You are on page 1of 84

Snowmelt Runoff Investigation in River Swat Upper

Basin Using Snowmelt Runoff Model, Remote


Sensing and GIS Techniques





























Mateeul Haq
March, 2008



Snowmelt Runoff Investigation in River Swat Upper
Basin Using Snowmelt Runoff Model, Remote Sensing
and GIS Techniques

by

Mateeul Haq



Thesis submitted to the International Institute for Geo-information Science and Earth
Observation in partial fulfilment of the requirements for the degree of Master of Science in
Geo-information Science and Earth Observation, Specialisation: (Geo-hazards)




Thesis Assessment Board

Prof.Dr. V.G. Jetten (Chairman)
Dr. T.W.J. van Asch (External Examiner)
Dr. D. Alkema (1
st
Supervisor)
Prof.Dr. V.G. Jetten (2
nd
Supervisor)
Dr. Z. Vekerdy (2
nd
Supervisor)
Drs. T.M. Loran (Course Director AES)












INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION
ENSCHEDE, THE NETHERLANDS



































Disclaimer

This document describes work undertaken as part of a programme of study at the
International Institute for Geo-information Science and Earth Observation. All views
and opinions expressed therein remain the sole responsibility of the author, and do
not necessarily represent those of the institute.

i
Abstract
Snow is a great water resource, but when melted abruptly it causes flooding. The River Swat,
which flows from the Hindukush mountains starting from Kalam valley, is flooded in summer
when abrupt melting of snow occurs. The aim of this study was to simulate the daily snowmelt
runoff using Snowmelt Runoff Model (SRM) in Kalam basin and use it for flood forecasting and
management of river Swat. An SRTM DEM was used to define the catchment area upstream of
Kalam hydrological station, and to get the elevation zones of 500 meters intervals. Snow cover in
different elevation zones were mapped, using MODIS cloud free images for the melting seasons
(April to August) of 2004-2006. After analyzing the available meteorological and field data,
input parameters were obtained to feed the model. SRM uses accuracy criteria, namely, the
coefficient of determination R
2
and volume difference D
v
, calculated from the measured river
discharge data and the simulated discharge by the model. The model was calibrated for the
melting season of 2004, with R
2
=0.74 and D
v
= -0.7%. When verified for 2005 and 2006, the
results were R
2
=0.54, D
v
= -16.1% and R
2
=0.72, D
v
= -0.3% respectively. Years 2004 and 2006
were very much alike in snowfall in winter and average temperatures in summer and hence their
simulated results were quite alike. However year 2005 received a record 30 years snowfall in
winter and above average temperatures in summer causing changes in runoff coefficients and
degree day factor. After adjustment of these parameters a new simulation resulted in R
2
=0.90 and
D
v
= 3.2%. The weekly forecast by the model from in the 1
st
week of September, and the melting
season of 2005 resulted with R
2
=0.80, D
v
=2.1% and R
2
=0.95, D
v
= -1.2% respectively. Keeping
in view the values of R
2
and D
v
for the melting season simulations and seasonal as well as
weekly forecasts, it is concluded that SRM model can be used for flood forecasts and water
resource management in the study area.



ii
Acknowledgements

Alhamdulillah, all glory be to ALLAH, who gave me the strength to complete my M.Sc. study at
ITC. I would like to acknowledge my family in Pakistan, especially my parents for their spiritual
support and prayers during my studies. Besides, thanks to Institute of Space and Technology
(IST) Pakistan, my parent department for sponsoring my M.Sc. studies here in ITC.

My special thanks go to my supervisors Professor Vector Jetten, Dr. Dinand Alkima for allowing
me to carry out this research and spending their valuable time, efforts and sharing their
knowledge. Especial regards to second supervisor Dr. Vekerdy Zoltan from Water Resource
(WRS) department for his critical comments and valuable suggestions which enabled me to
complete the research. Thanks to all lecturers in ESA department and to all my lovely
international friends and most respect to Dutch people.

Special thanks and regards to Mr. Raza Hussain Ali, Chairman (IST), who provided me this
opportunity of getting master degree in applied earth sciences. Thanks to Mr. Arshad H. Siraj,
Director General (IST), for his continuous support during my studies. Many thanks to Dr. Badar
Munir Khan Ghauri, Deputy Chief Manager and Mr. Rahmatullah Jilani, General Manager,
SPAS division for their struggle to get all the necessary data required for this research work.

Special regards to Pakistan Meteorological Department (PMD), Pakistan Water and Power
Development Authority (WAPDA) for providing the important Meteorological and river
discharge data.

I present my gratitude to Mr. Gulzar Khan, and Mr. Hydayattullah for their support during my
fieldwork. Thanks to my friends Mr. M. Shafique, Mr. Imtiaz Hasan and Mr. Zahir Ali and to all
my friends from Pakistan in ITC for their help during my research.

MAY ALLAH ALLAH ALLAH ALLAH FORGIVE US (AMIN)
MATEEUL HAQ
ENSCHEDE, THE NETHERLANDS

iii
Table of contents
1. INTRODUCTION ............................................................................................................................... 1
1.1 PROBLEM STATMENT........................................................................................................................ 1
1.2 GENERAL OBJECTIVES ...................................................................................................................... 2
1.2.1. Specific Objectives ............................................................................................................... 2
1.3 RESEARCH QUESTIONS...................................................................................................................... 3
1.4 RESEARCH HYPOTHESIS.................................................................................................................... 3
1.5 RESEARCH METHODOLOGY .............................................................................................................. 4
2. LITERATURE REVIEW.................................................................................................................... 6
2.1 OPERATIONAL SNOWMELT MODELS ................................................................................................. 6
2.2 COMPARISON OF SNOWMELT RUNOFF MODELS ................................................................................ 7
2.3 KABUL RIVER SRM STUDY .............................................................................................................. 9
2.4 WHY SRM IS SELECTED FOR THIS STUDY? ...................................................................................... 10
2.5 DESCRIPTION OF THE STUDY AREA...................................................................................... 10
2.6 PHYSIOGRAPHY............................................................................................................................... 11
2.7 CLIMATE OF THE STUDY AREA......................................................................................................... 12
3. DESCRIPTION OF SNOWMELT RUNOFF MODEL................................................................. 14
3.1 HISTORY OF SRM DEVELOPMENT................................................................................................... 15
3.2 DATA REQUIREMENT ...................................................................................................................... 19
3.3 BASIN CHARACTERISTICS ................................................................................................................ 19
3.3.1. Basin and zone areas ......................................................................................................... 19
3.3.2. Area elevation curve........................................................................................................... 20
3.4 SNOW AREA, S................................................................................................................................ 20
3.5 INPUT VARIABLES........................................................................................................................... 21
3.5.1. Temperature, T................................................................................................................... 21
3.5.2. Precipitation, P .................................................................................................................. 22
3.5.3. Snow Depletion Curve........................................................................................................ 23
3.6 INPUT PARAMETERS........................................................................................................................ 23
3.6.1. Runoff Coefficient, C.......................................................................................................... 24
3.6.2. Degree Day Factor: ........................................................................................................... 25
3.6.3. Temperature Lapse Rate: ................................................................................................... 26
3.6.3.1. Fieldwork...................................................................................................................................... 26
3.6.4. Critical temperature, T
CRIT
................................................................................................. 27
3.6.5. Rainfall contributing area, RCA ........................................................................................ 27
3.6.6. Recession coefficient, k ...................................................................................................... 27
4. GIS AND REMOTE SENSING........................................................................................................ 30
4.1 SOFTWARE USED ............................................................................................................................ 30
4.2 DIGITAL ELEVATION MODEL (DEM)............................................................................................... 30
4.2.1. Fill Sink .............................................................................................................................. 31
4.2.2. Flow Direction ................................................................................................................... 32
4.2.3. Flow Accumulation ............................................................................................................ 32

iv
4.2.4. Drainage Network Extraction ............................................................................................ 32
4.2.5. Drainage Network Ordering .............................................................................................. 32
4.2.6. Catchment Extraction......................................................................................................... 33
4.2.7. Catchment Merge ............................................................................................................... 34
4.3 REMOTE SENSING OF SNOW COVER................................................................................................ 35
4.4 MODERATE RESOLUTION IMAGING SPECTRORADIOMETER (MODIS) ............................................. 35
4.5 EXTRACTION OF STUDY AREA SUBSET ........................................................................................... 38
4.6 CLASSIFICATION OF IMAGES FOR EXTRACTION OF SNOW................................................................. 39
4.7 EXTRACTION OF SNOW FOR EACH ELEVATION ZONE...................................................................... 40
4.7.1. Extract by Attribute ............................................................................................................ 40
4.7.2. Resample ............................................................................................................................ 41
4.7.3. Extract by Mask.................................................................................................................. 41
4.7.4. Reclassify............................................................................................................................ 42
5. MODEL RESULTS........................................................................................................................... 45
5.1 ASSESSMENT OF THE MODEL ACCURACY.......................................................................... 45
5.1.1. Accuracy criteria................................................................................................................ 45
5.2 SIMULATIONS OF SRM.................................................................................................................... 46
5.3 MODEL CALIBRATION..................................................................................................................... 46
5.4 SRM VERIFICATION FOR 2005 AND 2006........................................................................................ 47
5.5 DISCUSSION ON MODEL SIMULATIONS ............................................................................................. 49
5.6 FORECASTS USING WINSRM........................................................................................................... 51
5.7 SENSITIVITY ANALYSIS ................................................................................................................... 53
5.8 EFFECT OF DISTORTED DEPLETION CURVE ..................................................................................... 54
6. CONCLUSION AND RECOMMENDATIONS ............................................................................. 55
6.1 CONCLUSION................................................................................................................................... 55
6.2 RECOMMENDATIONS....................................................................................................................... 57
7. REFERENCES: ................................................................................................................................. 58
APPENDICES.............................................................................................................................................. 60


v
List of figures
Figure 1.1: River Swat Pre and Post Flood Comparison. Left side image is showing normal
river flow, whereas right side image is showing overflow in the river. Bright blue in the
image is snow. Dots are showing villages around river Swat. Courtesy to MODIS rapid fire
team......................................................................................................................................... 2
Figure 1.2: Flow chart showing methodology followed in the research. ................................ 5
Figure 2.1: Combined representation of model performance using three criteria: R2, DG and
Dv. The volumes of the prisms indicate the average inaccuracies of the tested models from
all results for snowmelt seasons reported in the WMO project (Rango 1988). ...................... 8
Figure 2.2: Combined representation of model performance using three criteria: R
2
, DG and
D
v
. The volumes of the prisms indicate the maximum inaccuracies of the tested models from
all results as listed for snowmelt seasons and individual years in the WMO tables (Martinec
1989)........................................................................................................................................ 9
Figure 2.3: Location of the Catchment (Study Area) upstream the Kalam gauge station. The
district boundaries have been delineated by Humanitarian Information Centre for Pakistan.
............................................................................................................................................... 10
Figure 2.4: SRTM derived DEM including hill shading effect is showing the elevation
differences in the study area. The base station is at 2000 meters above sea level. ............... 11
Figure 3.1: The structure of the SRM................................................................................... 16
Figure 3.2: Kalam basin has been divided into 8 elevation zones of 500 meter interval...... 19
Figure 4.1: DEM of the study area obtained from Shuttle Radar Topography Mission
(SRTM) ................................................................................................................................. 31
Figure 4.2: Catchment extraction using Drainage network ordering and Flow direction as
inputs. .................................................................................................................................... 34
Figure 4.3: Extracted catchment above Kalam station.......................................................... 35
Figure 4.4: MODIS subset of Afghanistan covering parts of Pakistan including the study
area of Kalam. Grey white color is showing snow and ice. .................................................. 37
Figure 4.5: Subset of Afghanistan has been geo-referenced and re-projected to UTM co-
ordinates. ............................................................................................................................... 38
Figure 4.6: Satellite image of study area on the right after masking it from Afghanistan
subset. .................................................................................................................................... 39
Figure 4.7: Classified image of the study area. Blue color is showing snow whereas dark
brown is snow free. ............................................................................................................... 39
Figure 4.8: Main steps for getting snow cover using ERDAS and Arc GIS tools ................ 40
Figure 4.9: Extract by attribute tool and its result. Grey color is showing snow whereas
white is snow-free area.......................................................................................................... 41
Figure 4.10: Extract by mask tool and the resulted DEM of the area covered with snow.
White color is no-data (snow-free)........................................................................................ 42
Figure 4.11: Snow class in each elevation zone in raster format. Lowest height starts from
Zone1 and ends at Zone8. Zone1 has no snow in the month of May.................................... 42

