Post on 30-May-2015
description
Northern Gulf Institute - GeoSystems Research Institute
Mississippi State University
Scale-dependency and Sensitivity of Hydrological Estimations to Land Use
and Topography for a Coastal Watershed in Mississippi
Vladimir J. Alarcon and Charles G. O’Hara
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Introduction
• The spatial variability of the physical characteristics of the terrain influences the flow regime in watersheds.
• Through Internet, physiographic datasets can be easily downloaded for geo-processing.– Topographical and land-use/land-cover datasets
are used for the purposes of watershed delineation, land use characterization, geographical positioning of hydro-chemical point sources, etc.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Introduction
• Several previous researches have explored the sensitivity of hydrological estimations to topographical or land use datasets.
• Few have assessed the sensitivity of hydrological estimations to combined swapping of topographic and land use datasets.– Objective: This paper explores the effects of using
several different combinations of topographical and land use datasets in hydrological estimations of stream flow.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Study area:
– Two river catchments• Jourdan River
– Drains 88220 ha with an average stream flow of 24.5 m3/s.
• Wolf River – Drains slightly more than
98350 ha with an average stream flow of 20.1 m3/s.
• Major contributors of flow to the St. Louis Bay, MS, USA.
Jourdan River Catchment
Wolf River Catchment
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Topographical data:
– USGS DEM • 3 arc-second (or 1:250,000-scale) approx. 300 m spatial res.
– The National Elevation Data (NED) dataset • 1 arc second, approximately 30 m spatial res.
– NASA Shuttle Radar Topography Mission• 30 m spatial resolution near-global topography data product for
latitudes smaller than 60 degrees.
– Intermap’s Interferometric Synthetic Aperture Radar (IfSAR) Digital Elevation Models 1:24,000 scale topographic quadrangle map series, 5 m spatial res.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Land Use data:
– USGS GIRAS• Minimum mapping unit of 4 hectares and a maximum of 16
hectares (equivalent to 400 m spatial resolution).
– National Land Cover Data (NLCD)• 21-class land cover classification scheme, spatial resolution of the
data is 30 meters.
– The NASA MODIS MOD12Q1 Land Cover Product (MODIS/Terra Land Cover
• 1000 m spatial resolution, classified in 21 land use categories, covers most of the globe and is updated every year.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Geo-processing
DTED format conversion to ArcView raster (.BIL)
Mosaicing of appropriate data cubes
Delineation
Generation of the mosaiced grid
Conversion: binary floating point grid to
Arcinfo grid (floatgrid)
Mosaicing of appropriate data cubes
Pre-Delineation
Conversion to "integer grid“ with elevation in
centimeters.
Original data: floating point binary grid format
Mosaicing and clipping of the region of interest
Original data: DTED format
Optional clipping of the resulting grid
Joining of resulting polygons
Elimination of overlapping areas
SRTM IFSAR
Delineated watershed
Delineated watershed
LAND USETOPOGRAPHY
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Data processing: all the datasets were geo-processed for input into the Hydrological
Simulation Program Fortran (HSPF).
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Hydrological Modeling
– Hydrological Simulation Program Fortran (HSPF). • Simulation of non-point source watershed hydrology and
water quality. • Time-series of meteorological/water-quality data, land
use and topographical data are used to estimate stream flow hydrographs and polluto-graphs.
• The model simulates interception, soil moisture, surface runoff, interflow, base flow, snowpack depth and water content, snowmelt, evapo-transpiration, and ground-water recharge.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods• Processing steps for
generating the hydrological model’s data input.
• BASINS, GEOLEM, ArcGIS, and Arc Info were used.
• These GIS systems extracted land use and topographical information for input into HSPF.
• HSPF was launched from within the BASINS system.
