You are on page 1of 62

Published by:

Global Change Impact Studies Centre (GCISC) National Centre for Physics (NCP) Complex Quaid-i-Azam University Campus P.O. Box 3022, Islamabad-44000 Pakistan

ISBN: 978-969-9395-10-9

@GCISC

Copyright. This Report, or any part of it, may not be used for resale or any other commercial or gainful purpose without prior permission of Global Change Impact Studies Centre, Islamabad, Pakistan. For educational or non-profit use, however, any part of the Report may be reproduced with appropriate acknowledgement.

Published in: June 2009

This Report may be cited as follows: Saeed, F., S. Jehangir, M. Noaman-ul-Haq, W. Shafeeq, M. Z. Rashmi, G. Ali and A.M. Khan, (2009), Application of UBC and DHSVM Models for Selected Catchments of Indus Basin Pakistan, GCISC-RR-l1, Global Change Impact Studies Centre (GCISC), Islamabad, Pakistan.

2.1.1 Treatment of Meteorological Variables 2.1.2 Process Simulation 2.2 The Siran Watershed 2.2.1 Description 2.2.2 Meteorology 2.2.3 Preprocessing of Input Data 2.2.4 Meteorological Data 2.3 Calibration of DHSVM 2.3.1 Soil Water Content 2.3.2 Snow Water Equivalent 2.3.3 Evapotranspiration 2.3.4 Flow Simulation 2.3.5 Results and Discussion 2.4 Issues of Hydro-Met Data 2.5 Conclusion

25 25 28 28 28

28
31 32 32 32 33 33 35 37 37

Evaluating Spatial Resolution for Distributed Hydrological Modelling in the Indus River Basin
3 . Introduction 38 3.1 Criteria of Optimal Spatial Resolution 3.1.1 Field of Application of Spatial Databases 3.1.2 Extent of DHSVM Application 3.1.3 Commercial Availability of Spatial Digital Data 3.1.4 Computing Requirements 3.1.5 Resolution of Remote Sensing Data 3.1.6 .Restrictibility of Spatial Data 3.2 Development of Spatial Database 3.3 DHSVM Application in the Siran Watershed 3.3.1 Meteorological Data 3.3.2 Parameterization 3.3.3 Results of DHSVM Simulation of 80m 3.4 Comparative Efficiency of 1000m Resolution 3.5 Discussions 3.5.1 Issues of spatial data 3.5.2 Issues of Hydro-met data 3.6 Conclusions 49

38

39 39 39 39 39 39 40 40 43 43 43 43 47 48 48 48

References

50

FOREWORD

Global Change Impact Studies Centre (GCISC) was established in 2002 as a dedicated research centre for climate change and other global change related studies, at the initiative of Dr. Ishfaq Ahmad, NJ, HI, SI, the then Special Advisor to Chief Executive of Pakistan. The Centre has since been engaged in research on past and projected climate change in different sub regions of Pakistan; corresponding impacts on the country's key sectors; in particular Water and Agriculture; and adaptation measures to counter the negative impacts. The work described in this report was carried out at GCISC and was supported in part by APN (Asia Pacific Network for Global Change Research), Kobe, Japan, through its CAPaBLE Programme under a 3-year capacity enhancement cum research Project titled "Enhancement of national capabilities in the application of simulation models for assessment of climate change and its impacts on water resources, and food and agricultural production", awarded to GCISC in 2003 in collaboration with Pakistan Meteorological Department (PMD). It is hoped that the report will provide useful information to national planners and policymakers as well as to academic and research organizations in the country on issues related to impacts of climate change on Pakistan. The keen interest and support by Dr. Ishfaq Ahmad, Advisor (S & T) to the Planning Commission, and useful technical advice by Dr. Amir Muhammed, Rector, National University for Computer and Emerging Sciences and Member, Scientific Planning Group, APN, throughout the course of this work are gratefully acknowledged.

Dr. Arshad M. Khan Executive Director, GCISC

PREFACE

Climate Change is one of the most serious problems the world is facing today and has become a growing concern for nations across the globe particularly the developing countries. Human activities have now reached a level where Green House Gases (GHGs) of anthropogenic origin are significantly affecting the global environment. The research conducted by the Intergovernmental Panel on Climate Change (IPCC) over the last twenty years has established that the increasing concentration of GHGs in the atmosphere has caused an increase in the average global temperature by about 0.6 C over the last century. Much larger changes are expected in the 21st century, with the average global temperature rising in the order of 1.1oC - 6.4 C. In order to assess the impact of expected climatic change on river inflows, one needs to make use of appropriate watershed simulation models to work out the flows on the basis of climatic parameters and. basin characteristics. This report describes the suitability of UBC model (developed at the University of British Columbia by Quick and Pipes in 1977) and DHSVM (Distributed Hydrology Soil and Vegetation Model, developed by Wigmosta et aI. in 1994 at the University of Washington) for different rivers of the Upper Indus Basin (UIE). The UBC model is calibrated and validated for five sub-basins (Siran, Hunza, Astore, Kabul and Jhelum) having different hydrological regimes and DHSVM for only Siran River Basin of the UIB, for generating future flows under different climate change scenarios. The provision of hydro-meteorological data by Water and Power Development Authority (WAPDA) for the calibration/validation of watershed models is gratefully acknowledged.

11

LIST OF TABLES

Table 1.1 Table 1.2 Table 1.3

Meteorological and discharge gauging stations used for different basin Geophysical characteristics of Hunza Watershed per elevation zone Calibration and validation statistics of the UBC Watershed model for Hunza Watershed

11 13 17

Table 1.4

Calibration and validation statistics of the VBC Watershed model for Astor Watershed

20

Table 1.5

Calibration and validation statistics of the UBC Watershed model for Siran Watershed

23

Table 1.6

Calibration and validation statistics of the UBC Watershed model for Jhelum Basin

26

Table 1.7

Calibration and validation statistics of the UBC Watershed model for Kabul Basin

29

Table 2.1 Table 2.2 Table 2.3 Table 3.1 Table 3.2

Percentage area of different land use types for the years 1979 & 1989 Overall and seasonal accuracies of DHSVM outputs for 1979 and 1980 Accumulated Discharge of DHSVM simulations of two years Overall and seasonal accuracies of DHSVM outputs for 1979 and 1980 Comparative results of 80 ill and 1000 ill resolutions

38 43 44 59 61

iii

LIST OF FIGURES

Figure 1.1 Figure 1.2

Flow diagram of UBC Watershed model Northern Pakistan map of Upper Indus Basin and the location of study watershed

2 9

Figure 1.3

Characterization of elevation zones of Hunza watershed through GIS

13

Figure 1.4

Comparison of observed and simulated flow along with various components of runoff - Hunza watershed

19

Figure 1.5

Comparison of observed and simulated flow along with various components of runoff - Astor watershed

21

Figure 1.6

Comparison of observed and simulated flow along with various components of runoff - Siran watershed

24

Figure 1.7

Comparison of observed and simulated flow along with various components of runoff - Jhelum watershed

27

Figure 1.8

Comparison of observed and simulated flow along with various components of runoff - Kabul watershed

29

Figure 2.1 Figure 2.2

Model Representation of Watershed Schematic representation of DHSVM inputs, preprocessing requirements, outputs and interaction with GIS software.

31 34

Figure 2.3 Figure 2.4

Location of sir an Watershed in UIB Digital Elevation Model of Siran Watershed and location of meteorological stations

35 36

Figure 2.5 Figure 2.6

Land use grids for siran watershed Conversion of land use types dense cover forests & medium cover forests from 1979 into land use types of 1989

37 38

Figure 2.7

Scenarios of land use grids for siran watershed

39

iv

Figure 2.8

Daily fluctuation in soil water contents of siran watershed for .1979 and 1980

40

Figure 2.9 Figure 2.10 Figure 2.11 Figure 2.12 Figure 2.13 Figure 3.1 Figure 3.2 Figure 3.3 Figure 3.4 Figure 3.5 Figure 3.6 Figure 3.7 Figure 3.8 Figure 3.9 Figure 3.10

Snow water equivalent of Siran watershed for 1979 and 1980 Daily actual evapo-transpiration from Siran watershed Simulation and observed hydrograph Evaluation Map outputs of DHSVM Accumulation Evapotranspiration of DHSVM Digital Elevation Model of the Indus River Basin Extent of Indus Basin Database Land cover of the upper Indus basin Location of siran watershed in DIB Digital Elevation Model of the Siran Watershed Daily fluctuation in soil water content of Siran watershed Snow water equivalent of Sir an watershed for 1979 and 1980 Daily actual evapo-transpiration from Siran watershed Simulated and observed hydrographs Comparative hydro graphs simulated

41 41 42 45 46 51 51 52 53 54 56 57 57 58 60

APPLICATION OF UNIVERSITY OF BRITISH COLUMBIA WATERSHED MODEL (UBC WM) FOR BASINS OF INTEREST
1.1 Background
The economic life of Pakistan depends to a large extent on its agriculture, which in turn is dependent on irrigation through a vast network of barrages, diversions and channels from the River Indus and its tributaries. Almost 60% of total inflows of Pakistan are attributed to River Indus. Most of the flow abstracted for irrigation from the River Indus originates in the Hindu-Kush, Karakoram and Himalaya (HKH) mountains and is fed by a combination of melt water from seasonal and permanent snow fields and glaciers, and direct runoff from rainfall both during the winter and the Monsoon season from July to September. Almost 80% of the flows of IRS are attributed to snow and glacial melt. Glaciers of HKH are reported to be specifically vulnerable to the drastic impacts of climate change as a result; hydrological equilibrium is being threatened by these changes (Rees, 2004). These scenarios call for revisiting the medium-term and long-term plans to meet the national water needs for agriculture and other sectors through the use of state of the art Climate and Hydrological models. In previous studies, which concentrated primarily on the regression between climatic variables and stream flow, the Upper Indus Basin (UIB) was divided into three contrasting hydrologic regimes (Archer, 2003). In these regimes, the summer discharges are governed by the melt of glaciers and permanent snow (thermal control in the current summer), melt of seasonal snow (control by preceding winter and spring precipitation) and winter or monsoon rainfall (precipitation control in current season). Considering the diverse hydrologic nature of UIB, selection of a suitable watershed model is of utmost importance. This indeed, is a tough task, largely, due to the recognized challenges in modeling the hydrology of snow and glacier fed regions, complex topography and deficiencies in meteorological network. Distributed hydrologic models generally fit into the categories of either large-scale models and tend to be applied at coarser resolutions, for example the Variable Infiltration Capacity (VIC) model (Liang et al., 1994), or models developed for application at finer resolutions, for example Distributed Hydrology Soil and Vegetation Model (DHSVM) (Wigmosta et al., 1994). The typical resolution for VIC model used in literature is about 10-15 km, a resolution that would lump a number of topographic and geographic features in UIB, thereby making it unsuitable for application in this region with rapidly changing topography. On the other hand DHSVM was designed for application at 30m to 150m resolution and has rather large spatial data requirements. The available data for this study do not justify use of models like DHSVM either. A more suitable option is the use of a semi-distributed model which can be operated with freely available Geographical Information System (GIS) inputs. In this study, University of British Colombia's (UBC) Watershed Model is applied to three mountainous watersheds of the UJB, namely Hunza, Astore and Siran which represent three different hydrologic regimes using loose coupling technique with GIS data (AI-Abed et aI., 2005). The work is mainly concentrated on the calibration and validation of UBC Watershed Model in these catchments using limited meteorological data of automatic weather stations.