vi
Figure 4.12: Sequence of snow cover maps from MODIS, Upper Swat River at Kalam, 3032
km2 , 1991- 5790 m a.s.l. Blue is snow covered area. .......................................................... 44
Figure 5.1: Statistics of the daily simulations for the melting season (April-August) 2004. 47
Figure 5.2: Statistics of 2005 melting season simulations following accuracy criteria........ 48
Figure 5.3: MODIS Land surface temperatures and NDVI data showing high temperatures
and stressed vegetation in 2005 in comparison to 2004 where temperatures are low and
vegetation is healthy. Red color in upper half of the figure is showing high temperatures and
dark green in lower half is showing healthy vegetation. ....................................................... 49
Figure 5.4: Statistics of the daily simulations for the melting season 2005.......................... 51
Figure 5.5: Depletion was distorted in the middle of the melting season............................. 54
Figure 5.6: Depletion curve was distorted at the end of the depleted curve. ........................ 54
vii
List of tables
Table 2.1: Runoff models selected for WMO study on Intercomparison of runoff models. adapted
from (Russell 2003) and (WMO 1986). .......................................................................................... 6
Table 2.2: Results of model performance in the WMO project (10 years, snowmelt season) ....... 8
Table 3.1: Temperature data recorded during fieldwork .............................................................. 26
Table 5.1: Sensitivity analysis results ........................................................................................... 53



viii
List of Graphs

Graph 2.1: Annual Precipitation in comparison with precipitation in summer for the last 15 years
has been shown here...................................................................................................................... 12
Graph 2.2: Temperature averages of monthly min, max, and average for the last 20 years
recorded data in Kalam Catchment. .............................................................................................. 13
Graph 3.1: Countries name against the frequency of SRM applications..................................... 18
Graph 3.2: Cumulative area elevation curve of the study area. There are 8 Small dots on the
curve and they are representing hypsometric mean elevation of each zone starting from zone1 to
zone8. ............................................................................................................................................ 20
Graph 3.3: Daily average temperatures of melting seasons of 2004, 2005, and 2006 are used as
input to the model for the respective years. These are then extrapolated to the hypsometric mean
altitudes of all eight zones............................................................................................................. 21
Graph 3.4: Daily rainfall recorded at base station Kalam in 2004 and 2005................................ 22
Graph 3.5: Depletion curves of the snow coverage for 8 elevation zones of the basin Kalam,
derived from the MODIS imagery Zone1: 1991 - 2490 m a.s.l., Zone2: 2491 - 2990 m a.s.l.,
Zone3: 2991 - 3490 m a.s.l., Zone4: 3491 - 3990 m a.s.l., Zone5: 3991-4490 m a.s.l, Zone6: 4491
- 4990 m a.s.l, Zone7: 4991-5490 m a.s.l, Zone5: 5491-5990 m a.s.l. .......................................... 23
Graph 3.6: Runoff coefficients for rain and snow used for the model calibration for the melting
season (April-August) 2004. ......................................................................................................... 24
Graph 3.7: Degree day factor values against the months of the melting season 2004, 2005 and
2006 for Kalam basin, Pakistan..................................................................................................... 25
Graph 3.8: Recession flow plot Qn Vs Qn+1 for Kalam basin in Pakistan. Either the solid lower
envelope line or the dashed medium line is used to determine k-values for computing the
constants x and y. .......................................................................................................................... 28
Graph 4.1: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2004.
....................................................................................................................................................... 43
Graph 4.2: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2005.
....................................................................................................................................................... 43
Graph 4.3: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2006.
....................................................................................................................................................... 44
Graph 5.1: Runoff simulations for the melting season 2004 of Kalam basin. The dashed line is
the measured discharge at Kalam station whereas the solid line is simulated runoff. .................. 46
Graph 5.2: Runoff simulations for the melting season 2005 of Kalam basin. The dashed line is
the measured discharge at Kalam station whereas the solid line is simulated runoff. .................. 47
Graph 5.3: Daily runoff simulations for melting season, 2006. The triangular points measured
discharge values whereas the gray circular points are computed values. ..................................... 48
Graph 5.4: Same period temperatures comparison between 2004 and 2005 ................................ 50
Graph 5.5: Runoff simulations for the melting season 2005. R
2
=0.90 and D
v
=3.2%................... 50
Graph 5.6: Weekly forecast of runoff in the Kalam catchment, 1st week of Aug, 2005.............. 52
ix
Graph 5.7: Melting season runoff forecast for the year 2005 ....................................................... 53


x
List of Appendices
Appendix 1: Average temperatures at Kalam station for the melting season 2004. ..................... 60
Appendix 2: Average temperatures at Kalam station for the melting season 2005. ..................... 61
Appendix 3: Average temperatures at Kalam station for the melting season 2006. ..................... 62
Appendix 4: Measured Vs Computed River Discharge (m
3
s
-1
) 2004 (Melting Season), Kalam
Basin.............................................................................................................................................. 63
Appendix 5: Measured Vs Computed River Discharge (m
3
s
-1
) 2005 (Melting Season), Kalam
Basin.............................................................................................................................................. 64
Appendix 6: Measured Vs Computed River Discharge (m
3
s
-1
) 2006 (Melting Season), Kalam
Basin.............................................................................................................................................. 65
Appendix 7: Precipitation data for 2004, Kalam station............................................................... 66
Appendix 8: Precipitation data for 2005, Kalam station............................................................... 67
Appendix 9: Precipitation data for 2006, Kalam station............................................................... 68

xi
List of Abbreviations
WinSRM Window based Snowmelt Runoff Model
MODIS Moderate Resolution Imaging Spectroradiometer
NDVI Normalized Difference Vegetation Index
N.W.F.P. North West Frontier Province
LST Land Surface Temperature
SRTM Shuttle Radar Topography Mission

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
1
1. INTRODUCTION
Precipitation can be in many different forms, such as rain, freezing rain, hail, sleet, and snow.
Snow is composed of small ice particles and hence is a granular material. It is a soft structure,
unless packed by external pressure. Snow is a great source of water as it acts as natural reservoir
for many water supply systems.
Nature has blessed Pakistan with the icy peaks of the Himalaya, Karakoram and Hindukush in
the north of the country. These mountain ranges receive heavy snowfall in winter, which is a
great water resource feeding most of the rivers in Pakistan throughout the year. Snow is precious
water resource, but when it melts rapidly it causes flooding in most of the rivers in Pakistan.

1.1 Problem Statment

River Swat flows from Hindukush Mountains starting with River Ushu, and Utror, meeting in
Kalam valley and merges into Kabul River in Peshawar valley, North West Frontier Province
(N.W.F.P), Pakistan. River Swat is flooded in summer (Figure 1-1) when abrupt melting of snow
occurs under the combined effects of sunlight, winds, rainfall and heat waves in the region.
There is only one river discharge measurement station and one meteorological station at Kalam
city. Snow monitoring is only limited to finding the depth of snowfall in winter and the river
gauge station is used to warn about flood like situation when water discharge is increased.
Therefore, monitoring of snow-related processes can play a vital role in controlling the flood
problem along with water resource evaluation and management. Snowmelt Runoff Model,
developed by (Martinec 1975) will be used to evaluate the snowmelt runoff in Kalam basin
(Swat District) a tributary of Swat River.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
2

Figure 1.1: River Swat Pre and Post Flood Comparison. Left side image is showing normal river
flow, whereas right side image is showing overflow in the river. Bright blue in the image is snow.
Dots are showing villages around river Swat. Courtesy to MODIS rapid fire team
1.2 General Objectives
The general objective of this study is to simulate and estimate the daily snowmelt runoff in
snowmelt seasons for the study area and apply it for flood forecasting and management.

1.2.1. Specific Objectives
The following specific objectives will contribute to get the general objective.
Simulation of river flows on daily basis in a snowmelt season.
Short term and seasonal runoff forecasts.
Providing the forecasts for flood prediction and water resource management
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
3
Based on the specific objectives of the research the following are the possible research questions:
1.3 Research Questions
What are the parameters to which SRM is more sensitive?
What may be the effect of occasional summer snowfall on runoff simulation?
What may be the effect of long dry periods or above average temperatures in a
melting season on runoff coefficients?
What may be the effect of wet or dry years on the model performance?
How useful the runoff simulation and forecasts would be for flood prediction and
water resource management?
Keeping in view the structure of the model and the previous studies using the same model, the
following are the hypothesis to answer the above research questions.
1.4 Research Hypothesis
SRM is a degree day model; therefore, it is expected to be more sensitive to
temperature lapse rate as this is used to extrapolate temperatures from base station to
higher altitudes.
The depletion curves derived from the measured points are used by SRM to simulate
runoff. These curves are distorted by occasional summer snowfalls (Short lived)
which may lead to excessive melt-water especially if the time interval between too
successive measured points is too long.
Long dry periods or above average temperatures both increase the chances of water
losses due to infiltration or utilization by the stressed vegetation, and hence may
reduce the runoff coefficients for snow and rain.
The model calibrated for one year, is supposed to be verified by other years as well.
Some parameters used in the calibration may vary in different weather conditions.
Therefore, if the calibrated year weather conditions (Dry Year) are drastically
different from the verification year conditions (Wet Year) then problems may occur
in verification.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
4
The daily discharge and the seasonal average discharge measured at the river gauge
station, and the computed daily discharge by the model along with the seasonal
measured volumes and seasonal computed volumes, are used for calculating the
coefficient of determination, R
2
and the volume difference, D
v
. R
2
and D
v
are used to
find the accuracy of the model simulations or forecasts. If the simulated results are
found accurate (Based on the accuracy criteria), then it will be justified to say that
the model can play a significant role in the flood forecasting and water resources
management.
1.5 Research Methodology

SRM will be used for estimating the runoff due to snowmelt. Keeping in view the necessary data
required running the model and the research questions supposed to be answered, following are
the steps followed in the research.
Extraction of the river basin above Kalam city, using SRTM data.
Extractions of snow cover extent from a series of MODIS images for the years 2004,
2005, and 2006.
Calculation of snow covers extent time series in selected elevation zones.
Analysis of hydrological and meteorological data of the respective years of the base
station Kalam.
Calculation of SRM model parameters from the above mentioned hydrological and
meteorological data.
Calculation of the temperature lapse rate from the temperature data recorded during
fieldwork.
Calibration of the model using the above information.
Verification of the model.
Using the model for weekly as well as seasonal runoff forecasts
Sensitivity analysis.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
5
Checking the model run for distorted depletion curve.
Assessment of the model accuracy.
A schematic diagram below shows the methodology to use SRM model for the current study.


