Prof. D. Nagesh Kumar
Department of Civil Engineering
Indian Institute of Science
Bangalore – 560012
URL: http://www.civil.iisc.ernet.in/~nagesh
Acknowledgement:
Drs A Anandhi , V V Srinivas &
Prof Ravi S Nanjundiah
November 22, 2013
1
Outline
Introduction
Climate Change
IPCC Scenarios
Downscaling for Hydrologic Investigations
Downscaling of Hydro-meteorological Variables for a River Basin
Remote Sensing (RS), GIS & DEM for Hydrology
RS for Land Use/ Land Cover
GIS for Watershed Delineation
DEM for Drainage Pattern Estimation using SRTM Data
Streamflow Projections using SWAT
AV SWAT Model
Inputs for AV SWAT Model
Streamflow Projections
Conclusions
November 22, 2013
3
2
Climate Change Scenario
Climate Change scenario refers to a plausible future climate that has been
constructed for explicit use in investigating the potential consequences
of anthropogenic climate change and natural climate variability
IPCC - Intergovernmental Panel on Climate Change
The 4 storylines describe the way world population, land use changes,
new technologies, energy resources, economies and political structure
may evolve over the next few decades.
November 22, 2013
3
3
Dynamics in GCM’s
Any GCM must be formulated
with some fundamental
considerations for
Conservation of momentum
Conservation of mass
Conservation of energy
Ideal gas law
Computed either in rectangular
grid space or in spectral space
November 22, 2013
4
Downscaling to River Basin Scale
(a) real world
Schematic illustration of the
(b) world as represented by GCMs
To enhance the information from GCMs
in order to bridge the gap between
what is supplied by climate models and
what is required by impacts
researchers
Small-scale affects (such as
topography) important to local
climate are poorly represented
in GCM
300km
Courtesy: The Canadian Climate Impacts Scenarios (CCIS) Project
50km
Impact
models
require ...
10km
Point
downscaling
1m
Global Climate
Models supply...
impacts
GCMs
November 22, 2013
Courtesy: The Canadian Climate Impacts Scenarios (CCIS) Project
large gap
5
F is modeled dynamically
(through RCM)
19 vertical levelsatmosphere
2.5
lat
300 km
3.75
long
F is modeled empirically from
observational (or modeled) data
Climate
Model Grid
Points
1.25 km
1.25 km
20 vertical
Levels-soil
GCM
two way
nesting
one way
nesting
RCM and GCM
run simultaneously
Feedback from
RCM to GCM
Driving RCM with
GCM
No feedback from
RCM to GCM
November 22, 2013
RCM
6
Support Vector Machine
• The SVM uses kernel functions to transform the data to a higher, possibly
infinite, dimensional space.
• A linear solution, in the higher dimensional feature space, corresponds to a
non-linear solution in the original lower dimensional input space.
• The mapped objects are linearly separable
• An optimal line is constructed rather than a complex curve to separate objects
November 22, 2013
7
Select probable predictors (PPs)
Methodology for
Downscaling using SVM
Prepare scatter plots, and compute cross-correlations (CCs) between PPs in NCEP and GCM
data sets, and between PPs in NCEP and the predictand data
Select/update thresholds Tng & Tnp
No
CCs = thresholds
Yes
Potential predictors (POPs) for downscaling
Standardize the monthly data of POPs extracted from NCEP and GCM datasets
Extract PCs and PDs from
standardized NCEP data of POPs to
prepare feature vectors (FVs)
depicting months
75% of FVs form the
training set
Calibration and verification of SVM
Model
PDs
Obtain PCs of standardized GCM
data of POPs along PDs extracted
from NCEP data to prepare feature
vectors (depicting months)
Validated SVM model
Parameters
SVM model
validation
Grid search method to find
optimum parameter range
Genetic algorithm to
determine optimum
parameter value
November 22, 2013
Comparison of downscaled
values with observed values
from past record and
estimation of error
No
25% of FVs form the
validation set
Yes
Future projections of
predictand
Is the performance
accepted
8
Malaprabha Basin
Part of Krishna Basin in
Karnataka
The climate of the study
region is dry, except in
monsoon months
Avg. annual
rainfall:1051mm
Catchment Area:2564 sq
km
GCM output is downscaled
India
Arabian
Sea
using Support Vector
Machine (SVM)
Six Cardinal variables are
downscaled
Map showing
NCEP & GCM grid points and rain gauge
November 22, 2013
locations in Malaprabha reservoir catchment
Precipitation
Maximum Temperature
Minimum Temperature
Wind Speed
9
Relative Humidity
Solar Radiation
Predictors for downscaling different
hydrometeorological variables
Sl.
No.
Predictors
(Spatial resolution of 2.5o X 2.5o
1
Temperature at 925mb, 700mb, 500mb, 200mb
≈
75,000 km2 )
Predictand
(river basin scale
≈ 2500 km2)
Precipitation
(Ta 925, Ta 700, Ta 500, Ta 200),
geo-potential height at 925mb, 500mb, 200mb
(Zg 925, Zg 500, Zg 200),
specific humidity at 925mb, 850mb
(Hus 925, Hus 850),
zonal (Ua) and meridional wind velocities (Va) at 925mb,
200mb (Ua 925, Va 925, Ua 200, Va 200),
precipitable water (prw) and surface pressure (ps).