BASINSWatershed delineation
GEOLEM/ArcGISLand use and topographical
datasets parameterization for HSPF feed
WinHSPF/GenScnGeneration and calibration of
HSPF applications
starter.uci(base file for generation of
scenarios)
*.WSD(watershed file for
generation of scenarios)
Delineated watershed(*.shp file)
Land use datasets(NLCD, GIRAS,
MODIS)
Topographical datasets
(SRTM, IFSAR, NED, DEM)
WinHSPFScenarios simulation
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Methods
• HSPF models for Jourdan and Wolf watersheds.
• USGS stream flow gauge stations at: – Catahoula (Jourdan
River)
– Lyman, and Landon (Wolf River)
were used for hydrological calibration of the models.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Results• Results of statistical fit between simulated and measured
stream flow for Jourdan River watershed
Model fit efficiency (Nash-Sutcliff, NS)
MODIS (1000 m) GIRAS (400 m) NLCD (30 m)
USGS DEM (300m) 0.75 0.74 0.75
NED (30m) 0.73 0.72 0.72
SRTM (30m) 0.73 0.72 0.72
IFSAR (5m) 0.74 0.73 0.73
Coefficient of determination R2
MODIS (1000 m) GIRAS (400 m) NLCD (30 m)
USGS DEM (300m) 0.8 0.8 0.79
NED (30m) 0.77 0.78 0.76
SRTM (30m) 0.77 0.77 0.76
IFSAR (5m) 0.78 0.78 0.77
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
NLCD (30 m)
0.72
0.725
0.73
0.735
0.74
0.745
0.75
0.745-0.75
0.74-0.745
0.735-0.74
0.73-0.735
0.725-0.73
0.72-0.725DEM
(300m)NED(30m)SRTM
(30m)IFSAR(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.765
0.77
0.775
0.78
0.785
0.79
0.795
0.8
0.795-0.8
0.79-0.795
0.785-0.79
0.78-0.785
0.775-0.78
0.77-0.775
0.765-0.77
0.76-0.765
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.74
0.75
0.76
0.77
0.78
Model reliability coefficient
0.77-0.78
0.76-0.77
0.75-0.76
0.74-0.75
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.77
0.78
0.79
0.8
0.81
0.82
0.83
Nash-Sutcliff, NS, model fit efficiency
0.82-0.83
0.81-0.82
0.8-0.81
0.79-0.8
0.78-0.79
0.77-0.78
0.76-0.77 DEM (300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.81
0.815
0.82
0.825
0.83
0.835
0.84
Coefficient of determination (R^2)
0.835-0.84
0.83-0.835
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.8
0.805
0.81
0.815
0.82
0.825
0.83
Model reliability coefficient
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
0.805-0.81
0.8-0.805
JourdanRiver Watershed
Wolf River Watershed
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
NLCD (30 m)
0.72
0.725
0.73
0.735
0.74
0.745
0.75
0.745-0.75
0.74-0.745
0.735-0.74
0.73-0.735
0.725-0.73
0.72-0.725DEM
(300m)NED(30m)SRTM
(30m)IFSAR(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.765
0.77
0.775
0.78
0.785
0.79
0.795
0.8
0.795-0.8
0.79-0.795
0.785-0.79
0.78-0.785
0.775-0.78
0.77-0.775
0.765-0.77
0.76-0.765
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.74
0.75
0.76
0.77
0.78
Model reliability coefficient
0.77-0.78
0.76-0.77
0.75-0.76
0.74-0.75
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.77
0.78
0.79
0.8
0.81
0.82
0.83
Nash-Sutcliff, NS, model fit efficiency
0.82-0.83
0.81-0.82
0.8-0.81
0.79-0.8
0.78-0.79
0.77-0.78
0.76-0.77 DEM (300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.81
0.815
0.82
0.825
0.83
0.835
0.84
Coefficient of determination (R^2)
0.835-0.84
0.83-0.835
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.8
0.805
0.81
0.815
0.82
0.825
0.83
Model reliability coefficient
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
0.805-0.81