Figure 1.2: Flow chart showing methodology followed in the research.


SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
6
2. Literature Review
2.1 Operational Snowmelt Models

Many models have been created around the world over the last four decades to describe
snowmelt runoff. These can be divided into two broad categories, statistical and physical. A
statistical model utilizes statistical relationships between inputs and outputs. A physically based
model describes the physical processes that relate inputs to outputs. In turn, these models can be
applied in a lumped or distributed mode. A lumped model describes catchment processes with
single catchment average values. A distributed model divides a catchment into sections and
carries out model calculations for each section. A lumped model can be considered a single
section distributed model, or equally, a distributed model can be considered a series of small
lumped models. The two more common ways of sub dividing an area of interest for snowmelt
modelling is into elevation zones, or into grid squares (KERR 2005). Some of these models are
briefly discussed in Table 2.1:

Table 2.1: Runoff models selected for WMO study on Intercomparison of runoff models. adapted
from (Russell 2003) and (WMO 1986).
S.No Model Author(s) (Year) Data Required Remarks
1
University of
British Columbia
(UBC)
(Quick 1976)
Temperature,
precipitation,
discharge
Process-oriented,
lumped parameter,
continuous
simulation model
2
CEQUEAU

(Charbonneau
1977)
Temperature,
precipitation,
snowfall, discharge
Hydrological
repercussions
3 ERM (Turcan 1981)
Temperature,
precipitation,
discharge
Water management,
Reservoir operation
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
7
4 NAM (Gotleib 1980)
Temperature,
precipitation,
discharge
Lumped, continuous
rainfall-runoff model
5 TANK
Sugawara et al.
(1974)
Temperature,
evapotranspiration,
precipitation,
discharge
Process-oriented,
semi-distributed or
lumped continuous
simulation model
6
Hydrological
Simulations (HBV)
(Bergstrm 1975)
Temperature,
precipitation,
evapotranspiration,
discharge
Process-oriented,
lumped, continuous
streamflow
simulation model
7
Snowmelt Runoff
Model (SRM)
(Martinec 1975)
Temperature,
precipitation,
snow cover,
discharge
Lumped continuous
snowmelt-runoff
simulation model
8
Stream flow
Synthesis and
Reservoir
Regulation
(SSARR)
(Anderson 1973)
Temperature,
precipitation,
discharge
Lumped continuous
stream flow
simulation model
9
Precipitation
Runoff Modeling
System Model
(PRMS)
(Leavesley 1983)
Temperature,
precipitation,
solar radiation
Deterministic
physical-process
model for runoff
simulations
10
National Weather
Service (NWS)
(Burnash 1973)
Temperature,
precipitation,
discharge
Lumped continuous
river forecast system

2.2 Comparison of Snowmelt Runoff Models

The World Meteorological Organization (WMO) organized an international comparison of
snowmelt runoff models (WMO 1986) in which hundreds of model runs were performed in six
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
8
selected test basins. Table 2.2 compares the numerical results of each model in the WMO
project.
Table 2.2: Results of model performance in the WMO project (10 years, snowmelt season)

Figure 2.1 shows a summary of all numerical values of R
2
, DG (Coefficient of Gain) and D
v

published by WMO (WMO 1986). DG and R
2
have the same formula, but DG uses the average
measured discharge from a number of past years whereas R
2
uses the average measured discharge
of the particular year used in the simulations. Each prism refers to a tested model. The length
along the x-axis corresponds to the arithmetic mean of all (1- R
2
) values, length along the y-axis
to the arithmetic mean of all (1-DG) values, and length along the z-axis to the arithmetic mean of
all D
v
values as achieved in the snowmelt seasons of 10 test years.


Figure 2.1: Combined representation of model performance using three criteria: R2, DG and Dv.
The volumes of the prisms indicate the average inaccuracies of the tested models from all results
for snowmelt seasons reported in the WMO project (Rango 1988).
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
9


In Figure 2.2, the dimensions of the prisms are determined by the worst results of each model for
R
2
, DG and D
v
, i.e., (1-R
2
)
MAX
, (1-DG)
MAX
and (D
v
)
MAX
. All available data for the individual
years and snowmelt seasons, as listed in Table 2.2, are thus contained within each prism. The
differences between models are larger than in Figure 2.2, which means that some calibrating
models had difficulties in the years with unusual runoff conditions but improved their average
results (Figure 2.1) by the more normal years. Prisms in Figures 2.1 and 2.2 labelled "G" refer to
SRM.

Figure 2.2: Combined representation of model performance using three criteria: R
2
, DG and D
v
.
The volumes of the prisms indicate the maximum inaccuracies of the tested models from all
results as listed for snowmelt seasons and individual years in the WMO tables (Martinec 1989).
2.3 Kabul River SRM Study
SRM has been applied only once in Pakistan for the snowmelt season of 1976 and the model
results were published in 1989 (Dey 1989). The basin to which SRM was applied was of River
Kabul with a total area of 63657 sq km. A measure of model efficiency (R
2
) and percentage
difference between the total measured and simulated runoff, Dv were 0.66 and 6% respectively.
The main theme of this research was focused on water resource management for power
generation and irrigation practices. (Dey 1989) concluded that relatively bad simulations for the
Kabul basin are most likely caused by the eccentric position of the climatic station, and not by a
structural deficiency in the Snowmelt-Runoff Model.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
10
2.4 Why SRM is selected for this study?

There are two main reasons why SRM was selected for this study:
1. The strength of SRM is its primary reliance on snow cover areal extent. This allows for
limited data input needs, and the snow-covered-area can be derived from satellite,
aircraft, or ground measurements.
2. Secondly the (WMO 1986) study of Intercomparison of various models proves SRM the
best of all (Figures 2.1, and 2.2).

2.5 DESCRIPTION OF THE STUDY AREA
Kalam catchment (Study Area) is situated in the North West Frontier Province (N.W.F.P) of
Pakistan, between 72 12 & 72 51 E and 35 17 & 35 54 N. The study area is surrounded by
Ghizer district in the North, Chitral in the North West, Upper Dir in the West, Swat in South and
Kohistan in the East (Figure 2.3). SRTM DEM data has been used for the extraction of the
catchment boundary. This catchment belongs to tributaries of the Indus River, starting with Swat
River and Joining Kabul River which finally becomes part of river Indus. The whole area of the
catchment upstream the Kalam station is approximately 2032 sq km.


Figure 2.3: Location of the Catchment (Study Area) upstream the Kalam gauge station. The
district boundaries have been delineated by Humanitarian Information Centre for Pakistan.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
11
2.6 Physiography

Physiography is the study of the physical characteristics and morphological conditions of a
catchment. These have important role and effect on the hydrological characteristics and water
regime. To have a good quantitative and qualitative assessment of the hydrological system, it is
important to know the physiographic characteristics of the catchment under study.

Kalam catchment is mountainous with high internal relief. The elevation range between lowest
and highest altitudes is 1991-5790 a.m.s.l. The huge elevation gradient affects precipitation and
temperature value in the catchment. There are a few lakes in the catchment which can affect the
runoff coefficients for both rain and snow.


Figure 2.4: SRTM derived DEM including hill shading effect is showing the elevation differences
in the study area. The base station is at 2000 meters above sea level.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
12
2.7 Climate of the study area

Kalam valley is situated in the upper reaches of Swat district, in the North-West Frontier
Province. In Kalam catchment the annual rainfall averages around 1030 mm, with about 101.6
mm expected between July and September. October and November are the driest months with
rainfalls generally under 30 mm per month except in the most exposed areas. Graph of annual
precipitation in comparison with precipitation in summer, recorded at Kalam Station, has been
shown in Graph 2.1. Precipitation in winter season is mostly in the form of snow. High mountain
peaks can receive snow even in summer.
Kalam Annual Precipitation in Comparison with Precipitation in Summer
0
200
400
600
800
1000
1200
1400
1600
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Years
P
r
e
c
i
p
i
t
a
t
i
o
n

(
m
m
)
Annual
Summer

Graph 2.1: Annual Precipitation in comparison with precipitation in summer for the last 15 years
has been shown here

Monthly average temperatures in this region (Kalam Catchment) remains below 5 C in the
months of Dec, Jan and Feb, which touches 20 C on the average in June, July and August
(Graph 2.2). Maximum temperatures can reach up-to 26 C whereas minimum can go below -6
C.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
13
Kalam 20 years Min, Max, and Ave Temperature Values
-10
-5
0
5
10
15
20
25
30
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
T
e
m
p
e
r
a
t
u
r
e

(
C
e
n
t
i
g
r
a
d
e
)
Max
Min
Ave

Graph 2.2: Temperature averages of monthly min, max, and average for the last 20 years
recorded data in Kalam Catchment.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
14
3. Description of Snowmelt Runoff Model
The Snowmelt-Runoff Model (SRM) is designed to simulate and forecast daily stream flow in
mountain basins where snowmelt is a major runoff factor (Jaroslav Martinec 2005). SRM was
developed by (Martinec 1975) for small European basins. Thanks to the progress of satellite
remote sensing of snow cover, SRM has been applied to larger and larger basins. Recently, the
runoff was modelled in the basin of the Ganges River, which has an area of 917,444 km2 and an
elevation range from 0 to 8,840 m a.s.l. Contrary to the original assumptions, there appear to be
no limits for application with regard to the basin size and the elevation range.

SRM uses percentage areal snow cover, air temperature, and precipitation as critical input
variables. SRM divides the watershed into elevation zones and accounts for degree-days in each
elevation zone to drive the amount of snowmelt. Specific basin characteristics include runoff
coefficients, degree-day factors, and historical recession coefficients (Shafer 1981). Definition of
the basin includes careful determination of basin areas and, once the elevation zones are
established, finding the area of each zone. The zonal mean hypsometric elevation is determined
for each zone from an area-elevation curve. It is also necessary to know the temperature lapse
rate for the basin. In SRM, Each day during the snow melt season, the water produced from
snow melt and from rainfall is computed, superimposed on the calculated recession flow and
transformed into daily discharge from the basin (Martinec 1983).

Through the use of the zonal mean hypsometric elevations, the actual elevation of the
temperature measurement station and the temperature lapse rate, the melting degree-days for
each elevation zone are calculated. The precipitation for each zone is determined to be either rain
or snow, depending on the average zonal temperature and a critical temperature selected to be
slightly above freezing. The snow coverage for each zone is determined by ground observation,
aircraft photography, or by satellite and is arrayed as a depletion curve over the snowmelt period.
Runoff coefficient estimation requires knowledge of the basin and its hydrology, and it varies
over the year (Martinec 1983). The snowmelt-degree-day factor can be varied throughout the
snowmelt period to account for the changing density and albedo of the snowpack. The recession
coefficient is estimated from historical records of the actual daily average flows.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
15

SRM accumulates the number of degree-days in each elevation zone over the snowmelt period
and discriminates the input precipitation into snow or rain by comparing the assigned critical
temperature to the average daily temperature. Snowmelt is calculated using a degree-day factor
that is applied to the portion of the elevation zone that is snow covered. Within each elevation
zone, an average snow cover depletion curve is used to estimate the temporal change in the snow-
covered area. The snowmelt is distributed according to the chosen elevation zones and summed
to give total average daily runoff from the entire watershed.