1.2 Salient features of UBC Watershed Model


The model selected for the study is of University of British Columbia, named University of British Columbia Watershed Model. This model was developed by Quick and Pipes in 1977, primarily for mountainous watersheds and calculates the total contribution from glacier melt, snowmelt and rainfall runoff. The structure of the model is based on hydrological behavior as a function of elevation in the watershed. The model takes maximum and minimum temperature, precipitation as an input. Daily watershed stream flow is the main output of the model along with information on snow pack water equivalent, snow covered area, current soil moisture status and groundwater storage of the watershed. The watershed is divided into area-elevation bands; therefore the above mentioned information can also be obtained for each band separately. Each e1evation band has a separate. variable description of the watershed, such as the forested fraction, impermeable fraction, glaciated areas and aspects. The complete structure of model is discussed by Quick and Pipes (1977), however, some important factors which will be used extensively in the present research work are as follows.

1.2.1 Temperature Lapse Rate


Lapse rates are known to be quite variable, ranging from high values of dry adiabatic lapse rate (10 C/ 1000 m) to low values representing inversion conditions. A complete and detailed representation of lapse rate variability is not possible within the model, but the main features of lapse rate can be represented as a function of daily temperature range. The following major features of lapse rate variation are recognized by the temperature lapse rate algorithm:

1.

During continuous rainstorm conditions the lapse rate will approximate the saturated adiabatic lapse rate. Under these conditions the daily temperature range will tend to collapse to zero. Under clear sky and dry weather conditions, the lapse rate during the warm part of the day will tend to be dry adiabatic rate. During the night, under the clear sky conditions, radiation cooling will cause the temperature to fall to the dew point temperature, and this is particularly true for a moist air mass. As a result, night-time lapse rate under clear skies will tend to be quite low, and at times even zero lapse rates will occur. Based on these considerations, two lapse rates are specified in the model, one for the maximum temperature and one for the minimum temperature. The lapse rate is calculated for each day using the daily temperature range (diurnal range) as an index. The general form of the equations used in the program is given below. Maximum Temperature Lapse Rate --- TXLAPS = TZLAPS + (TLXM - TZLAPS) * TD/AOTERM
(1)

2.

Minimum Temperature Lapse Rate --- TNLAPS = TZLAPS - (TZLAPS - TLMN) * TD/AOTERM (2)

Where TD = daily temperature range (TX-TN) And TZLAPS = TZ - (PP/PPM) * (TZ - TZP) For the above, PP = daily precipitation

TZ = 6.4C/km, that is, a reference lapse rate for rain-free conditions TZP = 6.4C/km, that is, a reference lapse rate when PP > PPM PPM = 5 mrn/day The calibration parameter AOTERM equals the maximum temperature range under open sky conditions (selected from the data set (TX-TN).

1.2.2 Form of Precipitation


The model must distinguish between precipitation in the form of snow and precipitation falling as rain and this distinction must be made for each elevation band. Snow is stored until melted, whereas rain is immediately processed by the soil moisture model. The form of precipitation is controlled by three logical statements and the temperature, T, used in these statements is normally the mean daily temperature in each band but it can be specified to be the maximum or the minimum daily temperature in each band. 1fT < OC all precipitation is SNOW 1fT >AOFORM all precipitation is RAIN.

AOFORM = Temperature above which all precipitation is rain AOFORM is specified in the parameter list. If it is not specified or set equal to or below 0C, it will take a default value of 2C.

1.2.3 Precipitation Adjustment Factor


POPAD] can be used to increase the amount of precipitation in a band as an alternative to adjusting the amount of precipitation at the station. Sometimes, it is useful to assist the model in building glaciers or modeling avalanches. However, this parameter is rarely used for routine calibration.' A POPAD] value of 0.5 increases the precipitation by 50%, and a value of 1.0 doubles the amount of precipitation.

1.2.4

Precipitation Elevation Gradients

The precipitation in any elevation band is calculated from the precipitation in the band immediately above or below using the equation PI, J, L+l = PI, J, L*(1 + a).6. elev/l00 Where PI, J, L is the precipitation from meteorological station I for day J and elevation band L. A = elev = precipitation gradient (%) difference in elevation between the AES stations and the band
(3)

The 1 +a multiplier produce a logarithmic increase in precipitation with elevation. The enhancement factor 'a' is separately defined above and below a certain elevation which may be specified (EOLMID). Three precipitation gradients are possible: POGRADL, for bands below EOLMID; POGRADM, for bands between EOLMID and EOLHI; and POGRADU, for bands above EOLHL

1.2.5

Snowmelt

The Watershed Model accumulates precipitation falling as snow and then depletes these snow packs according to the calculated melt rate. This snow accumulation and depletion is carried out separately in each area elevation band. The final depletion of snow is a gradual recession of snowline from the bottom to the top of the band. The model redistributes the snow pack in such a way that there is more snow in the higher part of band and less in the lower part of band. The snow pack is then melted in such a way that less and less band area is covered by snow. The UBC Watershed Model uses, an energy budget approach to calculate snowmelt. The energy exchange at the surface of a snow pack is made up of four major components: 1. The shortwave radiation exchange, which consists of incoming solar radiation, and the reflected outgoing shortwave radiation. This shortwave component depends on the time of year, the site exposure, cloud cover and snow albedo. The longwave radiation exchange depends on black body radiation from the snow itself and from clouds and tree cover, and gray body radiation from the overlying air mass. Under clear sky, the net longwave is outgoing, or negative, unless air temperature exceeds about 20C. Under cloudy conditions' and also under tree cover, net longwave can be positive at temperature above freezing.

2.

3.

Convective heat transfer is produced by turbulent heat exchange between the air mass immediately above the snow pack This heat transfer is dependent on both wind and air temperature and particularly on the stability of the air mass above the snowpack, A warm air mass above the cold snow surface tends to be stable, resisting any downward transport of heat to the snow pack, unless there is enough wind to produce turbulent mixing. This turbulent heat transfer is governed by the Richardson number, RI, which is a measure of stability. The bulk Richardson number, defined below, is essentially the air temperature divided by the wind speed squared. As the air temperature increases, and if the wind is only moderate, the stability can increase to the point where very little convective heat transfer can occur. Convective heat transfer is therefore self limiting and becomes quite small at higher air temperature, unless there is very strong wind. Advective heat transfer, often termed condensation melt, is caused by the transport of moisture to and from the snow pack. Whether condensation occurs, cooling the snowpack depends on the relative vapor pressures of the air and snow surface. Wind is once again an important factor and so is stability, as was discussed for convective transport. Advective heat transport can therefore produce snowmelt if the dew point temperature is above freezing, but, like the convective heat transport, becomes limited at higher temperatures by the stability of the warm air mass. A simple set of equations which expresses the various snowmelt components in terms of millimeters of snowpack per day is given and discussed by Quick (1995).

4.

1.2.6

Model Efficiencies

The model calibration is measured by the Nash-Sutcliffe efficiency (Eff) and Coefficient of determination (R2). The Eff relates how well the estimated hydrographs compare in shape and volume to the observed hydrographs, and is calculated as follows:

(4)

Where

The coefficient of determination, R2 is a factor that relates how well the shape of the estimated hydrographs corresponds to the shapes of the observed hydrographs, and it is independent of volume. However, timing does affect the result of this statistics. The coefficient of determination is calculated as follows:

(5)

3.

Where:

1.3

Application of UBC Watershed Model to rivers of the UIB having different Hydrological Regimes

1.3.1 Study Area For this study three watersheds are selected from the DIB on the basis of their diversified hydrological environment for exploratory analysis. These are the Hunza River the flow of which observed at Daniyor Bridge, the Astore River at Doyian and Siran River at Phulra. Hunza watershed is located in high Karakoram mountain ranges with an elevation range of about 1500 masl to 7500 masl. The catchments area of Hunza river up to Daniyor Bridge is 13,157 km2 (Figure 1.2). Hunza is a highly glacier fed river with about 28% of the catchments area covered with glaciers and contains high mountain peaks like Batura (7786m) and Rakaposhi (7788m).