2
Group A – Ta 925, Ua 925, Va 925
Group B – latent heat, sensible heat, shortwave
Max. & Min.
Temperature
radiation and longwave radiation fluxes
Group C – comprises of all predictors in Groups A and B
November 22, 2013
10
Predictors for downscaling different
hydrometeorological variables – contd..
Sl. Predictors
No. (Spatial resolution of 2.5o X 2.5o)
Predictand
(river basin scale)
3
Ua 925, Va 925
Wind speed
4
Ta 925, Hus 925, Ta sur, and LH
Relative humidity
5
Precipitable water
Cloud cover
6
Group A – 15 predictors, same as predictors
selected for precipitation
*Group B –precipitation, max. & min. temperature,
wind speed, relative humidity, cloud cover
Streamflow
* at river basin scale
November 22, 2013
11
95% quantile
BOX PLOT
Upper whisker
Upper Quartile (75% quantile)
Historical (observed) mean
Median (50% quantile)
Mean value of simulations
Lower Quartile
(25% quantile)
Lower whisker
5% quantile
November 22, 2013
12
Downscaling RH* from CGCM3 for different Scenarios
Past – generally
1978-2000
Indicates the mean annual
RH (historical )
Depicts the mean trend of RH projected by GCM
Predictand
value
Downscaled Downscaled
using NCEP using GCM
November 22,predictors
2013
predictors
The future 100 year simulations are divided into five 20-year periods for each of
the IPCC scenarios
13
* Theoretical and Applied Climatology, Springer, 2012
Downscaled Rainfall
from CGCM3 for different Scenarios*
* International Journal of Climatology, Royal Meteorological Society (RMetS), UK, 2008
November 22, 2013
14
Downscaled Maximum Temperature*
• Results show that Tmax is projected to increase in future for A1B, A2, and B1 scenarios.
• Projected increase in Tmax is high for A2 scenario, whereas it is least for B1 scenario.
This is because among the scenarios considered, the scenario A2 has the highest
concentration of carbon dioxide (CO2) equal to 850 ppm, while the same for A1B,
B2 and COMMIT scenarios are 720 ppm, 550 ppm and ≈370 ppm respectively.
* International Journal of Climatology, Royal Meteorological Society (RMetS), UK, 2009
November 22, 2013
15
Downscaled Minimum Temperature*
* International Journal of Climatology, Royal Meteorological Society (RMetS), UK, 2009
November 22, 2013
16
Downscaled Wind Speed
November 22, 2013
17
Downscaled Cloud Cover
Solar radiation is obtained from downscaled cloud cover and temperature data
November 22, 2013
18
Integration of RS, GIS, DEM and Hydrological Models
RS
November 22, 2013
GIS
DEM
Hydrological Model
19
Integration of RS, GIS, DEM and
Hydrological Models
• Hydrological model is a good tool for understanding and
managing phenomena related to hydrological processes
• RS provides essential inputs for hydrologic models
• GIS provides a platform for simulation of hydrological model
• DEM provides inputs essential for topography
• RS, GIS & DEM combined with mathematical models provide
a convenient platform for handling, compiling and presenting
large amounts of spatial data essential to river basin management
November 22, 2013
20
Soil and Water Assessment Tool (SWAT)
SWAT is a river basin, or watershed scale model
(Neitsch et al., 2002)
Physically based & Continuous time model
To predict the impact of land management practices on
water, sediment and agricultural chemical yields in large
complex watersheds with varying soils, land use and
management conditions over long periods of time.
Due to its easy adaptability to situations with limited data
availability, it has become very popular even to study the
climate change impact on a river basin scale
21
November 22, 2013
November 22, 2013
22
Land use/ land cover theme of the Malaprabha
reservoir catchment derived from IRS data
November 22, 2013
23
Major soil groups in the Malaprabha reservoir
catchment derived from RS and NBSS & LUP, Nagpur
November 22, 2013
24
DEM of the catchment of Malaprabha reservoir
November 22, 2013
25
Stream network in the catchment of Malaprabha reservoir
obtained from AVSWAT model using DEM
November 22, 2013
26
Drainage sub-basin outlets that are fixed
in the catchment of Malaprabha
November 22, 2013
27
Sub-basins formed by AVSWAT for the outlets fixed
in the catchment of Malaprabha reservoir
November 22, 2013
28
Longest stream in each sub-basin obtained from AVSWAT model
November 22, 2013
29
Model performance during the validation period
(1994 –2000)
350
Observed
SW AT simulated
300
P = 0.97
250
200
150
100
Jul-00
Jan-00
Jul-99
Jan-99
Jul-98
Jan-98
Jul-97
Jan-97
Jul-96
Jan-96
Jul-95
Jan-95
Jul-94
0
Jan-94
50
Observed and simulated monthly streamflows (in mm)
November 22, 2013
30
Streamflow projections made for various IPCC scenarios
for Malaprabha Basin obtained from AV-SWAT model
Streamflow in mm
November 22, 2013
31
Conclusions
• Climate change and its variability will have significant impact
on Hydrology of river basin
• It is important to understand and quantify these impacts
• Remote Sensing, GIS and DEM are very useful tools to assess
water resources of a river basin
• Impact of climate change on hydrology and water resources
should be investigated for effective planning and management of
water resources
November 22, 2013
110
32
Thank you
November 22, 2013
33