3.1 History of SRM Development
SRM model is originated by Jaroslav Martinec, attached at the time to the Swiss Snow and
Avalanche Research Institute at Davos. Martinec (1975) described a snowmelt runoff model
which has come to be referred to as SRM. It has undergone substantial development since 1975,
by Martinec himself in collaboration with Al Rango (US NASA, later US ARS) and Michael
Baumgartner (University of Bern). The model has been amended and extended several times in
the light of operational experience.
SRM was developed specifically to predict snowmelt runoff, unlike HBV and other general-
purpose hydrological models containing a snowmelt routine. It has been extensively applied to
snowmelt modelling in mountainous terrain, and recently also to climate-change scenarios.
The basic structure of SRM is shown schematically in Figure 3.1. The basin is subdivided into
elevation zones. Runoff from all elevation zones is added together before routing, so location is
not taken into account in the model. A daily time step is used. Snowmelt in each zone is
predicted from air temperature, any rainfall is added on, and the total new water is routed
through a single store.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
16

Figure 3.1: The structure of the SRM
The original version of the model can be represented by a single equation (3.1):
1
(1 ) ( )
+
= + +
n n i i i i
i
Q kQ k a T A P
( ) 3.1
Where:
n is the day number
i are indexes of the elevation zones
Q is discharge from the basin (m
3
per second)
T is the air temperature (C) extrapolated to hypsometric mean height of each zone
P is the precipitation (mm and cm both can be used) falling as rain in the zone (when T
> Tcrit)
A is the current snow covered area (SCA) in the zone (% of zone area)
k is a recession coefficient
a is degree day factor (cm per C per day)

One other parameter is implicit in the equation above: a temperature lapse rate used to
extrapolate Ti from temperature at a base station.
The version 4 of SRM is rather more complicated and contains additional parameters, though the
basic structure is unchanged. It can be represented as (equation 3.2):

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
17
1 1 1
(1 ) ( )
+ + +
= + +
n n n n Sn i i i Rn i
Q k Q k c a T A c P
( ) 3.2
Where:

Sn
c
is a correction factor for losses from snowmelt [0, 1]

Rn
c
is a correction factor for losses from rainfall [0, 1]

1 + n
k
is a function of
n
Q

(Other parameters as original SRM equation (3.1))

So far the SRM philosophy was opposed to calibration; users were urged to use standard default
values based on worldwide experience and physically reasonable limits. The only exception was
that k must be determined for the basin concerned by analysis of recession flow.

The current version of SRM is represented by equation (3.3): Each day, the water produced from
snowmelt and from rainfall is computed, superimposed on the calculated recession flow and
transformed into daily discharge from the basin.
[ ]
1 1 1
10000
( ) (1 )
86400
n Sn n n n n Rn n n n n
A
Q c a T T S c P k Q k
+ + +
= + + +


( ) 3.3
Where: Q = average daily discharge [m3s-1]
c = runoff coefficient expressing the losses as a ratio (runoff/precipitation), with
cs referring to snowmelt and cr to rain
a = degree-day factor [cm per C per day] indicating the snowmelt depth
resulting from 1 degree-day
T = number of degree-days [C d]
T = the adjustment by temperature lapse rate when extrapolating the
temperature from the station to the average hypsometric elevation of the basin or
zone [C d]
S = ratio of the snow covered area to the total area
P = precipitation contributing to runoff [cm]. A pre-selected threshold
temperature, T
CRIT
, determines whether this contribution is rainfall or snow. If
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
18
precipitation is determined by T
CRIT
to be new snow, it is kept on storage over
the hitherto snow free area until melting conditions occur.
A = area of the basin or zone [km
2
]
k = recession coefficient indicating the decline of discharge in a period
without snowmelt or rainfall:
k = Q
m
/Q
m+1
(m, m + 1) are the sequence of days during a true recession flow
period).
n = sequence of days during the discharge computation period. Equation (3.3) is
written for a time lag between the daily temperature cycle and the resulting
discharge cycle of 18 hours. In this case, the number of degree-days measured on
the nth day corresponds to the discharge on the n + 1 day. Various lag times can
be introduced by a subroutine.
10000/86400= conversion from cmkm
2
d
-1
to m
3
s
-1


T, S and P are variables to be measured or determined each day, whereas, c
s
, c
r
, lapse rate, T
CRIT
,
k and the lag time are parameters which are characteristic for a given basin. If the elevation range
of the basin exceeds 500 m, it is recommended that the basin should be subdivided into elevation
zones of about 500 m each (Figure 3.2). To date the model has been applied by various agencies,
institutes and universities in over 100 basins, situated in 29 different countries. Almost 50% of
these studies have been done in USA, Spain and Switzerland (Graph 3.1).

Graph 3.1: Countries name against the frequency of SRM applications
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
19
3.2 Data Requirement
Actually there are three input tables required for running the model.
1. The characteristics of the basin with inputs area of each elevation zone and hypsometric
mean elevation of that zone, which can be obtained from area elevation curve of the
study area.
2. The snow cover, precipitation, and temperature data as input variables.
3. The parameters for the basin, which are; runoff coefficient for snow, critical temperature,
runoff coefficient for rain, rainfall contributing area, degree day factor, recession
coefficient, temperature lapse rate and time lag. These parameters are used to calibrate
the model.
3.3 Basin characteristics
3.3.1. Basin and zone areas
The basin boundary is defined by the location of the stream gauge (which is Kalam in this study)
and the watershed divide is identified on a topographic map or DEM. In this study SRTM data
was used for getting the basin boundary above Kalam which is the base station as well. This was
done in ILWIS software using DEM-hydro processing tools. The basin boundary is obtained as
polygon map. The polygon map is used for masking only the basin part of SRTM DEM. The
resulted DEM is then re-sampled into different elevation zones of 500m as recommended by the
model developers. The elevation difference in the study for this research is from 1991m a.m.s.l to
5790m a.m.s.l hence it has 8 elevation zones as shown below (Figure 3.2).

Figure 3.2: Kalam basin has been divided into 8 elevation zones of 500 meter interval.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
20
3.3.2. Area elevation curve

The area elevation curve is obtained from digital elevation model. The model needs hypsometric
mean elevation of each zone to extrapolate various variables to it with a known rate. Temperature
and precipitation are variables which are extrapolated from base station to hypsometric mean
elevation of each, as measurement stations are not available in each zone. The zonal hypsometric
mean elevation, h, was determined from the attribute table, by calculating half area for each zone
as well as adding it to total area of zones below it. This area was then compared against the
elevation in the attribute table (Graph 3.2).


Graph 3.2: Cumulative area elevation curve of the study area. There are 8 Small dots on the curve
and they are representing hypsometric mean elevation of each zone starting from zone1 to zone8.
3.4 Snow Area, S
Snow cover is the most important input variable for this study. Its areal extent gradually
decreases during a snowmelt season. Snow cover has been mapped using images of Moderate
Resolution Imaging Radio-spectrometer (MODIS) sensor onboard Aqua and Terra satellites.
Thanks to MODIS Rapid Response Project at NASA/GSFC who are providing subsets of my
study area free of charge on their website. Detail given in section 4.3
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
21
3.5 Input Variables

SRM needs three input variables, namely daily temperature, daily precipitation data and
snow cover as discussed below.
3.5.1. Temperature, T
SRM estimate the daily snow water equivalent by computing the daily snowmelt depth using the
number of degree days. A degree day is a day with an average temperature one degree above 2
C. For example a day with an average temperature of 15 C, gives 13 degree-days. SRM model
calculate degree days itself from daily average temperatures stored in the database for each
simulation. Average daily temperatures for the years 2004, 2005, and 2006 are used for the
current study (Graph 3.3).

Graph 3.3: Daily average temperatures of melting seasons of 2004, 2005, and 2006 are used as input to the
model for the respective years. These are then extrapolated to the hypsometric mean altitudes of all eight
zones.

The program accepts either temperature data from a single station or from several stations, by
zone. In the current case basin wide temperature data from a single station was used. The altitude
of the station is entered and temperature data are extrapolated to the hypsometric mean
elevations of all zones using the temperature lapse rate.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
22
3.5.2. Precipitation, P

Precipitation is a main component of the hydrological cycle, and as such is of primary
importance in hydrology. Mapping of areal precipitation is generally difficult in mountainous
basins, because it shows large and spatial variations and usually exhibits rapid temporal
variations. WinSRM accepts either a single, basin-wide precipitation input from one station
(which is Kalam in this study) or different precipitation inputs zone by zone. Graph 3.4 shows
the daily rainfall in Kalam catchment (2004-2005).

The precipitation can be treated as snow if the temperature values are less than a critical
temperature (section: critical temperature). The new snow that falls over the previously snow-
covered area is assumed to become part of the seasonal snowpack. This precipitation is stored by
SRM and then melted as soon as a sufficient number of degree-days have occurred. Rainfall data
of 2004, 2005, and 2006 have been stored in the database for Kalam station.

0
10
20
30
40
50
60
70
80
90
100
R
a
i
n

(
m
m
)
1
/
1
/
2
0
0
4
3
/
1
/
2
0
0
4
5
/
1
/
2
0
0
4
7
/
1
/
2
0
0
4
9
/
1
/
2
0
0
4
1
1
/
1
/
2
0
0
4
1
/
1
/
2
0
0
5
3
/
1
/
2
0
0
5
5
/
1
/
2
0
0
5
7
/
1
/
2
0
0
5
9
/
1
/
2
0
0
5
1
1
/
1
/
2
0
0
5
1
/
1
/
2
0
0
6
3
/
1
/
2
0
0
6
5
/
1
/
2
0
0
6
Day of year
Daily Rainfall (mm) Recorded at Kalam Station (2004 and 2005)

Graph 3.4: Daily rainfall recorded at base station Kalam in 2004 and 2005

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
23
3.5.3. Snow Depletion Curve

A depletion curve of the snow coverage is the curve showing how the snow cover in each zone is
depleted temporally. It can be interpolated from periodical snow cover mapping so that the daily
values can be read off as an important input variable to SRM. Together with temperature and
precipitation data, such depletion curves enable SRM to simulate runoff in a past year. The
depletion curves in zone7 and zone8 refer to the glaciers and permanently snow-covered areas, so
that is why the values do not change by time.


Graph 3.5: Depletion curves of the snow coverage for 8 elevation zones of the basin Kalam,
derived from the MODIS imagery Zone1: 1991 - 2490 m a.s.l., Zone2: 2491 - 2990 m a.s.l.,
Zone3: 2991 - 3490 m a.s.l., Zone4: 3491 - 3990 m a.s.l., Zone5: 3991-4490 m a.s.l, Zone6: 4491
- 4990 m a.s.l, Zone7: 4991-5490 m a.s.l, Zone5: 5491-5990 m a.s.l.