Figure 1.1 Northern Pakistan map of Upper Indus Basin and the location of study watersheds This watershed is arid and receives less than 150 mm of precipitation annually. Mean annual runoff of this watershed is 364 m3/s and most of the precipitation, which is normally generated due to

westerly disturbance, is in the fonn of snow and in winter season up to 90% of the watershed is snow covered. The largest part of the total annual runoff is generated in summer months by the melting of glaciers and the seasonal snow cover. The mean annual stream flow accounts for more than 13% of the inflow to the downstream Tarbela reservoir (Figure 1.2). Astore watershed is located in high Himalayan mountain ranges with an elevation range of about 1200 masl to 7500 masl, The catchment's area of Astore river up to Doyian is 4040 km2 (Figure 1.2). Astore River is fed by both glacier and snowmelt with about 6% of the catchments area covered with glaciers. The average annual precipitation for the entire basin is about 700 mm, 90% of which is in the form of snow. Most of the precipitation occurring in this watershed is generated from westerly depressions but occasionally monsoon currents also reach in this catchment. Mean annual discharge of this watershed is 124 m3/s, which account for 5% of the inflow to Tarbela reservoir. The largest part of the total annual runoff is generated in summer months mainly by the melting of seasonal snow cover and a small proportion due to glacial melt. The Siran watershed lies in the monsoon-fed part of the Indus Basin and measures 1060 km2 up to Phulra gauge (Figure1.2). The elevation of the Siran watershed ranges between 834 masl and 2199 masl and terrain comprises of two valleys, surrounded by moderate to steeply sloping mountains. Siran is glacier free watershed with uppermost reaches receive snow from December to February. The mean annual precipitation of the watershed is 1243 mm and about 70% is received during July, to September. This watershed receives both winter rainfall resulting from the westerly depressions and the monsoon from either the south westerly or the south easterly currents originating over Arabian Sea and Bay of Bengal respectively. The mean annual discharge of this watershed is about 16 m3/s at phulara,

1.3.2 Database
The UBC Watershed model needs daily values of maximum and minimum temperature and precipitation as an input from the observed station values. The main shortcoming of these manual meteorological stations is their lower elevations. In the watersheds of relatively higher elevation peaks, these valley based meteorological stations do not accurately represent the meteorological processes occurring at higher elevations. In order to cope with this problem, Water and Power Development Authority (W APDA), installed a network of 20 automatic weather stations, called as Data Collection Platforms (DCPs) in 1990's (Archer, 2003). These DCPs collect meteorological data within an elevation band of 2500-5500 masl in the mountain ranges of the Hindukush, Karakoram and Himalayas. DCPs are used for modeling Hunza and Astore watersheds because of their higher elevation ranges, whereas manual valley based stations are used for Siran Watershed (Table 1.1). Many watershed scale data were derived using GIS and utilized in the modeling procedure. Since the acquisition and extraction of the physical parameters pertaining to description of the watershed using topographic, geological, soil and vegetation maps is a cumbersome process, so it is decided to use freely available GIS data sets for this purpose. For the three studied watersheds, the spatial data or the physio-graphic features of the watersheds such as Elevation zones, land cover, Aspect, have been derived from the Digital Elevation Model.

Table 1.1 Meteorological and discharge gauging stations used for different basin

Slope grids developed by USGS/DAAC at l-km spatial resolution. Special programs are encoded in C++' compiler, which read these high-resolution grids and calculate the physical parameters pertaining to each watershed. These programs require the manual information about the elevation ranges for the creation of elevation bands after reading the information from the DEM for a particular watershed. However, elevation ranges should be given in such a way that each elevation zone should get a significant area so as to keep the soul of semi distribution watershed alive. Once the elevation zones are fanned, the programs then read the information from spatial grids of land cover, aspect and slope for obtaining physical parameters such as vegetation cover, tree density, orientation etc. The output for Hunza watershed is given in Figure 1.3 and Table 1.2.

1.3.3 Previous Application of UBC Watershed Model


The UBC Watershed Model was designed primarily for the estimation of stream flow from mountainous watersheds where stream flow consists of snowmelt, rain and glacier outflow. With subsequent updates, it has been used earlier for diverse climatic regions, ranging from coastal to inland mountain regions of British Columbia including the Rocky Mountains and the sub arctic region of Canada (Assaf and Quick, 1991; Hudson and Quick, 1991; Quick et al., 1998; Micovic and Quick, 1999; Druce, 2001; Morrison et al., 2002).

4.
Figure 1.2 Characterization of elevation zones of Hunza watershed through

Table 1.2 Geophysical characteristics of Hunza Watershed per elevation zone

The model has also been successful in the Himalayas in India (Singh and Kumar, 1997). The UBC Watershed model calculates watershed outflow using point measurements of precipitation and temperature data combined with physical watershed characteristics as input. Since the hydrologic behavior of the mountainous watershed is a function of elevation, the model uses the area-elevation zones concept, which makes it a semi-distributed modeL This concept accounts for orographic gradients of precipitation and temperature, which are assumed to behave similarly for each storm event so that, based on temperature, the model estimates whether precipitation falls as rain or snow and estimates snowpack accumulation as a function of elevation. A simplified energy budget approach, based only on data of maximum and minimum temperature is used to estimate snowmelt (Quick, 1995). Furthermore, the geophysical characteristics of each elevation zone in a watershed, such as impermeable area, forested areas, vegetation density, open areas, aspect and glaciated areas can be estimated from maps or remotely sensed data, on the assumption of homogeneity of the characteristics within each elevation zone . For each zone, the runoff from rainfall, snowmelt and glacier melt is distributed into four runoff components by a soil moisture control mechanism, which represents the nonlinear behavior of the watershed. This mechanism apportions the water from rainfall, snowmelt and glacier melt into fast or surface runoff, medium or interflow runoff, slow or upper zone groundwater runoff and the very slow or deep zone groundwater runoff. For each elevation band, the water that reaches the soil surface, after interception and sublimation, is divided into fast surface runoff and infiltrated water. The water infiltrated satisfies the soil moisture deficit and the evapotranspiration and then passes to groundwater or runs off as interflow. The ground water is further divided into upper and lower groundwater zones. When the soil moisture accounting sub-routine has been applied to all the elevation zones, each runoff component is routed to the watershed outlet by the flow routing subroutine. The simplest flow diagram of the model is shown in figure 1.3.

10

Figure 1.3: Flow diagram of UBC Watershed model Source: http://www.cig.ensmp.fr/~iahs/hsj/220/hysL22_01_0153.pdf The flow routing uses the linear reservoir cascade technique, which simplifies, significantly, the model structure, conserves the water mass, and provides a simple and accurate water budget balance. The flow variables are: the snowmelt and rainfall fast runoff time constants, the snowmelt and rainfall interflow time constants, the glacier time constant, the upper groundwater time constant and the deep zone groundwater time constant. The summation of the runoff from all bands and for all runoff components gives the catchments runoff for the time step. Apart from the total runoff, the UBC model provides information on snow-covered area, snowpack water equivalent, potential and actual evapotranspiration, soil moisture, interception losses, groundwater storage, surface and subsurface runoff for each elevation band separately and for the whole watershed. The UBC watershed model has more than 90 parameters. These parameters are separated into three groups: the precipitation distribution parameters, the water allocation parameters, and the flow routing parameters. Application of the model to various climatic regions and experience have shown that values of only 17 general parameters and two precipitation representation factors for each meteorological station have to be optimised and adjusted during calibration, and the majority of the parameters take standard constant values. These parameters are optimized through a two-stage procedure. At the first stage, a sensittvity analysis of each parameter is performed to estimate the range of parameter values for which the simulation results are the most sensitive. At the second stage, a Monte Carlo simulation is performed for each parameter of each group by keeping all other parameters constant. The parameter values are sampled from the respective parameter range defined in the first stage (sensitivity analysis) of the procedure. The procedure starts with the optimization of the

11

precipitation distribution parameters and ends with the optimization of the flow routing parameters. The objective function of the above optimization procedure is defined as:

(6)

Where, Vsim and Vobs are the simulated and the observed flow volumes, respectively and Eff is the Nash-Sutcliffe efficiency (Nash and Sutcliffe, 1970) defined as:

(7)

Where, Qobs i is the observed flow on day i, Qsim, is the simulated flow on day i, Qobs is the average observed flow and n is the number of days for the simulation period. The range of Eff lies between 1 (perfect fit) and -00. Physically, Effis the ratio of the Mean Square Error (MSE) to the variance in the observed data, subtracted from unity.

1.3.4 Application of UBCWM to Selected Watersheds


The daily stream flow of the three studied watersheds .has been simulated using the UBC Watershed Model. The application of the model followed the split sample test (Donigian, 2002). According to this test, the model was calibrated to the observed flow for the three studied watersheds for three hydrological years, i.e. 1999-2000, 2000-2001, 2001-2002, and it was validated for four hydrological years, i.e. 1995-1996,1996-1997,1997-1998, and 1998-1999. The simulated and observed discharges were compared graphically and statistically. Various statistical indices have been used to evaluate the accuracy and performance of the two simulation methods, namely, the Nash-Sutcliffe model efficiency, Eff, (Eq. 4), the percent runoff volume error: %DV = Vsim - Vobsx 100 Vobs (8)

Where, Vobs is the observed runoff volume, and Vsim is the simulated runoff volume for the simulation period and the coefficient of determination: (9)

12

The model efficiency (Eff) has been widely used in hydrological simulation studies. It compares the scale and the shape of the simulated and the observed hydrographs and its optimal value is 1. The percent runoff volume (%DV) is a scale parameter and measures the percent error in volume under the observed and the simulated hydrographs for the period of simulation. Positive values of %DV indicate overestimation of the observed runoff volume, negative values of %DV indicate underestimation of the observed runoff volume, and %D V equal to zero indicate perfect agreement between simulated and observed runoff volumes. The coefficient of determination (R2) is a shape statistical parameter that measures the linear correlation between the observed and simulated flows with optimal value of 1. The model efficiency represents an improvement over the coefficient of determination for model evaluation purposes because it is sensitive to differences in the observed and modeled means and variances. Due to the squared differences, however, E is more sensitive to extreme values, as is R2.

RESUL TS AND DISCUSSIONS


The results of all the three simulated watersheds are discussed one by one hereafter.

Hunza Watershed
The results for Hunza watershed for both calibration and validation periods are given in Table 1 .3. For the whole calibration period of Hunza watershed, the Nash-Sutcliffe efficiency is 0.91, the coefficient of determination is 0.91 and runoff volume difference is-2.28%. In individual years the Nash-Sutcliffe efficiency ranges from 0.89 to 0.95, the coefficient of determination ranges from 0.89 to 0.97 and runoff volume ranges from -10.48% to 4.58%.

