Transcript Geoprocessing using GEOLEM

Geoprocessing using GEOLEM
and
HSPF in the RPC Framework
Vladimir Alarcon
Chuck O’Hara
July 10, 2007
NASA quarterly briefing
GEOLEM
• Library of basic geoprocessing functions,
e.g., “flow direction”, “reclassify”
• Library of complex geoprocessing logic,
e.g., “make map of hillslopes”, “make map
of affected areas”
• Knowledge handling infrastructure
• System to encode modeling knowledge
into metadata
• Metadata handling infrastructure
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
• GEOLEM was customized to provide landuse
and topographical parameters to be ingested by
HSPF
• How was it modified?
– GEOLEM main code: changes to include new
schema files:
• config.xml
• concept.xml
• instance.xml
– New methods codes were written:
• SlopeHspfMethod.java
• LanduseHspfMethod.java
• ReclassLanduseHspfMethod.java
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
• How was it modified? (continued)
– New parameter java codes were written:
• LandUseHspfParameters.java
• SlopeHspfParameters.java
• ReclassHspfLanduseParameters.java
– New providers to parameters:
• LandUseHspfParametersProvider.java
• SlopeHspfParametersProvider.java
• ReclassHspfLanduseParametersProvider.java
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
• How was it modified? (continued)
– New commands were coded into existing
JacobCommands class:
• SlopePercent
• TabulateArea
• ReclassHSPF
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
• How was it modified? (Version without changes)
HSPF
July 10, 2007
NASA quarterly briefing
GEOLEM and HSPF
• How was it modified? (changed version)
HSPF
getParam: HSPFLanduseArea
getParam: HSPFSlope
getParam: HSPFLanduseArea
getParam: HSPFSlope
return: Area
return: Slope
getDimension: Sub_basin
return: Sub_basin
Subbasin: HSPFslope
Now the user
has the option
of re-using
EXISTING
ParamHspfSlope
ParamHspfLanduseArea
July 10,
ASCII HSPF
2007input files
Zonal Statistics: Slopes and landuse area per sub-basin
Delineated
subbasins
instead of
delineating all
over again
several times
LandUseParam.dbf
SlopeParam.dbf
NASA quarterly briefing
HSPF in RPC
Vladimir Alarcon
Chuck O’Hara
July 10, 2007
NASA quarterly briefing
How does HSPF fit into RPC?
Application
schema
User
User
UCI
Control
Input file
Watershed
boundary
polygon
GEOLEM
GEOLEM
GUI
HSPF
GenScen
WDMutil
Modified
UCI file
Delineated
watershed
polygon
Characterization --Stream
Land use
-Sub-basin Topography
User’s domain
July 10, 2007
Python
connector
Rainfall, ET, Soil
Moisture time
series
Results
G
E
O
L
E
M
Geo-processing
server
Server’s
domain
Topography
Delineated
SRTM
Watershed
Land
Land use
use
MODIS,
classes VIIRS
LIS-generated
data: Precip.,
ET, Soil Moist.
NASA quarterly briefing
HSPF in RPC
• How can HSPF be used within the RPC
environment?
– Study the effects of topographical datasets in
hydrograph
• Some results were presented in previous briefings
– Alarcon and O’Hara, 2006
– Alarcon, O’Hara et al., 2006
– Study the effects of landuse datasets in hydrograph
• Some results were presented in previous briefings
– Diaz, Alarcon, O’Hara, et al.
– For this quarterly report we have prepared what could
be a typical RPC application using Geolem HSPF
• Concurrent effects of landuse and topographical datasets on
streamflow hydrograph simulation in a coastal watershed
July 10, 2007
NASA quarterly briefing
HSPF in RPC: research question
• Question: if NASA would like to design/launch/release a
sensor/mission/product related with topographical and
landuse data, what resolution would be useful for
watershed hydrology modeling?
• A factorial experiment with existing LULC and
topography datasets has been performed.
• GEOLEM was used to generate 12 concurrent scenarios
of topographical/LULC cases for HSPF ingestion.
• HSPF was used to simulate streamflow hydrograph for
each of these 12 cases.