3.6 Input Parameters

Input parameters are used for the calibration of SRM model. These parameters include
runoff coefficient for snow, and rain, temperature lapse rate, degree day factor, lag time,
rainfall contributing area (RCA), and recession coefficients.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
24
3.6.1. Runoff Coefficient, C

The difference between the available water volume (snowmelt + rainfall) and the outflow from
the basin are the losses due to evaporation, Evapo-transpiration, interception and infiltration.
Runoff coefficient takes care of such losses. Comparison of historical precipitation and runoff
ratios provide a starting point for the runoff coefficient. Daily precipitation values recorded by a
station inside the basin multiplied by the total area of the basin and then dividing the recorded
daily discharge at the outlet of the basin by the results gives the runoff coefficient. Of the SRM
parameters, the runoff coefficient appears to be the primary candidate for adjustment if a runoff
simulation is not at once successful. The river discharge and meteorological data were provided
by the Pakistan Meteorological Department and Pakistan Water and Power Development
Authority (WAPDA). No recorded information about runoff coefficients is available about the
study area.

In the study area losses are supposed to be high in the start of the melting season because out of 8
elevation zones 7 zones are still covered with snow, which may capture most of the precipitation
on it. In the middle of the melting season the losses are dependent on weather conditions, which
mean both low and high losses are possible. Towards the end of the snowmelt season, direct
channel flow from the remaining snowfields and glaciers may prevail in some basins which lead
to a decrease of losses and to an increase of the runoff coefficient. Moreover, c is usually
different for snowmelt c
S
, and for rainfall c
R
.

Graph 3.6: Runoff coefficients for rain and snow used for the model calibration for the melting season
(April-August) 2004.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
25
3.6.2. Degree Day Factor:

The degree day factor is used to compute the snowmelt depths. It increases as the snow becomes
older because of the reduction in albedo of the snow. The degree day factor for snow can be
determined using snow density measurements since they are good indicators for the albedo
(Rango 1995). No data about the snow density in the area are available.

The degree factor was considered for the study area as low in the start of the melting season and
high at the end (Graph 3.7). Because the start of the melting season (April-May) receives most of
the precipitation, and the temperatures extrapolated from base station to altitudes higher than
4000m a.m.s.l are below critical temperature (2 C), which increases the chances of summer
snow fall and hence low snow densities.

Degree Day Factor Values, Kalam Basin
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
26-Mar 15-Apr 5-May 25-May 14-Jun 4-Jul 24-Jul 13-Aug 2-Sep
Dat e
Zones 1-4 (2004,2006)
Zones 5-8 (2004, 2006)
All Zones 2005

Graph 3.7: Degree day factor values against the months of the melting season 2004, 2005 and
2006 for Kalam basin, Pakistan.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
26
3.6.3. Temperature Lapse Rate:

Temperature decreases with the increase in altitude above mean sea level with a certain rate
called temperature lapse rate. It can be predetermined from historical data if temperature stations
at different altitudes are available. Otherwise one has to go to field for some duration to install
temperature sensors at different altitudes and record the temperature data to get the lapse rate.
The same was done for the current study as no permanent stations at different altitudes are
available in the study area or in the neighbourhoods.

3.6.3.1. Fieldwork

During the fieldwork, eight temperature sensors were installed at an altitude difference of 100
meters. The data was collected for four weeks, but due to unskilled manpower only last 8 days
temperature recorded data was reliable. The sensors were installed in different villages inside the
houses. Care was taken in installing the sensors under a shadow but due to winds and other
reason few were exposed to direct sun light. This resulted in high temperatures recorded by the
sensors. Therefore only minimum temperatures were used for lapse rate calculations (Table 3.1).
After analyzing the data, the average temperature lapse rate was determined as 0.63 per 100 m
change in altitude which is close to 0.65 proposed by the developers of the SRM model. This
figure (0.63) is used for extrapolating the temperature values of the base station to higher
altitudes.
Table 3.1: Temperature data recorded during fieldwork

Lower
Liakot
(1830m)
Kass
Aryanai
(1910m)
Bishmal
Port
(1960m)
Kalam
Tehsil
(2016m)
Gaheel
(2100m)
Utror
(2229m)
Gibrall
(2300m)
Chorat
(2340m)
Date T
min
(C) T
min
(C) T
min
(C) T
min
(C) T
min
(C) T
min
(C) T
min
(C) T
min
(C)
9/18/2007 13.6 14 15.3 12.5 13.6 9.2 12.9 11.5
9/19/2007 13.1 13.3 14.4 12 13.1 9.1 13.1 10.9
9/20/2007 12.9 12.7 14.2 11.5 12.3 13.4 10.9 10.6
9/21/2007 12.8 12.5 12.8 11 11.5 12.7 12.6 10
9/22/2007 11.3 10.6 11.2 9 9.7 9.2 9.2 7
9/23/2007 10.4 8.7 9.9 8.5 8.7 10.5 8.1 6.9
9/24/2007 10.4 12.3 9.5 8.3 8.1 11.7 7.5 7.1
9/25/2007 10.4 7.5 9 7.7 7.7 14.1 6.4 7.1
9/26/2007 9.7 8.3 9.9 8.1 7 14 7.1 6.5
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
27

3.6.4. Critical temperature, T
CRIT


The critical temperature determines whether the measured or forecasted precipitation is rain or
snow. SRM needs the critical temperature only in the snowmelt season (unless a year round
computer run is made) in order to decide whether precipitation immediately contributes to runoff
(rain), or if T < T
CRIT
, whether snowfall took place.

Unfortunately event (snow or rain) dependent temperature data is not available in Kalam basin;
however, a nearby station in the west of the study area, namely Drosh has historical records of
hourly temperature values against the events of snow or rain. The data reveals that snow event
has mostly occurred at temperatures below or equal to 2 C in the winter season i.e. from
October to March. No snowfall has been recorded by the station in the melting season, making it
difficult to know the critical temperature for the melting season. Hence compromise on
considering 2 C (April-June) and 0.75 C (July, August) as critical temperature has been made.

3.6.5. Rainfall contributing area, RCA

When precipitation is determined to be rain, it can be treated in two ways. In the initial situation
(option 0), it is assumed that rain falling on the snowpack early in the snowmelt season is
retained by the snow which is usually dry and deep. Rainfall runoff is added to snowmelt runoff
only from the snow-free area, that is to say the rainfall depth is reduced by the ratio snow-free
area/zone area. At some later stage, the snow cover becomes ripe (the user must decide on which
date) and the computer program should be switched to option 1, this was done for the current
study on 1
st
June and thereafter. Now, if rain falls on this snow cover, it is assumed that the same
amount of water is released from the snowpack so that rain from the entire zone area is added to
snowmelt. For this study 0 was selected from April to May, and 1 from June-August.

3.6.6. Recession coefficient, k

The recession coefficient is an important feature of SRM because (1-k) is the proportion of the
daily meltwater production which immediately appears in the runoff. Analysis of historical
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
28
discharge data is usually a good way to determine k. Values of Q
n
and Q
n+1
are plotted against
each other and the lower envelope line of all points is considered to indicate the k-values as
shown in the figure below (Graph 3.6).

Graph 3.8: Recession flow plot Qn Vs Qn+1 for Kalam basin in Pakistan. Either the solid lower
envelope line or the dashed medium line is used to determine k-values for computing the
constants x and y.

Based on the relation
1 n
n
Q
k
Q
+
= (3.4)
In the study area
1
k
= 0.85 for
n
Q
= 200 m3s-1 and
2
k
= 0.90 for
n
Q
= 30 m3s-1. This means
that
k
is not constant, but increases with the decreasing river discharge according to the
equation:
1
y
n n
k x Q

+
=
(3.5)

Where the constants x and y must be determined for a given basin by solving the equations:
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
29
1 1
y
k x Q

=
(3.6)
2 2
y
k x Q

=
(3.7)
1 1
log log (log ) k x y Q =
(3.8)
2 2
log log (log ) k x y Q =
(3.9)

Putting values of
1
k
,
2
k
and
1
Q
,
2
Q
in equations (3.8) and (3.9)

log(0.85) log( ) log(200) x y =
(3.10)
log(0.90) log( ) log(30) x y =
(3.11)

x
= 0.997
y
= 0.030

Care should be taken in finding the recession flow. It was experienced in this study that when
recession coefficients were calculated considering the discharge data from the middle of the
recession period, the resulted recession coefficients were giving completely strange runoffs.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
30
4. GIS AND REMOTE SENSING
4.1 Software Used

The SRM model needs daily snow cover as input variable. Daily remote sensing data from
MODIS on board Terra and Aqua satellites were processed using ERDAS and Arc GIS. Exact
and accurate boundary of the catchment is very important as this is used for the masking of
satellite image of the study area. Also the SRM recommend dividing the study area into different
elevation zones if the difference in the base station and the highest altitude is greater than 500
meters. Furthermore, measurements of snow cover in different elevation zones require a DEM as
input. Extraction of catchment/basin of the study area was done using SRTM DEM data with the
help of Integrated Land and Water Information System (ILWIS).
4.2 Digital Elevation Model (DEM)

The Shuttle Radar Topography Mission (SRTM) obtained elevation data on a near-global scale
to generate the most complete high-resolution digital topographic database of Earth. Elevation
data of the area from SRTM was used for the extraction of catchment boundaries above Kalam
river discharge measurement station on river Swat (Figure 4.1).

DEM hydro-processing operations are a set of tools which can lead to the extraction of
catchment as well as its boundary using digital elevation data. Those tools which were used in
this study are fill sink, flow direction, flow accumulation, drainage network extraction, drainage
network ordering, catchment extraction and catchment merge. Detail discussion on each of these
tools and their out put results has been given below.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
31

Figure 4.1: DEM of the study area obtained from Shuttle Radar Topography Mission (SRTM)
4.2.1. Fill Sink
Before using the Flow Direction operation, clean up of local depressions from DEM data is of
utmost importance. This is done by the Fill Sink operation which performs the following on a
Digital Elevation Model (DEM):
Any pixel with a smaller height value than all of its 8 neighbouring pixels, will be
increased in height to the smallest value of its 8 neighbour pixels
Any group of adjacent pixels where the pixels that have smaller height values than all
pixels that surround a depression will be increased to the smallest value of a pixel that is
both adjacent to the outlet for the depression, and that would discharge into the initial
depression.
The resulting output map of the Fill Sink operation is a so-called sink-free or depression-free
DEM. This means that for every pixel in the DEM, a flow direction will be found towards the
edges of the map.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
32
4.2.2. Flow Direction
In a (sink-free) Digital Elevation Model (DEM), the Flow Direction operation determines into
which neighbouring pixel any water in a central pixel will flow naturally.
Flow Direction is calculated for every central pixel of input blocks of 3 by 3 pixels, each time
comparing the value of the central pixel with the value of its 8 neighbours. The output map
contains flow direction as N (to the North), NE (to the North East), etc.
4.2.3. Flow Accumulation
The Flow accumulation operation performs a cumulative count of the number of pixels that
naturally drain into outlets. The operation can be used to find the drainage pattern of a terrain.
As input the operation uses the output map of the flow direction operation.
The output map contains cumulative hydrologic flow values that represent the number of
input pixels which contribute any water to any outlets; the outlets of the largest streams,
rivers etc. will have the largest values.
4.2.4. Drainage Network Extraction
The Drainage Network Extraction operation extracts a basic drainage network (Boolean raster
map). The output raster map will show the basic drainage as pixels with value True, while other
pixels have value False.
The output raster map of the Flow Accumulation which is required as input map. This map
contains a cumulative drainage count for each pixel.
4.2.5. Drainage Network Ordering
The Drainage network ordering operation:
Examines all drainage lines in the drainage network map
Finds the nodes where two or more streams meet
Assigns a unique ID to each stream in between these nodes, as well as to the streams that
only have a single node
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
33
The output of this operation is a raster map, a segment map and an attribute table that all use a
newly created ID domain.
The output maps and the attribute table of the Drainage network ordering operation are used as
input in many other DEM-hydro processing operations, among others:
The Catchment extraction operation
The Catchment merge operation
4.2.6. Catchment Extraction
The Catchment Extraction operation constructs catchments; a catchment will be calculated for
each stream found in the output map of the Drainage Network Ordering operation.
As input is required:
The output raster map of the Drainage Network Ordering operation
The output raster map of the Flow Direction operation.
As output a raster map, a polygon map and an attribute table are produced which all use the ID
domain of the input Drainage Network Ordering (Figure 4.2).
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
34

Figure 4.2: Catchment extraction using Drainage network ordering and Flow direction as inputs.
4.2.7. Catchment Merge
The Catchment Merge operation is able to merge adjacent catchments, as found by the
Catchment Extraction operation. In fact, new catchments will be created on the basis of the
Drainage Network Ordering map and its attribute table.
As input is required:
The output map and attribute table of the Drainage Network Ordering operation
The output map of the Flow Direction operation
The output map of the Flow Accumulation operation
Merge catchments can be done by specifying a point map that contains locations of stream
outlets within a catchment; all adjacent catchments that drain into such outlets will be merged.
As output a new catchment raster map, polygon map and attribute table are produced. These all
use a new ID domain. A point map containing one point (catchment outlet) taken at Kalam river
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
35
discharge measurement station was selected to merge all catchments of the drainages that drain
into this discharge measurement station into a bigger catchment (Figure 4.3).