0.8-0.805
JourdanRiver Watershed
Wolf River Watershed
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
NLCD (30 m)
0.72
0.725
0.73
0.735
0.74
0.745
0.75
0.745-0.75
0.74-0.745
0.735-0.74
0.73-0.735
0.725-0.73
0.72-0.725DEM
(300m)NED(30m)SRTM
(30m)IFSAR(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.765
0.77
0.775
0.78
0.785
0.79
0.795
0.8
0.795-0.8
0.79-0.795
0.785-0.79
0.78-0.785
0.775-0.78
0.77-0.775
0.765-0.77
0.76-0.765
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.74
0.75
0.76
0.77
0.78
Model reliability coefficient
0.77-0.78
0.76-0.77
0.75-0.76
0.74-0.75
DEM(300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.76
0.77
0.78
0.79
0.8
0.81
0.82
0.83
Nash-Sutcliff, NS, model fit efficiency
0.82-0.83
0.81-0.82
0.8-0.81
0.79-0.8
0.78-0.79
0.77-0.78
0.76-0.77 DEM (300m)NED
(30m)SRTM(30m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400 m)
NLCD (30 m)
0.81
0.815
0.82
0.825
0.83
0.835
0.84
Coefficient of determination (R^2)
0.835-0.84
0.83-0.835
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
DEM(300m)NED
(30m)SRTM(30 m)IFSAR
(5m)
MODIS (1000 m)
GIRAS (400)
NLCD (30 m)
0.8
0.805
0.81
0.815
0.82
0.825
0.83
Model reliability coefficient
0.825-0.83
0.82-0.825
0.815-0.82
0.81-0.815
0.805-0.81
0.8-0.805
JourdanRiver Watershed
Wolf River Watershed
.)(
)(
1
1
2
1
2
n
jjj
n
jjj
OO
SO
NS
GIRAS (400 m)
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Results• Results of statistical fit between simulated and measured
stream flow for Wolf River watershed
JourdanRiver Watershed
Wolf River Watershed
JourdanRiver Watershed
Wolf River Watershed
JourdanRiver Watershed
Wolf River Watershed
Model fit efficiency (Nash-Sutcliff, NS)
MODIS (1000 m) GIRAS (400 m) NLCD (30 m)
USGS DEM (300m) 0.82 0.76 0.81
NED (30m) 0.81 0.81 0.8
SRTM (30m) 0.81 0.81 0.8
IFSAR (5m) 0.82 0.82 0.81
Coefficient of determination R2
MODIS (1000 m) GIRAS (400 m) NLCD (30 m)
USGS DEM (300m) 0.83 0.83 0.82
NED (30m) 0.82 0.82 0.82
SRTM (30m) 0.82 0.82 0.82
IFSAR (5m) 0.83 0.83 0.82
.)(
)(
1
1
2
1
2
n
jjj
n
jjj
OO
SO
NS
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Conclusions• Scale and spatial resolution of digital land use and topography
datasets do not affect the statistical fit of measured and simulated stream flow time-series, when using the Hydrological Program Fortran (HSPF) as the hydrological model recipient of land use and topographical data.
• When the input to HSPF comes from the coarsest spatial resolution datasets (MODIS land use and USGS DEM topography), HSPF simulated stream flow show equivalent statistical fit to measured stream flow as when finer spatial resolution datasets are combined.
• This suggests that stream flow simulation is not sensitive to scale or spatial resolution of land use and/or topographical datasets.
ICCSA 2010 Conference, March 23-26, 2010, Fukuoka, Japan
Conclusions• Since HSPF is a lumped-parameter hydrological model, the
summarization of physiographic information (before it is input to the model) is a required step.
• This pre-processing of raw land use and topographical data (re-classification, averaging of data per sub-basin, etc.) may be a factor that explains the insensitivity of simulated stream flow to land use and topographical datasets swapping, when using HSPF.
• Future studies should explore the effects that this swapping may have when using distributed hydrological models.