• Those simulated streamflow hydrographs were
compared to measured streamflow and the simulatedoutput reliability was assessed
July 10, 2007
NASA quarterly briefing
HSPF in RPC: factorial experiment
Topography\Landuse
MODIS
(1000 m)
GIRAS
(900 m)
NLCD
(30 m)
DEM (300 m)
NED (30 m)
SRTM (30 m)
IFSAR (5 m)
Statistical indicators of fit between HSPF
simulated streamflow and measured
streamflow
•Nash-Sutcliff (NS) number
•Coefficient of determination
•Model reliability coefficient:
R2
R   NS  
2 2
2
Good fit when
these coefficients
are close to 1
2
July 10, 2007
NASA quarterly briefing
HSPF in RPC: Study area
• Jourdan River:
– Located in the Saint Louis
Bay watershed (Mississippi
Gulf coast)
• Largest contributor of flow to
the Saint Louis Bay
• Drains 882 sq. km
• Average flow: 24.5 cms
Jourdan River
Catchment
July 10, 2007
NASA quarterly briefing
HSPF in RPC: NED & NLCD
July 10, 2007
(GOOD FIT)
NASA quarterly briefing
HSPF in RPC: DEM & MODIS
July 10, 2007
(BETTER FIT)
NASA quarterly briefing
HSPF in RPC:
fit between simulated and measured streamflow
0.75
0.745
0.74
0.735
0.73
0.745-0.75
0.74-0.745
0.735-0.74
0.73-0.735
0.725
0.72
MODIS (1000 m)
GIRAS (900 m)
NLCD (30 m)
IFSAR
(5m)
SRTM
(30m)
NED
(30m)
0.725-0.73
0.72-0.725
DEM
(300m)
Model fit efficiency (Nash-Sutcliff NS)
MODIS (1000 m) GIRAS (900 m) NLCD (30 m)
NLCD (30 m)
0.74
0.75
0.75
0.72
0.72
0.72
0.72
0.72
0.72
0.73
0.73
0.73
MODIS (1000 m) GIRAS (900 m)
DEM
(300m)
0.75
DEM
(300m)
0.75
0.74
NED
(30m)
0.73
NED
(30m)
0.73
0.72
SRTM
(30m)
0.73
0.72
SRTM (30m)
0.73
IFSAR
(5m)
0.74
0.73
IFSAR (5m)
0.74
July 10, 2007
NASA quarterly briefing
HSPF in RPC:
fit between simulated and measured streamflow
0.8
0.795
0.79
0.795-0.8
0.785
0.79-0.795
0.78
0.785-0.79
0.775
0.78-0.785
0.77
0.775-0.78
0.765
0.76
MODIS (1000 m)
GIRAS (900 m)
NLCD (30 m)
IFSAR
(5m)
SRTM
(30m)
NED
(30m)
0.77-0.775
0.765-0.77
DEM
(300m)
Coefficient of determination
R
2
0.76-0.765
2
Coefficient of determination R
MODIS (1000 m) GIRAS (900 m)
MODIS (1000 m) GIRAS (900 m) NLCD (30 m)
0.8
DEM (300m) DEM (300m)
0.8
0.8
0.79 0.8
0.77
NED (30m) NED (30m)
0.77
0.78
0.760.78
SRTM (30m) SRTM (30m)
0.77
0.77
0.760.77
0.77
IFSAR (5m) IFSAR (5m)
0.78
0.78
0.770.78
0.78
July 10, 2007
NLCD (30 m)
0.79
0.76
0.76
0.77
NASA quarterly briefing
HSPF in RPC:
fit between simulated and measured streamflow
0.78
0.775
0.77
0.765
0.76
0.755
0.75
0.745
0.74
MODIS (1000 m)
0.775-0.78
0.77-0.775
0.765-0.77
0.76-0.765
0.755-0.76
0.75-0.755
0.745-0.75
GIRAS (900)
NLCD (30 m)
IFSAR
(5m)
SRTM
(30 m)
NED
(30m)
DEM
(300m)
0.74-0.745
Model reliability coefficient
MODIS (1000 m) GIRAS (900)
Model reliability
DEMMODIS
(300m)
0.78
(1000 m) GIRAS
(900)
NLCD (300.77
m)
NED
(30m)
0.75
0.75
DEM (300m)
0.78
0.77
0.77
NED (30m) SRTM (300.75
0.75
0.74
m)
0.75
0.75
SRTM (30 m)IFSAR (5m)
0.75
0.75
0.74
0.76
0.76
IFSAR (5m)
July 10, 2007
0.76
0.76
0.75
NLCD (30 m)
0.77
0.74
0.74
0.75
NASA quarterly briefing
Conclusions from the experiment
• The combination of low resolution topographical datasets
(such as DEM, 300m) and low resolution landuse
datasets (such as MODIS, 1000m) produce good
statistical fit between simulated and measured
streamflow hydrographs.
• Also: the finer the topographical grid (such as IFSAR,
5m) combined with coarse resolution landuse datasets
(such as MODIS or GIRAS) seem to produce good
statistical fit.
• Medium-resolution topographical datasets(such as
SRTM or NED, 30m) combined with medium-resolution
landuse datasets (NLCD, 30 m) give the lowest
goodness of fit.
July 10, 2007
NASA quarterly briefing
Future steps
• Include simulated VIIRS in similar
topography/LULC and streamflow
hydrograph assessments
• Include distributed meteorological forcings
in the exploration:
– NASA LIS precipitation
– NASA LIS soil moisture
– NASA LIS evapotranspiration
July 10, 2007
NASA quarterly briefing