Figure 4.3: Extracted catchment above Kalam station.
4.3 Remote Sensing of Snow Cover
Snow cover is an important variable for climate and hydrologic models due to its effects on
energy and moisture budget (Najafi Eigdir 2003). Remote sensing is a valuable tool for snow
cover mapping which can be used for predicting snowmelt runoff. From a remote sensing
perspective, snow cover is one of the most readily identifiable measures of water resources from
aerial photography or satellite imagery (Engman 1991).
4.4 Moderate Resolution Imaging Spectroradiometer (MODIS)

With the launch of MODIS in December 1999, a new era in hyperspectral satellite remote
sensing began. MODIS makes it possible to monitor the environment by measuring atmospheric
trace gases and aerosol density, and mapping the surface of clouds, land and sea in a variety of
spectral ranges from the blue to the thermal infra-red.
The first MODIS sensor went into orbit with the launch of the TERRA satellite on December 18,
1999. With the successful launch of AQUA from Vandenberg Air Force Base, CA, on May 4,
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
36
2002, a second MODIS sensor was put into orbit for studying the Earth's water cycle and our
environment. TERRA and AQUA (both with a 705km orbit) have a sun-synchronous, near polar,
circular orbit. AQUA will cross the equator daily at 1:30 p.m. as it heads north (ascending mode)
in contrast to TERRA, which crosses the equator at 10:30 a.m. daily (descending mode). With
this formation it is expected that AQUA's afternoon observations combined with TERRA's
morning observations will provide important insights into the daily cycling of global
precipitation and ocean circulation.
MODIS is a 36 band spectrometer providing a global data set every 1-2 days with a 16-day repeat
cycle. The spatial resolution of MODIS (pixel size at nadir) is 250m for channel 1 and 2 (0.6m -
0.9m), 500m for channel 3 to 7 (0.4m - 2.1m) and 1000m for channel 8 to 36 (0.4m -
14.4m), respectively.
Satellite images are made by combining the reflected light detected by the sensor at various
wavelengths (spectral bands) and making them into a single image. The MODIS Rapid Response
System makes use of MODIS broad range of spectral observations by creating both true-color
and false-color images, each tailored to highlight different land surface, atmospheric, and oceanic
features. One such band combination is 721. In this composite, MODIS Bands 7, 2, and 1, are
assigned to the red, green, and blue portions of the digital image. Vegetation appears bright green
and bare soil red, and water appears dark black in this combination leaving snow light blue very
prominent and hence making it easy to distinguish snow from all three classes i.e. vegetation,
bare soil, and water.
MODIS Rapid Response System provides on its website a number of image subsets (Figure 4.4)
that are automatically generated in near-real-time for various applications users.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
37

Figure 4.4: MODIS subset of Afghanistan covering parts of Pakistan including the study area of
Kalam. Grey white color is showing snow and ice.
The archive imagery is available online; images of the area for 2004, 2005, and 2006 have been
downloaded from MODIS rapid response system web site (http://rapidfire.sci.gsfc.nasa.gov/subsets).

Georeferencing information is available on the same web site, and those have been used to for
georeferencing. Projection: Plate Carree and ellipsoid: WGS84. Georeferencing has been done
using ERDAS image processing software. To see the snow cover the images should be cloud
free. Normally one can get around 140-150 cloud free images throughout the year, and for four
years the total number of images to be processed for snow extraction was up to 600.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
38

Figure 4.5: Subset of Afghanistan has been geo-referenced and re-projected to UTM co-ordinates.
Extraction of daily snow cover in the study area has been done in a number of steps using both
ERDAS and ArcGIS image processing softwares.

The MODIS jpeg format images were selected for this study inspite of their limitation due to the
following reasons;
1. MODIS snow cover products were available but with 500m pixel resolution, which is
poor as compared to 250m jpeg images.
2. There was a possibility of getting HDF file format by ordering to MODIS Science Data
Support Team, but the file size was too large, and so were the number of images required
for the study. Secondly they have removed all 250m resolution HDF files from their data
base and they were recreating each file, which needed a lot of time to wait.
3. Georeferencing were available with jpeg images and therefore, it was relatively easy to
carry out the georeferencing.
4.5 Extraction of Study Area Subset

It is clear to see that the subset Afghanistan is covering a large area, but our study area is too
small. This means that subset of study area should be extracted from the whole subset before
going to snow cover mapping. A lot of time on image processing is saved this way. The merged
catchment polygon is used for masking only the image of the study area (Figure 4.6). Extract by
Mask toll in ArcGIS extracts the cells of a raster that correspond with the areas defined by a
mask, which in the present case is the polygon in the form of merged catchment.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
39

Figure 4.6: Satellite image of study area on the right after masking it from Afghanistan subset.

4.6 Classification of images for extraction of snow

After all the subsets of the study area, they are classified into snow and snow free classes (Figure
4.7). Both supervised and unsupervised methods have been applied. Classification has been done
in ERDAS.

Figure 4.7: Classified image of the study area. Blue color is showing snow whereas dark brown is
snow free.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
40
4.7 Extraction of Snow for Each Elevation Zone

Snow cover mapping is the most laborious job in this research, because daily satellite images are
downloaded and if the study area is cloud free then it is processed for the extraction of snow
cover. This leads to processing of almost 400 satellite images for 3 years i.e. 2004-2006. MODIS
images are available in jpeg format as subsets of different regions of the world. The following
steps are involved in getting snow cover from an image.
1. Downloading MODIS subsets (Afghanistan) from the rapid fire website.
2. Rectification of the satellite image using the available corner coordinates from info file
with the image.
3. Classification of each image into snow and snow free classes.
4. Converting the classified image into polygons.
5. Extraction of DEM by masking the snow class.
6. Extraction of snow covers for each elevation zone, using DEM which is classified into
different elevation zones.

Figure 4.8: Main steps for getting snow cover using ERDAS and Arc GIS tools

These have been discussed in detail one by one.

4.7.1. Extract by Attribute

After classification of the image of the catchment into snow and snow-free classes, only snow
class is extracted from the image using the option extract by attribute tool. Extract by attribute
extracts the cells of a raster based on a logical query e.g. where Class_name = Snow.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
41

Figure 4.9: Extract by attribute tool and its result. Grey color is showing snow whereas white is
snow-free area.

4.7.2. Resample

The extracted snow class map has a 250 m spatial resolution, whereas the DEM is in 90 m spatial
resolution. It is important to extract DEM covered only with snow on a particular day, because
the extracted DEM then can be used to extract snow for each elevation zone. That is why
resample tool is used. Resample tool alters the proportions of a raster dataset by changing the cell
size. The cell size will be changed but the extent of the raster will remain the same.

4.7.3. Extract by Mask

Now the resampled image can be used to extract DEM covered only with snow. Extract by Mask
tool in ArcGIS extracts the cells of a raster that correspond with the areas defined by a mask,
which in this case is the resample image after extraction of snow. This way DEM where snow
exists is extracted and which will be used for extracting snow in each elevation zone.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
42

Figure 4.10: Extract by mask tool and the resulted DEM of the area covered with snow. White
color is no-data (snow-free).
4.7.4. Reclassify

It is recommended in the SRM model to divide the catchment into a number of zones if the
elevation range is more than 500 meters between the streamgauge and the highest elevation in the
catchment. Elevation zones should be delineated in intervals of about 500 meters. The present
study area has the elevation range 1991-5790 m.a.s.l. Therefore the whole catchment has been
divided into 8 elevation zones with 500 meters interval.


Figure 4.11: Snow class in each elevation zone in raster format. Lowest height starts from Zone1
and ends at Zone8. Zone1 has no snow in the month of May.

This step is followed by conversion of the above raster format into polygons to make the area
calculation easy. Each polygon has a unique ID based on the elevation zones. Area calculator is
used to quantify the area of snow in each elevation zone. This process is repeated for all the
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
43
images available for the study. The snow cover area is converted into fraction of snow cover in a
particular zone before feeding it to the model. Snow depletion curves have been obtained for the
snowmelt season in 2004, 2005, and 2006 (Graph 4.1, 4.2, 4.3).


Graph 4.1: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2004.


Graph 4.2: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2005.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
44

Graph 4.3: Snow Deletion Curves of Snowmelt Seasons from April to August in the year 2006.


Figure 4.12: Sequence of snow cover maps from MODIS, Upper Swat River at Kalam, 3032 km2
, 1991- 5790 m a.s.l. Blue is snow covered area.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
45
5. Model Results
5.1 ASSESSMENT OF THE MODEL ACCURACY
5.1.1. Accuracy criteria

The SRM computer program includes a graphical display of the computed hydrograph and of the
measured runoff. A visual inspection shows at the first glance whether the simulation is
successful or not. SRM additionally uses two well established accuracy criteria, namely, the
coefficient of determination, R
2
, and the percent volume difference, Dv.

The coefficient of determination is computed as follows:
( )
( )
2
'
2
1
2
1
1
n
i i
i
n
i
i
Q Q
R
Q Q
=
=

(5.1)
Where:
i
Q
is the measured daily discharge

'
i
Q
is the computed daily discharge

Q
is the average measured discharge of the given melting season

n
is the number of daily discharge values
The deviation of the runoff volumes,
v
D
, is computed as follows:

[ ]
( )
'
% 100
R R
v
R
V V
D
V

=
(5.2)
Where:
R
V
is the measured melting season runoff volume

'
R
V
is the computed melting season runoff volume

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
46
5.2 Simulations of SRM

Daily runoff simulations for the snow melting season of 2004, 2005, and 2006 were carried out
after calibrating the SRM model for the year 2004. The model requires seven parameters, three
input variables and catchment / basin characteristics for its snowmelt runoff simulation. The
catchment receives heavy snowfall in winter (Dec to March) which melts in the summer.
However not all the snow is melted in the region as the catchment has permanent snow cover and
glaciers as well.