13

Table 1.3 Calibration and validation statistics of the UBC Watershed model for Hunza Watershed
Nash-Sutcliffe
Hydrologic years efficiency (Eff) Coefficient of determination diff. (%DV) Runoff volume Mean observed flow (m3/s) Mean simulated flow (m3/s)

(R )
Calibration period 1999-2000 2000-2001 2001-2002 0.89 0.90 0.95 0.89 0.91 0.97 1.53 -10.49 4.59 282.58 353.78 265.42 286.90 316.68 277.59

Whole period
(1999-2002) Validation period 1995-1996 1996-1997 1997-1998 1998-1999

0.91

0.91

-2.28

300.58

293.72

0.89 -0.49 0.85 0.87

0.89 0.93 0.86


0.88

2.53 70.07 -11.34 3.82

256.85 199.55 327.23 313.66 274.31

263.34 339.37 290.12 325.63

Whole period

(1995-1999)

0.73

0.78

11.04

304.59

For the whole validation period, the Nash-Sutcliffe efficiency and the coefficient of determination reduce to 0.73 and 0.78 respectively, whereas runoff volume difference is increased to 11.03%. Among the individual years, 1996-97 is an exceptional year which is poorly modeled. In this year the value of coefficient of determination is 0.93, indicating that the patterns of the flow has been captured very well but the volume is considerably overestimated for the whole year by 70.08%. Thorough examination of the meteorological data indicated that 1996-1997 hydrological year was quite dry for the northern and eastern part of the Upper Indus Basin. This could be also seen from the low mean annual flow of Hunza and Astore watersheds for this particular year (Tables 1.3 and 1.4). Furthermore, analysis of recorded temperature data indicated that daily temperatures were much higher than normal temperatures for that particular year and especially for the period of June to August. However, this abnormality may be due to a non-representative recording of the meteorological stations which in turn results in an erroneous larger simulated glacial and snowmelt outflow or it is real but the range of meteorological variation is outside the meteorological conditions of the calibration period. Also, due to its complex topography Hunza watershed is prone to the events like flooding due to landslides and glacial lake outburst (Archer, 2002). Personal communication with Director Snow and Ice Hydrology Project, W APDA also confirmed that there was an event of landslide in 1997 which blocked the Hunza River. Since this event did not create any havoc in the region, therefore it is not documented other than the local newspapers. In other individual years of validation the Nash-Sutcliffe efficiency ranges from 0.85 to 0.89, the coefficient of determination ranges from 0.86 to 0.89 and runoff volume ranges from -11.34% to 3.82%. The graphs between observed and simulated flows for the calibration year 2001-02 and the validation

14

year 1997-1998 are given in Figure 1.4. In winter season from October to March all the discharge is in the form of base flow. The flows during the months of May and June are a result of snowmelt whereas glacial melt accounts for flows during July to September. These results are inline with the findings of (Khan., 2001).

Figure 1.4: Comparison of observed and simulated flow along with various components of runoff - Hunza watershed for: a) calibration year (2001-2002) and b) validation year (1997-1998)

15

Astore Watershed
The results for Astore watershed for both calibration and validation periods are given in Table 1.4. For the whole calibration period of Astore watershed, the Nash-Sutcliffe efficiency is 0.84, the coefficient of determination is 0.87 and runoff volume difference is -3.40%. In individual years the ranges of Nash-Sutcliff efficiency, coefficient of determination and volume difference are 0.79 to 0.91,0.86 to 0.91and 10.38% to 4.74% respectively. Table 1.4 Calibration and validation statistics of the UBC Watershed Model for Astore Watershed

For the whole validation period, the Nash-Sutcliffe efficiency and coefficient of determination are 0.84 and 0.87, respectively, whereas runoff volume difference is reduced to -1.77%. The overall validation results looks better than calibration, but in individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and runoff volumes difference are simulated to be 0.67 to 0.92, 0.86 to 0.94 and -17.97% to 2.15% respectively The graphs between observed and simulated flows for the calibration year 2001-02 and the validation year 1997-1998 are given in Figure 1.5. In winter season, from October to March, almost whole of the runoff is inthe form of base flow. The flows during the months of May and June are a result of snowmelt, whereas glacial nielt accounts for flows during July to September. These results are inline with the findings of (Khan, 2001). Since Astore is predominantly a snow-fed catchment, therefore most of the contribution in flows is because of snow melt.

16

6. 5.

Figure 1.5: Comparison of observed and simulated flow along with various components of runoff - Astore watershed for: a) calibration year (2001-2002) and b) validation year (1997-1998)

17

Siran Watershed
The results for Siran watershed for both calibration and validation periods are given in Table 1.5. For the whole calibration period of Siran watershed, the Nash-Sutcliffe efficiency is 0.66, the coefficient of determination is 0.96 and runoff volume difference is -2.18%. In individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and volume difference are 0.60 to 0.68,0.63 to 0.82 and -13.38% to 2.87% respectively For the whole validation period, value of both the Nash-Sutcliffe efficiency and coefficient of determination is 0.74 and 0.76, whereas runoff volume difference is reduced to 0%. The overall validation results looks better than calibration, but in individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and runoff volume difference are simulated to be 0.57 to 0.85, 0.65 to 0.85 and -11.47% to 15.31 % respectively. The graph between observed and simulated flows for the year 2001-02 is given in Figure 1.6. The results of UBC watershed model for the Siran catchment are not as good in comparison with those for the Hunza and the Astore catchments, because of the reason that Siran is typically a flashy river watershed which comes under the influence of Monsoon. Due to erratic nature of Monsoon and high ratio of variation in topography to watershed area, it is rather difficult to model such a watershed.

Table 1.5 Calibration and Validation Statistics of the UBC Watershed Model for Siran Watershed
Nash-Sutcliffe

Hydrologic years Calibration period


1999-2000 2000-2001 2001-2002

efficiency (Eft)

Coefficient of determination . (R2}


0.63 0.82 0.69 0.69

Runoff volume diff (%DV)

Mean observed flow (m%}


10.79 7.43 9.27 9.16

Mean simulated flow (m3/s}


11.10 7.74 8.03 8.96

0.60 0.66 0.68 0.66

2.87 4.17 -13.38 -2.18

Whole period (1999-2002) Validation period


1995-1996 1996-1997 1997-1998 1998-1999

0.74 0.57 0.85 0.73 0.74

0.80 0.65 0.85 0.77 0.76

15.31 -11.47 -7.81 8.60 0.00

22.14 21.62 24.58 11.51 19.96

25.53 19.14 22.66 12.50 19.96

Whole period (1995-1999)

18

8. 7.

Figure 1.6 Comparison of observed and simulated flow along with various components of runoffSiran watershed for: a) calibration year (2001-2002) and b) validation year (1997-1998)

19.

1.4

UBC WATERSHED MODEL APPLICATION TO JHELUM AND KABUL BASINS

1.4.1 Jhelum Basin


The Jhelum River basin is located in the northern areas with the latitude and longitude range of 34 04' to 35 08' and 73 17' to 75 16' respectively. This river basin is bounded in the west by southwestern part of Indus River basin, in the north by Indus and Astor River basins and in the east by Shingo River basins. The basin stretches over an elevation range of 1,200 masl to more than 4,700 masl. The catchment area of Jhelum River up to Azad Patan is 26,485 km2. Jhelum River is mainly fed by both rain and snowmelt. Snowmelt accounts for more than 50% of the flow in the Jhelum River. The basin is much more dependent on the variable monsoon runoff than Indus basin. The largest part of the total annual runoff is generated in summer months initially by the melting of seasonal snow cover and then by monsoon rains.

1.4.2 Results
The results for Jhelum basin for both calibration and validation periods are given in Table 1.6. For the whole calibration period of Jhelum basin, the Nash-Sutcliffe efficiency is 0.79, the coefficient of determination is 0.80 and runoff volume difference is 5.95%. In individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and volume difference are 0.70 to 0.84, 0.75 to 0.86 and -0.64% to 25% respectively. For the whole validation period, value of both the Nash-Sutcliffe efficiency and coefficient of determination is 0.75 and 0.77, whereas runoff volume difference is -7.17%. The overall validation results look less better than calibration, but in individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and runoff volume difference are simulated to be 0.63 to 0.77, 0.69 to 0.89 and -9.95% to 22.24% respectively. The graph between observed and simulated flows for the year 2001-02 is given in Figure 1.7. The results of UBC watershed model for the Jhelum basin seem to be good keeping in mind the complex nature of the basin due. to erratic nature of Monsoon and high ratio of variation in topography to watershed area, it is rather difficult to model such a watershed.

20

9.

Table 1.6 Calibration and validation statistics of the UBC watershed model for Jhelum basin

Figure 1.7: Comparison of observed and simulated flow along with vanous components of runoff for Jhelum basin

21

1.4.3 Kabul Basin


The Kabul River originates from the Unai Pass of the Southern Hindukush at an elevation of 3,000 m above sea level (masl). It drains eastern Afghanistan and then enters Pakistan just north of the Khyber Pass. Kabul River is mainly snow fed with about 5% of the catchment area covered with glaciers. The average annual precipitation for the entire basin is more than 700 mm, 90% of which is in the form of snow. Most of the precipitation occurring in this watershed is generated from westerly depressions but occasionally monsoon currents also reach in this catchment. The annual flows in the Kabul River are less than one-third of that in the Indus River. The largest part of the total annual runoff is generated in summer months mainly by the melting of seasonal snow cover and a small proportion due to glacial melt.

1.4.4 Results
The results for Kabul basin for both calibration and validation periods are given in Table 1.7. For the whole calibration period of Kabul basin, the Nash-Sutcliffe efficiency is 0.80, the coefficient of determination is 0.83 and runoff volume difference is 12.54 %. In individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and volume difference are 0.70 to 0.83, 0.80 to 0.86 and 2.05 % to 23.08 % respectively. For the whole validation period, value of both the Nash-Sutcliffe efficiency and coefficient of determination is 0.76 and 0.77, whereas runoff volume difference is -3.91%. The overall validation results look very good but as compared to the calibration results the efficiency is reduced. In individual years the ranges of Nash-Sutcliffe efficiency, coefficient of determination and runoff volume difference are simulated to be 0.63 to 0.90, 0.65 to 0.92 and -2.42 % to 14.93 % respectively. The graph between observed and simulated flows for the year 1995~96 is given in Figure 1.8. The hydrological nature of the Kabul basin is bit complex as it involves snowmelt as well as rainfall outflow. Also it has a small portion of basin area covered with glaciers. Keeping in view these facts, the basin outflows are very well reproduced by the UBC watershed model.

Table 1.7 Calibration and Validation Statistics of the UBC Watershed Model for Kabul Basin
Coefficient of determination (R2)

Hydrologic years Calibration period 1999-2000 2000-2001 2001-2002

Nash-Sutcliffe efficiency (Eff)

Runoff volume diff. (dV) (%)

Mean observed

Mean simulated

flow (m3/s)

flow (m3/s)

0 .70 0.80 0 .83

0.84 0.80 0.86

23.08 2.05 12.92 12.54

471.79 496.15 580.78 516.2

580.69 506.30 655.81 580.93

Whole period (1999-2002)


Validation period 1995-1996 1996-1997 1997-1998 1998-1999

0.80

0.83

0.90 0.68 0.63 0.87

0.92 0.74 0.65 0.91

-2.42 -15.48 -9.05 14.93

899.07 879.83 882.34 711.86

877.07 743.62 802.49 818.11

Whole period
(1995-1999)

0.76

0.77

-3.91
22

843.31

810.37

Figure 1.8: Comparison of observed and simulated flow along with various components of runoff for Kabul basin
1.4.5 Conclusions

In this study, the UBC watershed model has been calibrated and validated in three watersheds of Upper Indus Basin. The three watersheds are Hunza, Astore and Siran, representing the glacial fed, snow fed and rain fed regimes of Northern Pakistan, respectively. The input parameters for the model have been derived from GIS and it proved to be a very useful source in helping with the calibration and validation of the models. This is because of GIS's ability to compile many different data sources to aid in calculating hydrologic parameters. The performance of UBC is found to be satisfactory although the model better simulates the flow from the glacial and snow fed watersheds than from the rain fed watershed. These findings suggest that the UBC is able to simulate the runoff in the region having limited data with satisfactory accuracy. Therefore, the next step will be the utilization of this model in testing different climate and land use change scenarios for exploring its effect on the water resources in the study area.

