (approximately 85 % of the area), hydrologic response is observed to be moderately sensitive to changes in the urban area. High sensitivity of streamflow from the regions to urban area can be attributed to the fact that increase in urban sprawl could reduce the infiltration resulting in the generation of higher surface runoff. In addition to this, it can be observed that hydrologic response to a change in forest area in midstream and downstream regions has a positive sign unlike in the upstream region, where the response has a negative sign. This is due to the fact that midstream and downstream regions are dominated by scrub forest, area under which has decreased over the time period, thereby increasing the streamflow. Thus all three regions of the UGB are observed to be moderately sensitive to a change in cropland area while moderately to highly sensitive to a change in urban area.
Runoff / LU ratio for different LU categories for upstream, midstream and downstream regions.
Change in ensemble mean of
Runoff ratio across time slices for upstream, midstream and downstream regions (terms in parentheses indicate the percent change from the baseline values).
Contribution of climate and LU to the streamflow for different time periods.
Streamflow observed at Bhimgodha (outlet for upstream region) and Ankinghat
(outlet for the midstream region) stations is examined for the presence of
a trend using the Mann–Kendall test. It is noticed that the observed streamflow
for upstream (1987–2005) and midstream (1977–2005) regions do not show any
trend. However, in order to investigate the individual impact of changing
climate on hydrology, simulations are carried out keeping LU fixed for 1971
and altering climate continuously for the baseline period (1971–2005) and
future emission scenarios (2010–2100). The simulation results obtained are
referred to as
From the Fig. 10, it can be observed that change in
Assessment of the monthly variations in the
The RR is a simple index that reflects the relationship between
In a real-world situation, change in LU and climate occurs simultaneously
and the impact of both these factors is reflected in the streamflow. To
carry out analysis pertaining to this scenario, one needs concurrent
information on LU and climate. Under this notion, VIC model is driven for
1971–2005 (baseline period) across the three regions in the UGB. It is to be
noted that the process of obtaining projections of future LU conditions in
the basin does not come under the purview of the present work. Therefore,
integrated impact of LU and climate change on future streamflow could not be
assessed. The results obtained from this analysis can be interpreted as the
streamflow simulations under simultaneous change in LU and climate
conditions (hereafter referred to as
In order to segregate the impacts of LU and climate, the proposed approach
primarily requires results of
In the present case study, simulations of
Contribution of LU and climate to streamflow during the T1 (2010–2020) time slice under RCP4.5 and RCP8.5 emission scenarios.
Results from Table 6 suggest that climate is the dominant contributor to
streamflow across all the regions. The contribution of LU, on the other hand, is
observed to be minimal. Further insight to the influence of LU to streamflow
is obtained from the inferences drawn from Sect. 3.3.1. It is observed
from the analysis in Sect. 3.3.1 that streamflow is highly sensitive to
changes in urban land in upstream and downstream regions, while it is
moderately sensitive to urban and cropland areas in the midstream region. The
spatial extent of urban area is observed to be much less in upstream and
downstream regions (less than 10 %), which could have resulted in
a negligible contribution of LU to streamflow. For the midstream region,
despite
Contribution of LU and climate on the streamflow response is isolated at a monthly scale as well. It is observed that climate is a major contributor to the streamflow across all three regions at a monthly scale as well (see the attached Supplement).
In the present study, the application of proposed methodology of isolating
the hydrologic impacts of LU and climate is limited only to the baseline
period due to unavailability of future LU information. However, this
approach can be applied to the future time periods as well upon obtaining
future LU projections along with climate simulations. This is illustrated by
conducting the analysis on T1 (2010–2020) wherein
From Table 7, it can be observed that the contribution of LU to streamflow from the upstream region has increased (compared to P4). This could be attributed to an increase in area under urban land by 65 % in T1 from P4 in the upstream region. No significant increase is observed in cropland and urban land areas in T1 from P4 for midstream and downstream regions, respectively (2 % increase in cropland in the midstream region and 20 % increase in urban area in the downstream region), which could have resulted in an unvarying contribution of LU to streamflow from P4 (Table 6) to T1 (Table 7) in these regions.
From the analysis, it can be concluded that the proposed approach can be applied over a catchment with a well-calibrated and validated hydrologic model. Future work involves generating LU projections for future time periods, which can be corroborated with climate projections described in Sect. 3.3.2, to isolate the impacts of LU and climate on future streamflow simulations. Although there is the presence of a snow covered region in the basin, segregating the contribution of snowmelt runoff from the total streamflow is not feasible at this stage due to a lack of observed data. This limits the assessment of impact of temperature changes on snowmelt and its consequences on the streamflow.
In the present paper a hydrologic modeling-based methodology is presented to isolate the impacts of LU and climate on streamflow in a river basin. To achieve this, three objectives are considered (i) assessing the sensitivity of the streamflow to the changes in LU, (ii) examining the impact of change in climate on the streamflow and (iii) integrated impact of LU and climate change on the streamflow of the UGB. These three objectives are translated to three scenarios and are used to segregate the influence of LU and climate change on the streamflow. Not many studies conducted earlier have considered the combined effect of LU and climate on the hydrology of the basin. The VIC hydrologic model is used to understand the impact of LU and climate change on the streamflow. The VIC model, owing to its comprehensive ability to simulate hydrological processes, has been used widely to perform impact assessment studies. However, being a physically based distributed model, there are concerns associated with the model structure and the number of calibration parameters. Furthermore, due to spatiotemporal variability in the input variables, a parameter set for the initial or reference time period may not be suitable for future periods (Viney et al., 2009). In the present study, these concerns are partially addressed by calibrating and validating the VIC model over upstream, midstream and downstream regions of the UGB.
LU change analysis of the study region indicated an increase in the areas of
crop and urban land categories to which streamflow is observed to be
moderately to highly sensitive. From the climate change analysis, it is
observed that rainfall may decrease during the monsoon months and increase
during the winter months which may result in a shift in seasonal rainfall
pattern. Annual means of
The integrated effect of LU and climate change on streamflow is observed to
be more prominent in the study area. From the analysis of isolating the
individual impacts of LU and climate from their integrative effects on
streamflow, it is observed that climate contributes more to the simulated
streamflow (
The proposed approach is generic and applicable to any river basin to isolate the relative impacts of LU and climate change on streamflow. However, the approach is based on the assumption of linear response of LU and climate to the streamflow. The case study analysis indicates that the change in climate may become a major concern in the UGB for water resources management.
The authors would like to thank Thomas Kjeldsen (manuscript handling editor), Ge Sun, Young-Oh Kim and the anonymous referees for providing very useful comments on the manuscript. The work is carried out as part of the MoES-NERC Changing Water Cycle (South Asia) project: hydro-meteorological feedbacks and changes in water storage and fluxes in Northern India (grant no. MoES/NERC/16/02/10 PC-II). Authors acknowledge the support of the Indian Meteorological Department (IMD) and the Indian Institute of Tropical Meteorology (IITM) for providing the data. Edited by: T. Kjeldsen