5.3 Model Calibration
The model results were calibrated using 2004 discharge data at Kalam station. The following
parameters were used in the calibration procedure. The best fit with measured discharge was
obtained using the following values. The simulations show over shouting as well under shooting
through out the melting season. Since the lapse rate considered constant through out the melting
season this may have caused, variations with changing weather conditions are possible. Also it is
not possible to check whether the average temperature represented the actual temperatures or
not? The runoff coefficients can also cause these discrepancies keeping in view their spatial and
as well as temporal variability nature.

Graph 5.1: Runoff simulations for the melting season 2004 of Kalam basin. The dashed line is
the measured discharge at Kalam station whereas the solid line is simulated runoff.

The statistics of this simulation show that the simulation is good enough both for daily as well as
for the whole melting season. Accuracy criteria (discussed in section-5.1) used in the model
gives volume difference D
v
= -0.7% and the coefficient of determination R
2
is 0.74 (Figure 5.1).
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
47

Figure 5.1: Statistics of the daily simulations for the melting season (April-August) 2004.
5.4 SRM Verification for 2005 and 2006
Once the model is calibrated it should be applicable to another year only changing the
precipitation (rain), snow cover, and temperature values for that year without changing the
parameters used in the calibrated year. This was done for the years 2005 and 2006 snowmelt
season with the same parameters used in 2004 simulations. For 2005 the results for Q are given
in Figure 5.2. For 2005 Volume Difference D
v
= -16.1% whereas the coefficient of determination
R
2
= 0.54 (Figure 5.2). It is important to mention that there was one drawback in the hydrological
data of the year 2006, as not all the daily discharge values were recorded on the measuring
station on the river. Only data of 4 to 7 days in a month was provided. Therefore, the computed
runoff values were compared only with the measured runoff values and calculations were done
for R
2
and D
v
, with R
2
= 0.83 and D
v
= 0.05% (Graph 5.3).


Graph 5.2: Runoff simulations for the melting season 2005 of Kalam basin. The dashed line is
the measured discharge at Kalam station whereas the solid line is simulated runoff.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
48

Figure 5.2: Statistics of 2005 melting season simulations following accuracy criteria

Model Computed Discharge Against the Measured Discharge
Melting Season (April-August), 2006
0
100
200
300
400
500
600
700
Mar-06 Apr-06 May-06 May-06 Jun-06 Jul-06 Jul-06 Aug-06
Date
R
i
v
e
r

D
i
s
c
h
a
r
g
e

(
c
u
b
i
c

m
e
t
e
r

p
e
r

s
e
c
o
n
d
)
Measured
Computed

Graph 5.3: Daily runoff simulations for melting season, 2006. The triangular points measured
discharge values whereas the gray circular points are computed values.

Year 2006 was more similar to year 2004 than 2005. Snowfall in the winter season of 2006
remained slightly below the snowfall during winter season 2004, which means far below 2005
(30 years record snowfall). Temperatures on the average during the melting season were closer to
2004 as compared to 2005; although at the end of the melting season they become closer to 2005
temperature values. Hence year 2004 parameters were used for 2006 melting season simulations,
and the results were better than 2004 simulations, and hence the model was verified for 2006.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
49
5.5 Discussion on model simulations

In SRM simulations of 2005, the accuracy criteria show worse results with R
2
= 0.54. Therefore,
the runoff coefficients and the degree day factor had to be changed, keeping the lower and upper
limits of these parameters well within the extreme values used during the calibration year 2004,
which means that runoff coefficients and degree day factor do not remain constant temporally i.e.
from year to year. Year 2005 received 30 years record snowfall in the winter (Pakistan
Meteorological Department) and that was followed by a heat wave in summer (NASAs Earth
Observatory Report, Southern Asia Heat Wave). Heavy snow fall means more snow available for
the melting season and for a much longer time. In fact snow that remains for a longer time is
denser and hence a higher degree day factor is needed as discussed earlier. Similarly if in a
particular year there are below average rains or above average high temperatures then the
catchment will be dry and vegetation will be stressed. MODIS land surface temperature (LST)
data and normalized difference vegetation index (NDVI) data both confirms the above argument
for 2005 (Figure 5.3).


Figure 5.3: MODIS Land surface temperatures and NDVI data showing high temperatures and
stressed vegetation in 2005 in comparison to 2004 where temperatures are low and vegetation is
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
50
healthy. Red color in upper half of the figure is showing high temperatures and dark green in
lower half is showing healthy vegetation.
Air temperature data recorded at Kalam station also confirms that temperatures on the average in
June-July for year 2005 were higher by 2.7 C as compared to year 2004 (Graph 5.5). At the
same time precipitation data revealed that there were longer dry spills in 2005 in comparison to
2004. As a result whenever the runoff occurred, there were more losses due to infiltration as well
as water utilization by plants to recover the deficiency caused during the dry period. These were
included in the model by adjusting the runoff coefficients and degree day factor.

Graph 5.4: Same period temperatures comparison between 2004 and 2005
After adjusting these parameters the simulations were fitting better, than the previous simulations
where the parameters of 2004 were used (Graph 5.6, and Figure 5.4).

Graph 5.5: Runoff simulations for the melting season 2005. R
2
=0.90 and D
v
=3.2%
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
51

Figure 5.4: Statistics of the daily simulations for the melting season 2005

5.6 Forecasts using WinSRM

SRM can be used for short term (for example weekly) forecasts of daily flows as well as for
longer time period forecasts such as monthly runoff volumes or seasonal runoff volumes. For
short term forecasts, temperature, precipitation and snow covered area must be forecasted or
predetermined for the coming days and substituted into the model. Temperature and even
precipitation forecasts are becoming increasingly available from meteorological services, but the
snow covered areas must be extrapolated by the model user.

New simulation was added to the model of 7 days duration from 09/01/2005 to 09/07/2005. This
simulation was used for weekly forecast (1st week of September 2005). Instead of forecasted
values for temperature and rainfall recorded data was used for the same duration. Also snow
cover data of august 2005 was extrapolated for next 7 days. The model was run for weekly
forecast in the melting season. The forecasts of runoff have been found quite satisfactory with
volume difference D
v
= 2.1% and Coefficient of Determination R
2
=0.80 (Graph 5.7).

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
52

Graph 5.6: Weekly forecast of runoff in the Kalam catchment, 1st week of Aug, 2005

In the above forecast, recorded data of temperature and precipitation have been used just to
emphasize the importance of accurate forecasts of temperature and precipitation, because the
forecast of input variables is still an important challenge for all snowmelt runoff models.

The model can be used for monthly and seasonal forecast as well, but the sensitive part will be
how to get the modified snow depletion curves for those periods. If there are proper historical
records of yearly snow accumulation and depletion, these can be averaged and then used for
runoff forecasts. However careful study about similar / different years is needed in this approach.

Suppose there is a record of 10 years of snow accumulation and snow depletion about a
particular catchment. Some of these may have had heavy snowfall while some may have had
below average snowfall during these 10 years. Similarly in some years there may be more runoff
(both in volume and frequency) while some may had the opposite. Thus different categories can
be formed and used for matching with any coming years to get forecast for runoff. Forecast the
melting season 2005 gives promising result with volume difference D
v
= -1.2% and coefficient of
determination R
2
= 0.95 (Graph 5.8).

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
53

Graph 5.7: Melting season runoff forecast for the year 2005
5.7 Sensitivity Analysis
The SRM model was tested for the sensitivity of its parameters. Almost every parameter was
increased or decreased by 10% of its original value used in the calibration, and results of
simulation were noted down. There was a considerable change noted in the values of R
2
and D
v
,
when temperature lapse rate was decreased or increased by 10%. Apart from lapse rate the
degree day factor and runoff coefficient for snow (C
s
) were having the same effect on both R
2

and D
v
. It is therefore concluded SRM is more sensitive to temperature lapse rate. However
degree day factor and runoff coefficient for snow should not be ignored for best results. Original
run results are with R
2
=0.90 & D
v
=2.6%.

Table 5.1: Sensitivity analysis results
S.No Parameter Name R
2
D
v

Lapse Rate + 10% of Lapse Rate 0.76 24.1%
1
Lapse Rate + 10% of Lapse Rate 0.67 -21.9%
Critical Temperature + 10% of Tcrit 0.90 2.6%
2 Critical Temperature - 10% of Tcrit 0.89 2.6%
Degree Day Factor (a) + 10% of a 0.84 -6.38%
3 Degree Day Factor (a) - 10% of a 0.89 11.6%
Cs + 10% of Cs 0.84 -6.4%
4 Cs - 10% of Cs 0.89 11.7%
Cr + 10% of Cr 0.90 2%
5 Cr - 10% of Cr 0.90 3.3%
Lag time 16 hrs 0.90 2.5%
6 Lag time 20 hrs 0.90 2.6%
RCA = 1 0.91 1.1%
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
54
7 RCA = 0 0.90 3.9%
5.8 Effect of Distorted Depletion Curve

Depletion curves derived for the melting season 2005 were used for this analysis. Some times a
long period of cloud cover brings occasional summer snowfall. Due to this when a satellite image
of the following day is used for snow extraction it distorts the snow depletion curve. Snow
depletion curves were distorted intentionally by adding large values in it. When the model was
run on distorted depletion curve as input, it was noted that when the distortion is at the end of the
snowmelt season it causes much more excessive meltwater than in the start or the middle of the
melting season (Figures 5.5, and 5.6). Especially when the depletion curve of a certain zone was
ending (all the snow was near to melt), the effect was more significant (Figure 5.6).



Figure 5.5: Depletion was distorted in the middle of the melting season



Figure 5.6: Depletion curve was distorted at the end of the depleted curve.


SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
55
6. Conclusion and Recommendations
6.1 Conclusion

River Swat is snow fed river, starting with tributaries Ushu, and Utror, meeting in Kalam. Hence
assessment of snowmelt in Kalam basin is very important to understand the behaviour of River
Swat. Although snow in the basin is an important water resource, it causes flooding in the lower
reaches of river Swat. When it is melted abruptly Flooding due to over flow in the melting season
is not only the problem in Swat district but also in Charsada, Peshawar and Noshehra districts
where river Swat meets the River Kabul and onwards. Very few studies have been found in
Pakistan on this topic. The River Kabul catchment has been studied in 1975 (section 2.3), but so
far there is no study about river Swat.

Most of the precipitation in Kalam basin is in the form of snow in the winter season (Dec to
March), followed by a bit more rainy season from April to May. The snow cover has been
mapped using data from the MODIS sensor onboard the Terra and Aqua satellites. Different band
combinations are available free of charge for scientific purposes. So far studies on snow cover
mapping for running SRM model has been done mostly using NOAA-AVHRR satellite images
with spatial resolution of 1km. This research has the upper edge in the snow cover mapping on
the previous studies because MODIS 250m spatial resolution data has been used for snow cover
mapping instead of 1km NOAA images.

Snowmelt runoff model (SRM) by Martinec 1975 has been used for this study. The model needs
snow extent, precipitation in the form of rain and temperature data as input variables, and some
catchment parameters for calibrating the model.

After getting the necessary data for the running the model, several simulations in the snowmelt
season have been carried out for different years i.e. 2004, 2005 and 2006. Graphical display and
statistics can tell about the accuracy of the models simulations. The model seems to be
satisfactory the values of D
v
and R
2
are within the acceptable range. The model has also been
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
56
used in this study for forecasts of runoffs, and if the correct inputs are available the model can
give good results in forecasts and can be used for flood forecasts in the lower reaches.