23

APPLICATION OF DISTRIBUTED HYDROLOGY SOIL AND VEGETATION MODEL (DHSVM) FOR SIRAN RIVER BASIN
2.1 Distributed Hydrology Soil and Vegetation Model (DHSVM)
Lumped hydrological models are being widely used in forecasting response of watersheds to given inputs of precipitation. A critical shortcoming of spatially lumped (often termed conceptual) hydrologic simulation models is their inability to represent the spatial variability of hydrologic processes and watershed parameters (Moore et al., 1991). Recent developments in generation and management of digital topographic and other land surface data and availability of high-spec desktop computers have now made it possible to develop and implement spatially distributed hydrological modeling. Spatially distributed hydrological models are capable of reading map-based or image-based land surface characteristics data (especially digital topography), soil and land cover data to predict the response of watershed in its true spatial context. They are particularly useful for assessment of the hydrologic effects of land surface change and climate change. Distributed Hydrology-Soil-Vegetation Model (DHSVM) initially developed by Wigmosta et al. (1994) at the University of Washington is one ofthe models which have full integrity with standard GIS databases and mesoscale climate models for the simulation of hydrological processes in their true spatial and temporal context. The attributes required by the DHSVM include topography, aspect, slope, soil properties, vegetation type and age as well as the distribution of stream channels and logging roads. It provides a dynamic (one day or shorter time step) representation of the spatial distribution of soil moisture, snow cover, evapotranspiration, and runoff production. It consists of a two-layer canopy representation for evapotranspiration, a two-layer energy-balance model for snow accumulation and melt, a multi-layer unsaturated soil model, and a saturated subsurface flow model. Meteorological inputs are precipitation, temperature, wind, humidity and incoming shortwave and long-wave radiation. Digital elevation data are used to model topographic controls on incoming shortwave radiation, precipitation, air temperature and down-slope water movement. Surface land cover and soil properties are assigned to each digital elevation model (DEM) grid cell or pixel. The DEM resolution is arbitrary, but the land surface is usually represented with pixels of dimension less than 100 m by 100 m, as represented in the figure 2.1.

24

Figure 2.1: Model Representation of Watershed Source: http://hydrology.pnl.govlimages/projects/c1imate1,jpg DHSYM codes were provided by the University of Washington Seattle with full permission to use and update. The GNU 'C' codes have adapted and compiled at Global Change Impact Studies Centre, Pakistan for the purpose of research and operational use. 2.1.1 Treatment of Meteorological Variables In each model pixel, the land surface may be composed of over story vegetation, understory vegetation, and soil. Meteorological conditions (precipitation, air temperature, solar radiation, wind speed, vapor pressure) are prescribed at a specified reference height well above the overstory. Solar radiation and wind speed are attenuated through the two canopies. Snow, if present, is assumed to cover the understory and thus affects radiation transfer and the wind profiles via increased albedo and decreased surface roughness. Temperature is extrapolated using standards lapse rates. Precipitation from meteorological stations is extrapolated using one of the preferred method such as INVDIST, V ARCESS or NEAREST Temperature and relative humidity are not adjusted through the canopy. Precipitation occurring below a threshold temperature is assumed to be snow.

2.1.2 Process Simulation An independent one-dimensional (vertical) water balance is calculated for each pixel (Wigmosta et al., 1994).

25

Evapotranspiration
Evaporation of intercepted water from the surfaces of wet vegetation is assumed to occur at the potential rate. Transpiration from dry vegetative surfaces is calculated using a Penman-Monteith approach. The model follows Entekhabi and Eagleson (1989) in using a soil physics-based approach to calculate soil evaporation.

Interception
Snow interception by the overstory is calculated as a function of Leaf Area Index and is adjusted downward for windy or cold conditions (Schmidt and Troendle, 1992). Intercepted snow can be removed from the canopy through snow melt, sublimation, and mass release. Melt of intercepted snow is calculated based on a single layer energy balance approach. Mass release occurs if sufficient melt water is generated during an individual time step such that the snow slides off the canopy (Bunnell et aL, 1985; Calder, 1990). Drip from the canopy is added to the ground snowpack (if present) as rains while the cold content of any mass release or un-intercepted snow is added directly to the ground snowpack.

Soil Water
Ground snow accumulation and melt are simulated using a two-layer energy-balance model at the snow surface, similar to that described by Anderson (1968). The model accounts for the energy advected by rain, throughfall or drip, as well as net radiation and sensible and latent heat. Bulk transfer coefficients for turbulent exchange are calculated based on the aerodynamic resistance from the snow surface to the calculated two-meter wind and adjusted for atmospheric stability. Unsaturated moisture movement through the soil layers is calculated using Darcy's law. This downward moisture flux recharges the grid cell water table. Each DEM grid cell in tum exchanges saturated subsurface flow with its eight adjacent neighbors according to topographic slope. This method allows a transient, three-dimensional representation of saturated subsurface flow. Return flow and saturation overland flow are generated in locations where grid cell water tables intersect the ground surface.

Routing of Runoff
Runoff generation can be routed toward the watershed outlet by a distributed velocity surface routing algorithm (Maidment et al., 1996). Surface runoff generated at a given pixel moves directly toward the watershed outlet based on a unit hydro graph that can include both a linear translation and storage component. The travel time for each pixel is calculated based on its flow path and the velocity vectors along that flow path. The local pixel velocity is determined by the upstream drainage area and the local slope. Although this method is computationally simple, it does not allow for downstream re-infiltration. An alternative method for surface runoff and open channel routing uses explicit information on the location of stream channels and road networks (Perkinset aI., 1996; Bowling and Lettenmaier, 1996). These two networks are imposed on the digital elevation model of the watershed topography as GIS coverage's of vectors mapped to specific pixels. Runoff is then routed though the combined

26

road and channel network using a Muskingum-Cunge scheme. An illustration of DHSVM hydrological simulation process is given in the following figure.

Figure 2.2: Schematic representation of DHSVM inputs, preprocessing requirements, outputs and interaction with GIS software. DHSVM options are indicated with dashed arrows

27

10. 2.2 The Siran Watershed 2.2.1 Description


The Siran watershed lies in the monsoon-fed part of the Indus Basin and measures 1060 Km2 up to Phulra gauge. It is one of 37 sub-basins of the Upper Indus Basin (UIB) as delineated on USGS/DAAC Basins coverage (Figure 2.3). The UIB is the upper part of the Indus Basin which partly falls in China, Jammu & Kashmir and Pakistan and feeds the Tarbela reservoir. The elevation of the Siran watershed ranges between 834 and 4199 m and terrain comprises of two valleys namely Pakhli and Chattar, surrounded by moderate to steepy sloping mountains. Main soil textural classes as identified by Soil Survey of Pakistan are Figure 2.3: Location of Siran Watershed in UIB silty loam (26%) sandy loam (20 %) and silt (16 %). As interpreted from Landsat MSS for March 1979, the dominant land covers are conifer forests (30 %), agriculture (34 %) and open shrubs and grass (27 %).

2.2.2 Meteorology
Analysis of data from 1969-1983 of the three meteorological station i.e., Oghi, Phulra and Shinkiari (Figure. 2.4) shows that mean annual precipitation of the watershed is 1243 mrn, of nearly 2/3rd is received during July, August and September. Uppermost reaches of the Siran watershed receive snow from December to February, which melts down completely by the month of April. This watershed receives both the winter rains from the passing westerly disturbances and monsoon rains from either the South westerly currents from the Arabian Sea or South easterly currents from the Bay of Bengal.

2.2.3 Preprocessing of Input Data:


The Siran DEM was generated from digitized topographic maps at 50-100 feet contour interval. The contour coverage was latticed TIN and rasterized into 80 m regular grid. Outside area was masked by imposing the mask for the Siran watershed (Figure 2.4). Soil information for the Siran watershed by digitization of soil survey maps at 1: 250000 scale and rasterized at 80 m within ARCINFO. Due to smaller size of the watershed and large drainage density, stream network is set to discharge directly to Phulra outlet. For Land use classification, four different land covers have been used. Land cover grid of 1979 has been derived from the image processing of the Landsat MSS (Figure 2.5(a)), while land cover grid of 1989 is derived by the processing of Landsat TM (Figure 2.5(b)), using maximum likelihood classifier.

11.

Figure 2.4: Digital Elevation Model of Siran Watershed and location of meteorological stations

Obvious change in the pattern of land cover between the two grids is observed. A prominent decrease in Dense Cover Conifers can be noticed from the Figure 2.5. However, special codes in C++ have been written to analyze the difference in two land covers within 10 years. Table 2.1 shows the vegetation change between the two land cover grids. Table 2.1 shows a significant change in vegetation types of the two grids. Out of all the vegetation It is obvious that two vegetation types i.e., Dense Cover Forests and Medium Cover Forests have got significant percentage area as well as a significant change. The noticeable thing is that this conversion of vegetation type between the two grids is not linear. As illustrated in Table 2.1, there is overall decrease of 54.89% of Dense Cover Forests from 1979-89 but Figure 4 shows that there is also some area which is converted into Dense Cover Forest in 1989 from Medium Cover Forests of 1979.

Figure 2.5: Land use grids for Siran Watershed (a) 1979 (b) 1989. Obvious change in land use pattern can be observed Table 2.1: Percentage area of different land use types for both the years 1979 & 1989

Figure 2.6: Conversion of Land use types Dense Cover Forests & Medium Cover Forests from 1979 into Land use types of 1989.

30

Special codes in C++ are written to generate two more land cover grids presuming two scenarios. In first scenario, all the Medium Cover Conifers in 1989 land use grid are converted into Dense Cover Conifers as shown in Figure 2.4(b). Analysis showed that percentage area of Dense Cover Conifers is increased to 32% in the presumed scenario as compared to 6.14% of 1989. In the second scenario, all the Grassland is converted into Cropland as shown in Figure 2.4(a). Analysis showed that percentage area of Cropland is increased to 61.35% in the presumed scenario as compared to 34.4% of 1989.

Figure 2.7: Scenarios of Land use grids for Siran Watershed (a) Cropland Increase (b) Dense Cover Forest Increase

2.2.4 Meteorological Data The meteorological data for the three stations located within the Siran watershed was used to run the DHSVM. Since hydro-meteorological data is not available for an interval less than 24 hours, DHSVM was tuned to run on daily time step. Daily meteorological data for the three stations namely Phulra, Oghi and Shinkiari (Figure 2.5) was formatted to input to DHSVM. These are temporary stations established and run by Surface Water Hydrology Project of Pakistan WAPDA. Discharge of the Siran River is also recorded by the same authority on 24-hourly basis. The radiation data for running DHSVM was acquired from Pakistan Meteorological Department which runs its station in a nearby watershed at Kakul station. The simulations were initially carried out from January 1979 to December 1980.

31

2.3 Calibration of DHSVM:


2.3.1 Soil Water Content

The daily water balance of the Siran watershed for the period 1979 and 1980 has shown that spatially-averaged soil water content fluctuates between 950 mm to 1100 mm. This temporal fluctuation is highly dependent on intensity of precipitation but overall soil water flux is negligible indicating that no excessive intake or outflow has occurred

2.3.2 Snow Water Equivalent

During winter 1979, the watershed received ample precipitation as snow which accumulated as snow and melted down completely somewhere in May-June 1979. This DHSVM-simulated behavior was compared with the snow information interpreted from Landsat MSS image acquired on 5 March 1979. The DHSVM simulations were found to be close to factual position on 5 March 1979.

Figure 2.9: Snow water equivalent of Siran watershed for 1979 and 1980 (Net Snow flux is zero).

32

2.3.3 Evapotranspiration DHSVM simulated the basin-averaged evapo-transpiration of the Siran ranges between 1.0 and 3.0 mm per day in summer. During winter months it ranged between 0.5 and 1.2 depending upon the climatic and soil moisture conditions (Figure 2.10).

Figure 2.10: Daily actual evapo-transpiration from Siran watershed. The fluctuations are highly dependent on precipitation and resultant soil moisture content. 2.3.4 Flow Simulation Daily discharge of the Siran River at Phulra as simulated by DHSVM with 80 m databases, for the years 1979 and 1980 is shown in Figure 2.11(a) and 2.11(b). For the purpose of results analyses, first two months of 1979 are omitted due to some mismatching initial conditions of snow and soil water. Total volume of water contributed by the Siran watershed from 1 March 1979 to 31 December 1980 was 881 million cubic meters. For the same period WAPDA had measured actual volume of water from the Siran at Phulra as 927 million cubic meters. The accuracy of total volume of water for this period was therefore calculated 88.3 % (R = 0.64). Mean daily discharge was simulated as 1.31 against the observed 1.49 million cubic meters.

33

12.
13.

Figure 2.11 (a): Simulated and observed hydro graphs for the winter months 1979 and 1980

Figure 2.11 (b): Simulated and observed hydrographs for the Monsoon 1989
The accuracy of simulated discharge varies significantly with season. In winter months, accuracy of volume goes as high as 99.73 %. During the monsoon 1979, the overall accuracy was 91.92 % but the very next year, the monsoon accuracy dropped to 62.33 % (Table 2.2).

34

Table 2.2: Overall and seasonal accuracies of DHSVM outputs for 1979 and 1980

During the winter season (from 1 October 1979 to 31 March 1980), the accuracy of simulated volume of water was as high as 99.7 % (R = 0.88). However, during the monsoon season of 1979 (from 1 July 1979 to 30 September 1979) the accuracy of simulations dropped to 91.92%. In 1980, the accuracy of simulated stream flow was even lower at 62.33 %, due to the probable reasons discussed later in this paper. During this period, DHSVM could not simulate most of the prominent peaks as observed at Phulra outlet. 2.3.5 Results and Discussion

The outputs of the DHSVM simulation by using four different land use grids for the Siran watershed are obtained by designating the position of the Phulra outlet in the stream network input file. As discussed above, out of four land use grids two grids represent condition of land cover for the year 1979 & 1989 where as the other two grids represent the scenarios of forest and vegetation increase whose results are shown in Table 2.3:
Table 2.3: Accumulated Discharges of DHSVM Simulations of two years using all the four Land use grids

Land Use Grids 1979 1989 Increase in Forests Increase in Cropland

Discharge (Million m3) 881 925 855 813

Trend w.r.t 1979 Increased Decreased Decreased

As it is obvious from the results that land use changes has a significant effect on discharge of water. There is a difference of 44 million cubic meters between 1979 and 1989. As discussed earlier, it is due to the intense deforestation in that period caused by the Afghan settlers who used the forest wood to spend their living. In forest increase scenario an expected decrease in accumulated discharge of 70 million cubic meters as compared to 1989 is observed showing this much amount of water is needed by the increased forests for their nourishment. Where as in the cropland increase scenario an obvious decrease in accumulated discharge of 113 million cubic meters is observed as compared to 1989.

35

In order to check the authenticity of the results the trend of evapotranspiration is also observed. Figure 10 shows the map outputs of evapotranspiration at 10th June which are quite acceptable for the given scenarios. For the year 1979, a significant amount of evapotranspiration is shown which is due to the high dense cover forests having 13.1 % of total area as compared to 1989 in which only 6% of dense cover forests were present. Similarly, for forest increase scenario, again a high amount of evapotranspiration is observed which is due to the high percentage of dense cover forests covering the most of the area. 14.

Figure 2.12: Evapotranspiration Map Outputs of DHSVM Simulations at 10 June 1979, (a) 1979 (b) 1989 (c) Cropland Increase (d) Forest Increase

th

Figure 2.13: Accumulated Evapotranspiration of DHSVM Simulations of two years using all the four Land use grids.

36

Total amount of evapotranspiration for the simulation of two years using all the four grids is shown in Figure 2.13. As in Cropland Increase scenario, cropland covers appox. 62% of the total area as compared to 34% of the total area covered in 1989; therefore cropland increase scenario has the maximum amount of accumulated evapotranspiration which is also in consensus with our results.

2.4 Issues of Hydro-Met Data


DHSVM ideally employs a 1-3 hour time step. At present, 1-3 hour meteorological data are not available for the stations operated by Pakistan Meteorological Department and Water & Power Development Authority (WAPDA). Even some daily data has restrictive use. Therefore validation of DHSVM has been conducted using limited daily meteorological data from stations that are sparsely located in valleys which do not represent the conditions prevailing at the higher elevations and on different sides of the slopes. The larger discrepancy between the observed and simulated discharge during the monsoon is indicative of the problems of the meteorological forcing, as a single precipitation lapse rate was prescribed based on limited data. WAPDA is now running 18 high-altitude data collection platforms (DCPs) in the northern area of Pakistan. Once this data becomes available, further testing using DHSVM can be performed to evaluate the model performance at different spatial resolutions.

2.5 Conclusions
Distributed hydrologic modeling can be a powerful tool for investigation of hydrologic effects of land use change. The four applications described here illustrated how such a model can be linked with modem GIS platforms to predict the effects of land use change on watershed hydrology. DHSVM requires a detailed description of the distribution of physical parameters affecting the water and energy balance at the land surface. Due to these data input requirements, the method is facilitated by use and processing of GIS vegetation, topographic and soil data layers. Model input files are derived directly from GIS layers, and model output is readily processed and interpreted using GIS software. The four examples, for Siran Watershed, illustrated how the distributed model output maps can provide insights into hydrologic changes in a watershed, and more importantly, how areas within the watershed that contribute most to these changes can be identified. For instance, if some organization wants to launch a forestation program in some catchment, then such model can ably predict the effect of that change on the streamflow, flood peaks, sediment control etc.

37

EVALUATING SPATIAL RESOLUTION FOR DISTRIBUTED HYDROLOGICAL MODELLING IN THE INDUS RIVER BASIN

3. INTRODUCTION
Semi-arid watersheds lying in the mid-latitudes are reported to be specifically vulnerable to the drastic impacts of climate changes. In western Himalayan watersheds of the Indus Basin, hydrologic equilibrium is threatened by a large scale changes in land use on one hand, and by climate change on the other hand. Some recent simulations of deglaciation have concluded that pace of deglaciation of the western Himalayan watersheds will be much higher than those feeding eastern rivers followed by a crisis of severe water shortage (Rees, 2004). These scenarios have compelled the water authorities to revisit their medium-term and long-term plans to meet the national water needs for agriculture and other sectors. In Pakistan, short-term forecasting of river inflows is conducted by Water & Power Development Authority (W APDA) through lumped and semi-distributed models like the University of British Columbia Watershed Model. The Water planners are now becoming more interested in spatial analyses of hydrologic processes in complex terrain of western Himalayas where lumped models are unsuccessful to assess the spatial variations in water balance. Distributed hydrologic simulation is now believed to be a feasible tool for the assessment of available water resources and for the quantification of impacts of land cover changes and global warming on these resources (Van Shaar et al. 1998). Distributed models have provided an opportunity to model the hydrologic processes in the international river basins like the Indus, where major watershed areas lie in more than one country. In particular, the distributed models which have capabilities to couple with regional and meso-scale climate models and to predict outflows of un-gauged watersheds are of specific focus for modeling in the Indus Basin. Physically-based distributed hydrologic models like DHSVM have been designed for use in a variety of geographical and environmental settings (Wigmosta et al. 1994) including the regions of complex terrain (Bowling et al. 2000). However the applications of distributed hydrologic models in the Indus Basin are restricted owing to unavailability of high-resolution spatial data. Although remote sensing has provided digital information on land covers, yet topographic and soil information is still badly lacking for the most parts of the Indus Basin. An important consideration in the implementation of DHSVM in the Upper Indus Basin would be regarding an optimal resolution of spatial databases. On experimental scales, distributed models have been applied to a wide range of resolutions. Strock et aL (1998) reported that the DHSVM has been typically applied at a resolution between 30 and 100 m mainly for the impact studies of forest operations. In those studies where meteorology is simulated by regional climate models like PNNL~RCM, the finest resolution of spatial databases tested at PNNL was 90 m . For majority of watersheds lying in the Upper Indus Basin, fine-resolutions spatial data would not be available in near future for inputting to DHSVM, For the Indus Basin, we have to compromise for a coarser-resolution to operationalize DHSVM. However, evaluation of comparative efficiencies of different resolutions is required because the spatial resolution is reported to affect simulated runoff and peak: discharge significantly.