Results have proved that:

The SRM model can be calibrated for the simulation of long term runoff volume for the
Kalam catchment.
The model is verified for 2006, whereas for 2005 slight adjustment of the runoff
coefficients and the degree day factor was needed before verification.
The values of D
v
and R
2
also prove it to be good for daily runoff simulations.
When tested for runoff forecasts the model was found to give good predictions.
The model results in the snowmelt seasons were very good in term of the volume of
water from the snowmelt in a melting season. Hence the values of D
v
prove the model to
be very good for long term water resource management.
It has been noticed that the model is more sensitive to the temperature lapse rate. The
degree day factor and runoff coefficient are found to be second to the temperature lapse
rate, to which SRM is more sensitive.
The runoff coefficient for rain and snow are important parameters of the basin, as these
tells about how much fraction of available water due to rain or melting of snow reaches
the river discharge station.
To avoid any error in recession coefficients, one should consider the whole recession
period of the year, not a part of recession period. This was experienced during this
research that when the recession coefficients were measured using only part of the
recession period the simulations were very strange. But when the whole recession period
i.e. July to August, was considered for the measurement, the resulted simulations were
quite logical and close to reality.
SRTM has been very useful in getting the catchment boundary and in the snow cover
estimation in different elevation zones of 500m intervals.
It is believed that the parameters used for calibration of the model should remain
constant for the same season in different years, however this study has proved that heat
waves and below average rainfall, or less frequent rainfall can affect the runoff
coefficients from year to year. Heat waves and below average rains can make the
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
57
vegetation stressed and hence can cause much loss of available rain water or snowmelt
water.
Degree day factor which is dependent on snow density has also been found slightly
different in 2005 from 2004 and 2006. The reason is that there was a record 30 years
heavy snowfall in 2005 which means that snow will remain for much longer time and
hence will be denser. This lead to increased values of degree day factor for 2005, as
compared to 2004 and 2006.

6.2 Recommendations

The SRM model is a degree day model, and is more sensitive to temperature values at different
altitudes. There is only one meteorological station in Kalam basin, which is in the lowest altitude
level. This compelled to extrapolate the base stations temperatures to higher altitudes with a
constant rate.

Moreover critical temperatures need at least half hourly precipitation data both in the form of
rain or snow against recorded temperatures.

Fieldwork is important for lapse rate, with at least one sensor in each elevation zone. This is only
possible if one can install recording sensors to avoid requirement of a number of skilled
manpower. This could make the performance of the model better.

Therefore there should be at least one automatic meteorological station in each elevation zone, if
one would like to carry a proper study and avoid assumptions.

The use of satellite surface temperature and rainfall data in SRM should be investigated.
Incorporation of these data could improve the insight into the spatial distribution and its variation
of the corresponding parameters, and their optimal quantification for the snowmelt runoff
modelling.

Similarly runoff coefficients for snow and rain can also be linked with Land surface temperatures
(LST) and normalized difference vegetation index (NDVI) with some limitations due to cloud
cover and unavailability of continuous coverage of the area through satellite.
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
58
7. References:
Anderson, E. A. (1973). National Weather Service Forecast System - Snow accumulation and
ablation model, NOAA Technical Memorandum NWS. HYDRO-17.
Bergstrm, D. (1975). "The development of a snow routine for HBV-2 Model." Nordic
Hydrology 6(2).
Burnash, R. J. C., Ferral, R. L. and McGuire, R. A. (1973). A generalized streamflow simulation
system: Joint Federal-State River Forecast Center, Sacramento, California, 204 pp
Charbonneau, R., Fortin, J. P. and Morin, G. (1977). "The CEQUEAU model: description and
examples of its use in problems related to water resources management." AIHS 22(1): 193-203.
Dey, B., V. K. Sharma, and A. Rango (1989). "A Test of Snowmelt Runoff Model for a Major
River Basin in Western Himalayas." Nordic Hydrology 20: 167-178.
Engman, E. T. a. R. J. G. (1991). Remote Sensing in Hydrology, Chapman and Hall.
Gotleib, L. (1980). A general runoff model for snowcovered and glacierized basins. 6th Nordic
Hydrological Conference Proceedings, Vemdalen, Sweden.
Jaroslav Martinec, A. R. R. R. (2005). SRM USERS MANUAL. New Mexico State University,
USDA.
KERR, T. (2005). SNOW STORAGE MODELLING IN THE LAKE PUKAKI CATCHMENT,
NEW ZEALAND: AN INVESTIGATION OF ENHANCEMENTS TO THE SNOWSIM
MODEL. Department of Geography, University of Canterbury. Master of Science.
Leavesley, G. H., Lichty, R. W., Troutman, B. M. and Saindon, L. G. (1983). Precipitationrunoff
modelling system: User's manual. Water Resources Investigations Report. U. S. G. Survey.
Martinec, J. (1975). "Snowmelt-Runoff Model for stream flow forecasts." Nordic Hydrol. 6(3):
145-154.
Martinec, J., Rango, A. & Major, E. (1983). "The Snowmelt-Runoff Model (SRM) User's
Manual. NASA Reference Publ. 1100, Washington, D.C., USA."
Martinec, J. R., A. (1989). "Merits of statistical criteria for the performance of hydrological
models." Wat Resour. Bull. 25(20): 421-432.
Najafi Eigdir, A. (2003). Investigation of the snowmelt runoff in the Orumiyeh region, using
modelling, GIS and RS techniques. Enschede, ITC: 54.
Quick, M. C. a. P., A. (1976). "A combined snowmelt and rainfall runoff model." Canadian
Journal of Civil Engineering 3(3): 449-460.
Rango, A. M., J. (1988). Results from international intercomparisons of snowmelt runoff model
performance. 45th Annual Eastern Snow Conference, Lake Placid, New York, USA.
Rango, A. M., J. (1995) "Revisiting the degree-day method for snowmelt computation." Water
Resource Bull. Volume, 657-669 DOI:
SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
59
Russell, J. S. (2003). Comparison of watershed models. Perspectives in Civil Engineering,
American Society of Civil Engineers: 347-349.
Shafer, B. A., Jones, E. B. & Frick, D. M. (1981). Snowmelt Runoff Simulations Using the
Martinec-Rango Model on the South Fork Rio Grande and Conejos River in Colorado.,
AgRISTARS, Goddard Space Flight Center, Greenbelt, MD, USA.
Turcan, J. (1981). Empirical-Regressive Forecasting Runoff Model. VUVH, Bratislava.
WMO (1986). Intercomparison of Models of Snowmelt Runoff. . Operational Hydrol. Geneva,
Switzerland., WMO.


SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
60
Appendices
Appendix 1: Average temperatures at Kalam station for the melting season 2004.













SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
61

Appendix 2: Average temperatures at Kalam station for the melting season 2005.







SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
62

Appendix 3: Average temperatures at Kalam station for the melting season 2006.

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
63

Appendix 4: Measured Vs Computed River Discharge (m
3
s
-1
) 2004 (Melting Season), Kalam Basin

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
64

Appendix 5: Measured Vs Computed River Discharge (m
3
s
-1
) 2005 (Melting Season), Kalam Basin

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
65

Appendix 6: Measured Vs Computed River Discharge (m
3
s
-1
) 2006 (Melting Season), Kalam Basin

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
66

Appendix 7: Precipitation data for 2004, Kalam station
Daily Precipitation (Millimeters) Year - 2004
STATION: KALAM
Province: North West Frontier Province
Country: Pakistan
Date Apr May Jun Jul Aug
1 0 3.8 13 0 0
2 0 0 0 0 1.3
3 19.1 0 0 4.1 0.3
4 0 0 0 0 0
5 0 0 0 0 0
6 0 0 0 0 0
7 3.8 0 0 0 0
8 0 17.8 0 0 0
9 11.4 10.2 0 1.5 4.6
10 10.2 0 0 0 0
11 0 0 0 0 6.9
12 0 0 0 0 0
13 3.8 0 0 0 0
14 0 0 0 2.5 0
15 0 0 3.8 4.1 0
16 0 3.8 0 10.2 0
17 0 0 0 0 0
18 3 0 0 0 0
19 0 0 0 2.5 3
20 29.2 0 0 0 0.3
21 0 0 0 1.3 0
22 10.9 0 0 0 0
23 19.1 0 0 2.5 0
24 0 0 0 0 0
25 0 0 0 0 7.6
26 0 4.1 1.3 0 0
27 34.3 7.6 1.3 0 0
28 0 0 0 0 0
29 21.6 0 0 0 0
30 3.8 0 0 0 1.3
31 0 0 0.5

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
67

Appendix 8: Precipitation data for 2005, Kalam station
Daily Precipitation (Millimeters) Year - 2005
STATION: KALAM
Province: North West Frontier Province
Country: Pakistan
Date Apr May Jun Jul Aug
1 0 0 0 0 0
2 0 0 0 0 3.8
3 0 0 0 0 0
4 0 4.1 0 0 0
5 0.3 3.3 0 1.5 0
6 50.8 0 0 0 0
7 34.3 0 0 0 0
8 9.7 4.1 0 0 4.8
9 2.5 15.2 0 0 5.6
10 0 0 0 0 0
11 0 0 4.6 1.5 0
12 0 0 10.2 0 0
13 0 0 0 7.9 8.4
14 38.1 0 0 9.9 0
15 0 0 0 0 0
16 0 0 0 0 0
17 0 0 0 0 13.2
18 0 0 0.5 0 0
19 0 1.8 0 0 0
20 0 1.3 0 0 0
21 0 0 0 0 0
22 0 0 0 0 0
23 0 35.8 0 0 0
24 15 8.1 0 0 0
25 43.2 6.9 0 0 0
26 31 25.4 0 0 0
27 5.3 0 0 0 0
28 0 0 0 0 0
29 0 0 0 0 0
30 0 7.6 1 0 0
31 10.7 0 0

SNOWMELT RUNOFF INVESTIGATION IN RIVER SWAT UPPER BASIN USING SNOWMELT RUNOFF MODEL
68

Appendix 9: Precipitation data for 2006, Kalam station
Daily Precipitation (Millimeters) Year - 2006
STATION: KALAM
Province: North West Frontier Province
Country: Pakistan
Date Apr May Jun Jul Aug
1 21 8.0 0 0 0
2 3 0 0 10.0 0
3 0 0 0 0 0
4 0 0 0 0 0
5 0 0 3.0 0 0
6 0 0 0 0 0
7 0 0 0 0 0
8 0 0 0 0 0
9 20 0 3.0 0 0
10 53 0 0 0 0
11 0 0 0 0 0
12 0 0 5.0 7.0 0
13 0 0 0 11.0 0
14 0 0 0 5.5 0
15 20.0 0 0 0 0
16 0 0 0 0 7.0
17 0 0 0 0 0
18 0 0 0 0 0
19 0 6.0 0 0 0
20 0 0 0 0 0
21 8.0 0 0 0 0
22 0 2.5 0 0 2.0
23 0 0 4.0 0 4.6
24 0 0 0 0 2.0
25 0 0 0 0 0
26 0 0 0 0 0
27 0 0 0 0 0
28 0 0 0 0 0
29 0 0 0 0 1.0
30 0.5 0 0 2.0 0
31 0 0.5 0

You might also like