38

3.1 Criteria of Optimal Spatial Resolution


Following facts were given due consideration while deciding spatial resolution of the Indus Basin spatial databases for inputting to the DHSVM; 3.1.1 Field of Application of Spatial Databases Pakistan is presently transforming from lumped and semi-distributed systems to completely distributed systems, therefore a medium resolution (l00-1000 m) may be considered for preliminary applications in the Indus Basin. However, we need to ascertain whether the DHSVM outputs at medium resolution are comparable to those obtained from high resolution (<100 m) resolution. 3.1.2 Extent of DHSVM Application To date higher resolutions have been largely employed for DHSVM application in small to medium-sized watersheds. The Indus Basin, a regular grid, would cover an area of 3.75 million Km2. This extent would cover whole of Pakistan and Jammu & Kashmir, and parts of India, China, Afghanistan and Iran. For desktop applications, the spatial databases for this large area should be portable. 3.1.3 Commercial Availability of Spatial Digital Data Development of high-resolution spatial databases for the Indus Basin could be highly cost and time demanding. Therefore first priority for DHSVM applications should be to utilize spatial databases developed and released by some national organizations like Pakistan Space and Upper Atmosphere Commission (SUPARCO), Survey of Pakistan, Soil Survey of Pakistan, and Water Resources Research Institute (WRRI) of Pakistan Agricultural Research Council and other international agencies like USGS Distributed Active Archive Center (DAAC) and US National Oceanic and Atmospheric Administration (NOAA). 3.1.4 Computing Requirements For ensuring wider applicability, DHSVM would be implemented on Linux or UNIX installed on Pentium systems. Size of input data, intermediate and final outputs of DHSVM should be manageable on PCs of available specification. 3.1.5 Resolution of Remote Sensing Data For large basins, high frequency and cost effective information on land cover can be obtained from NOAA AVHRR at approximately 1 Km resolution. Therefore, for the DHSVM, topographic and soil databases may be advantageously developed at the same resolution as that of satellite data.

39

3.1.6 Restrictibility of Spatial Data A spatial resolution which is devoid of any strategic detail is recommended for the Indus Basin, so that no restriction is imposed on the use of spatial databases by any organization. After evaluating the above criteria for DHSVM application at the Indus Basin scale, a spatial resolution of 1 Km was considered as optimal.

Figure 3.1: Digital Elevation Model of the Indus River Basin developed from USGA

3.2 Development of Spatial Database


(a) The Indus Basin at 1000 m Resolution Digital Elevation Model (DEM) has been developed for the whole Indus Basin at 1000 m resolution from USGS/DAAC HYDRO1k DEM for Asia. In Lambert Azimuthal Equal Area Projection, total land area covered by the DEM of the Indus Basin measures 3.75 million Km2 extending between the coordinates given in Figure 3.2. This includes the headwater areas of the Indus Basin falling partly in Pakistan, Jammu & Kashmir, and Tibet (China). The DEM of Upper Indus Basin is given in Figure 3.1.

Figure 3.2: Extent of Indus Basin Database UL (56.0670,30.8584), UR (82.4876, 39.0697), LL (61.5623,18.9681), LR (85.1437, 25.7855)

A mask of the UIB was generated from the USGS/DAAC polygon coverage showing the boundary of sub-basins. On the USGS/DAAC Streams coverage, the stream segments representing river outlets were identified and their locations were defined in stream network file of DHSVM.

40

Information on vegetation was derived from USGS/DAAC Land Cover Characteristics at 1000 m resolution by giving the same extent as that of OEM. The USGS/DAAC Land Cover Characteristics dataset was developed at Lambert Azimuthal Equal Area Projection, from NOAA AVHRR spanning April 1992 through March 1993. Dominant land covers of the Upper Indus Basin, as interpreted from USGS/DAAC LCC database are shown in Figure 3. The land covers were reclassed as per the vegetation categories of DHSVM. 15.

The information on soil texture and depth is as such not available at 1000 m resolution. For the development of soil database for the Indus Basin, Staub and Rosenzweig Zobler Near-Surface Soil Texture l-degree grid of the world released by National Geophysical Data Centre I NOAA was warped to the DEM and downscaled to 1000 m resolution.

Stream map for the Upper Indus Basin was generated from the vector coverage of streams provided in l k USGS/DAAC datasets for Asia. The x and y coordinates of each vertex were translated in to the row and column of pixels through which the streams pass. Stream network was generated by routing smaller order streams into the larger order streams by using the standard IDs associated with USGS datasets. (b) The Siran Watershed at 80 m Resolution The Siran watershed lies in the monsoon-fed part of the Indus Basin and measures 1 060Km2 up to Phulra gauge. It is one of 37 sub-basins of the Upper Indus Basin (UIB) as delineated on USGS/DAAC Basins coverage (Figure 3.4). The UIB is the upper part of the Indus Basin which partly falls in China, Jammu & Kashmir and Pakistan and feeds to the Tarbela reservoir. Mean annual precipitation of the watershed is 1100 mm, of nearly 2/3rd is received during July, August and September. Uppermost reaches of the Siran watershed receive snow from December to February, which melts down completely by the middle of March. This watershed was studied due to the fact that it feeds directly to the Tarbela reservoir without merging into the main Indus River and therefore the accumulated errors of inflows are obviated. The elevation of the Siran watershed ranges between 834 and 4199 m and terrain comprises of two valleys namely Pakhli and Chattar, surrounded by moderate to steeply sloping mountains. Main soil textural classes as identified by Figure 3.4: Location of Siran Watershed in UIB

41

Soil Survey of Pakistan are silty loam (26%) sandy loam (20 %) and silt (16 %). As interpreted fro Landsat MSS for March 1979, the dominant land covers are conifer forests (30 %), agriculture (34 %) and open shrubs and grass (27 %). 16. The Siran DEM was generated from digitized topographic maps at 50-100 feet contour interval. The contour coverage was latticed TIN and rasterized into 80 m regular grid. Outside area was masked by imposing the mask for the Siran watershed (Figure 3.5).

Figure 3.5: Digital Elevation Model of Siran Watershed and location of meteorological stations

Land cover was derived by the processing of Landsat TM acquired on 10 July 1989, using maximum likelihood classifier. Finally land cover classes were assigned DHSVM vegetation codes. Soil information for the Siran watershed by digitization of soil survey maps at 1: 250000 scale and rasterized at 80 m within ARCINFO. Due to smaller size of the watershed and large drainage density, stream network is set to discharge directly to Phulra outlet.

42

3.3 DHSVM Application in the Siran Watershed 3.3.1 Meteorological Data The meteorological data for the three stations located within the Siran watershed was used to run the DHSVM. Since hydro-meteorological data is not available for an interval less than 24 hours, DHSVM was tuned to run on daily time step. Daily meteorological data for the three stations namely Phulra, Oghi and Shinkiari was formatted to input to DHSVM. These are temporary stations established and run by Surface Water Hydrology Project of Pakistan W APDA. Discharge of the Siran River is also recorded by the same authority on 24-hourly basis. The radiation data for running DHSVM was acquired from Pakistan Meteorological Department which runs its station in a nearby watershed at Kakul station. The simulations were initially carried out from January 1979 to December 1980. 3.3.2 Parameterization It is assured that all other parameters as defined in the DHSVM input file are kept uniform to assess the effects of spatial resolution on the hydrologic processes. Precipitation is distributed across the watershed by using Inverse Distance Weightage algorithm with daily precipitation lapse rate calculated for the monsoon rains only. Winter precipitation is mostly under the influence of the Mediterranean system, and is more uniformly distributed over large areas. The monsoon rains, on the other hand, are subject to spatial distribution under the local orographic uplift. For the Siran watershed, an increase of precipitation up to 50 percent has been estimated from the lowest elevations at 1000 m to the highest elevations at 4000 m. correspondingly orographic lapse rate for precipitation is calculated for each rainfall event in the daily meteorological files.

3.3.3 Results of DHSVM Simulations at 80 m Soil Water Content The daily water balance of the Siran watershed for the period 1979 and 1980 has shown that spatially-averaged soil water content fluctuates between 950 mm to 1100. This temporal fluctuation is highly dependent on intensity of precipitation but overall soil water flux is negligible indicating that no excessive intake or outflow has occurred (Figure 3.6).

43

18. 17.

Figure 3.6: Daily fluctuation in soil water content of Siran watershed for 1979 and 1980

Snow Water Equivalent During February 1979, the watershed received ample precipitation as snow which accumulated as snow and melted down completely in the first week of April 1979. This DHSVM-simulated behavior was compared with the snow information interpreted from Landsat MSS image acquired on 5 March 1979. The DHSVM simulations were found to be close to factual position on 5 March 1979.

Evapotranspiration
19.

DHSVM simulated the basin-averaged evapo-transpiration of the Siran ranges between 1.0 and 2.5 mm per day in summer. During winter months it ranged between 0.5 and 0.15 depending upon the climatic and soil moisture conditions (Figure 3.8).

Figure 3.8: Daily actual evapo-transpiration from Siran watershed. The fluctuations are highly dependent on precipitation and resultant soil moisture content.

River Discharge
Daily discharge of the Siran River at Phulra as simulated by DHSVM with 80 m databases, for the years 1979 and 1980 is shown in Figure 3.9 (a) and 3.9 (b). For the purpose of results analyses, first three months of 1979 are omitted due to some mismatching initial conditions of snow and soil water. Total volume of water contributed by the Siran watershed from 1 April 1979 to 31 December 1980 was 731 million cubic meters. For the same period WAPDA had measured actual volume of water from the Siran at Phulra as 927 million cubic meters. The accuracy of total volume of water for this period was therefore calculated 79 % (R = 0.52). Mean daily discharge was simulated as 1.14 against the observed 1.45 million cubic meters.

45

Figure 3.9(b): Simulated and observed hydrographs for the Monsoon 1979 The accuracy of simulated discharge varies significantly with season. In winter months, accuracy of volume goes as high as 97.6 %. During the monsoon 1979, the overall accuracy was 83.6 % but the very next year, the monsoon accuracy dropped to 68.8 % (Table --). During the winter season (from 1 October 1979 to 31 March 1980), the accuracy of simulated volume of water was as high as 97.6 % (R = 0.88). However, during the monsoon season of 1979 (from 1 July 1979 to 30 September 1979) the accuracy of simulations dropped to 83.6%. In 1980, the accuracy of simulated stream flow was even lower at 68.8 %, due to the probable reasons discussed later in this paper. During this period, DHSVM could not simulate most of the prominent peaks as observed at Phulra outlet.

46

Table 3.1: Overall and seasonal accuracies of DHSVM outputs for 1979 and 1980
20.

Season I Period April 1979 to December 1980 Winter 1979-1980 Monsoon 1979 Monsoon 1980

Simulated (million m3) 731 216 92.4 184

Observed (million m3) 927 222 111 140

Accuracy% 78.8 97 .6 83.6 68.8

3.4 Comparative Efficiency of 1000 m Resolution


At 1000 m resolution, DHSVM was run for the whole Upper Indus Basin.The outputs for the Siran watershed were obtained by designating the position of the Phulra outlet in the stream network input file. Siran discharge at Phulra outlet simulated by using two resolutions is shown in Figure 3.10.

Figure 3.10: Comparative hydro graphs simulated at 80 m and 1000 m resolutions.

Total volume and mean daily discharge simulated with two resolutions are given in Table 3.2. At a coarser resolution, DHSVM tended to under-estimated the stream flows, particularly during 1979. Against the total volume of 731 million cubic meters simulated at 80 m resolution, DHSVM simulated 609 million cubic meters at 1000 m resolution. The results obtained by using 1000 m are 83.3 % accurate when compared to as those results obtained by using 80 m. The shapes of hydro graphs at 1000 m and 80 m are parallel during high flows and coinciding during low flows. There exists a strong correlation (R = 0.82) between daily stream flows simulated by using 80 m and 1000 m resolutions.

47

Table 3.2: Comparative results of 80 m and 1000 m resolutions Simulated (80 m) 731 1.14 Simulated (1000 m) 609 0.95 Correlation (R) 0.82

Accuracy % 83.3 83.3

Total Volume (million m3) (1 Apr 1979 to 31 Dec 1980) Mean Discharge (million m3/day) (1 Apr 1979 to 31 Dec 1980)

The effect of spatial resolution on the peak discharge is evident from the results obtained at 80 m and 1000 ill resolutions. At 1000 m resolution, DHSVM tended to flatten the hydro graph and peak discharge is significantly lower as compared to that simulated with 80m. .

3.5 Discussions 3.5.1 Issues of Spatial Data Hydrologically corrected DEM of Asia provided by USGSIDAAC was clipped in Global Mapper in float mode. At 1000 m elevation is presented within reasonable levels of accuracy when compared with high-resolution data sets, but slopes are smoothened. The slope is one of the controlling factors in complex terrain of the Himalayas which affects energy balance, infiltration, surface and sub-surface flows, and resultant river flows. It is evident from high-resolution runs for the same area and same period that higher the resolution, more the peak discharge.
Information on soil is not presently available for most of the watersheds of Indus Basin. This study utilized the soil textural information derived from Staub and Rosenzweig Zobler Near-Surface Soil Texture l-degree grid of the world released by National Geophysical Data Centre of NOAA. This coarse-resolution grid for downscaled to 1000 m resolution. Therefore boundaries of soil textural classes are only arbitrary and do not truly represent the field conditions. Soil description for Indus Basin will be improved once more reliable data is available. 3.5.2 Issues-of hydro-met data DHSVM ideally simulates at 1-3 hour frequencies, which is presently not available for the stations operated by Pakistan Meteorological Department and WAPDA. Daily data has also some restrictibility of use. Therefore validation of DHSVM has been conducted with limited daily meteorological data. Another constraint attached with the observed meteorological data is that the stations are sparsely located in valleys which do not represent the conditions prevailing on higher elevations and on different aspects, particularly during monsoon season. WAPDA is now running 18 high-altitude data collection platforms (DCPs) in northern areas of Pakistan. Once this data is available to use in DHSVM applications, the gap in projecting climatic conditions at higher elevations will filled through proper interpolation and the results will probably improve. Discharge measurements are recorded by W APDA only once in a day (0600 hrs) and presumed to apply for the whole day. There are some inherent errors in daily volumes calculated from river discharge records. It is anticipated that DHSVM results would significantly improve if more frequent (preferably telemetric) discharge data is used to assess the accuracy.

48

3.6 Conclusions

in Pakistan, the prospects of implementing distributed hydrologic modeling are linked with the issues of spatial databases. Although the strengths of theory-based distributed model DHSVM in short-term forecasting and long-term impact studies are now fully understood, the issues of availability of spatial databases and their appropriate resolutions need to be thoroughly addressed. We have evaluated the efficiency of two different spatial resolutions for the monsoon-fed Siran watershed of Pakistan. As anticipated, the accuracy of Siran discharge simulated by DHSVM was found to be proportional to resolution of spatial database. For large area applications in the Indus Basin, higher resolutions (80 m) would not be available in near future. Therefore the water planners and mangers may utilize coarser resolution (1000 m) in operational applications of DHSVM. However, the daily results obtained by two different resolutions are highly correlated (R = 0.83) and the accuracy of 1000 m may be improved to bring it comparable to that obtained by 80 m, We recommend that 1000 m resolution may be used in the Indus Basin for the impact assessment of land cover changes and climate change.

49

REFERENCES
Anderson, E. A., 1968: Development and testing of snowpack energy balance equations, Water Resources Research, 4(1), 19-37 Archer, D. R., 2003: Contrasting hydrological regimes in the Indus Basin. J. Hydrol., 274, 198 210. Assaf and Quick, 1991; Hudson and Quick, 1997; Quick et a1., 1998; Micovic and Quick, 1999; Druce, 2001; Morrison et al., 2002 Bowling, L.C. and D.P. Lettenmaier, 1996: Predicting the effects of forest roads using a distributed hydrological model, Eos Trans. AGU, 77(46), Fall Meet. Suppl., F453. Bowling, L.C, P. Str~ck and D.P. Lettenmaier, 2000: Hydrologic effects of logging in Western Washington, United States, Water Resources Research, 36, 3223-3240. Bunnel, F. L., R. S. McNay, C. C. Shank, 1985: Trees and snow: the deposition of snow on the ground - a review and quantitative synthesis, Research Branch, Ministries of Environment and Forests, IWIFR-17, Victoria, B. C., Canada. Calder, I. R., 1990: Evaporation in the Uplands, John Wiley and Sons, Chichester, 148pp Eltahir, E. A. B. and R. L. Bras, 1996: Precipitation Recycling, Reviews of Geophysics, 34(3): 367-379. Entekhabi, D., and P. S. Eagleson, 1989: Land surface hydrology parameterization for atmospheric general circulation models: Inclusion of subgrid scale variability and screening with a simple climate model, Rep. 325, Ralph M. Parsons Lab., 195pp., Mass. Inst. of Technology, Cambridge. Giambelluca, T.W., Holscher, D., Bastos, T.X., Frazao, R.R., Nullet, M.A., and Ziegler, A.D, 1997: Observations of albedo and radation balance over post-forest land surfaces in eastern Amazon Basin. Journal of Climate 10:919-928. Khan, A. R, 200 I: Analysis of hydro-meteorological time series: Searching evidence for climatic change in the Upper Indus Basin. Lahore, Pakistan: International Water Management Institute. M.C.Quick and A.Pipes, 1977: UBC Watershed Model, Department of Civil Engineering, University of British Columbia,Vancouver 8, B.C., Canada Maidment, D. R., IF. Olivera, A. Calver, E. EatheraH and W. Fraczeck, 1996: A unit hydrograph derived from a spatially distributed velocity field, Hydrological Processes, 1 O( 6), 831-844 Moore, LD., R.B. Grayson and A.R. Ladson, 1991: Digital terrain modeling: A review of hydrological, geomorphological, and biological applications, Hydrological Processes, 5, 3-30.
' J

50

NOAA, 2004: National Weather Service, Office of Climate, Water, and Weather Services: Hydrology Services Official Website of NWS/NOAA. Perkins, W.A., M.S. Wigmosta, B. Nijssen, 1996: Development and testing of road and stream drainage network simulations within a distributed hydrologic model, Eos Trans. AGU, 77(46), Fall Meet. Suppl., F453. Rees, G. and Collins, D. N, 2004: An assessment of the potential impacts of deglaciation on the water resources of the Himalayas, Project Report SAGARMATHA DFID KAR Project No. R7980. Schmidt, R.. A, and C .. A. Troendle, 1992: Sublimation of intercepted snow as a global source of water vapor, Proceedings, 60th Western Snow Conference, Colorado State University, Fort Collins. Strock, P, L. Bowling, P. Welterbee and D. Lettemnaier, 1998: Application of a GIS-based distributed hydrology model for prediction of forest harvest effects on peak stream flow in Pacific Northwest, Hydrologic Process, 12,889-904. Van Shaar. J, D.P Lettenmaier, L Goodman, and K. Cornnor, 1999: Effects of Land Cover Change on the Hydrologic Response of Pacific North West forested Catchments, AGU Poster 12/1999, AGU 1999 Fall Meeting, San Francisco, CA Wigmosta, M.S, L. Vail, and D.P Lettenmaier, 1994: A Distributed Hydrology-Sail-Vegetation Model for Complex Terrain, Water Resources Research, 30, 1665-1679. Wigmosta, M. S .. , D. P. Lettenmaier, and L. W. Vail, 1994: a distributed hydrology-vegetation model for complex terrain, Water Resources Research, 30(6), 1665-1679. Rakesh Kumar, R.D.Singh and K.D Sharma, Water resources of India, National Institute of Hydrology 247667, India. http://www .hydro. washington.eduiLettenmaierlModels/DHSVMlindex. shtml http://www.cig.ensmp.fr/~iahsJhsj/220IhysL22_01_0153.pdf

51

Global Change Impact Studies Centre (GCISC)


Global change science is being aggressively pursued around the world. The Global Change Impact Studies Centre was created in May 2002 to initiate this multidisciplinary effort in Pakistan. The main objective of the Centre is to comprehend the phenomenon of global change, scientifically determine its likely impacts on various socio-economic sectors in Pakistan and develop strategies to counter the adverse effects, if any" Another function of the Centre is to establish itself as a national focal point for providing cohesion to global change related activities at the national level and for linking it with international global research. An important function of the Centre is to help develop manpower that is capable of studying and participating in the international effort to study the global change phenomenon. The Centre also works to increase the awareness of the public, the scientific community and the policy planners in the country to global change.

Global Change Impact Studies Centre (GCISC)


National Centre for Physics (NCP) Complex Quaid-i-Azam University Campus P.O. Box 3022, Islamabad Pakistan
Telephone: (+92-51) 9230226 - 8, 2077386 Fax: (+92-51) 2077385 E-mail: gcisc@comsats.net.pk Web: www.gcisc.org.pk

You might also like