Вы находитесь на странице: 1из 56

Geospatial Stream Flow Model (GeoSFM)

USGS FEWS NET EROS Data Center Sioux Falls, SD 57198

U.S. Department of the Interior U.S. Geological Survey

Objectives
 To develop a model for wide-area flood hazard monitoring using existing geospatial datasets  To use the model to routinely monitor flood hazards across Africa and provide early warning to decision makers

GIS IN FLOOD MONITORING


 The Mid-West Floods of 1993 SAST Database  Creation of Global Elevation Datasets for hydrologic modeling in 1997  Initiation of GIS-based distributed flood modeling at the USGS in the late 1990s;  Now being applied in Southern Africa, East African and proposed for Mekong River tributaries 

Model Overview
 Leverage the vast geospatial data archived at EDC Initial parameters derived from existing datasets Input data generated daily from available datasets  Catchment scale modeling framework Semi-distributed hydrologic model Inputs aggregated to the catchment level  GIS based Modeling Takes advantage of existing spatial analysis algorithms Includes integration with external routing codes

FEWS Flood Risk Monitoring System Flow Diagram


GIS Preprocessing
Satellite Rainfall Estimates Rainfall Forecasts GDAS PET Fields FAO Soil Data Land Use/ Land Cover Elevation Data
Dist. Routing

Stream Flow Model

GIS Postprocessing

Water Balance

Lumped Routing

http:/www.fews.net

Flood Inundation Mapping

Stage Forecasting

Geospatial Stream Flow Model, An ArcView 3.2 Extension

Using Menus,Message Boxes and Tools

Hydrograph plotting tool

Tool for Dam Insertion

Model Components
 Terrain Analysis Module  Parameter Estimation Module  Data Preprocessing Module  Water Balance Module  Flow routing Module  Post-processing Module

Terrain Analysis Module

The goal of Terrain Analysis


 to divide the study area into smaller subbasin, rivers  to establish the connectivity between these elements  to compute topography dependent parameters

Using ArcViews Terrain Analysis Functions with USGS 1 km DEM

Flow Direction Flow Accumulation Flow Length Hill Length Slope Subbasins Downstream Subbasin

Key Lessons from Terrain Analysis


 Procedures for Terrain Analysis have been refined over the last decade, and they work very well  USGS 1km DEM (Hydro1k) is sufficient for delineation in most basins; it is currently being refined for trouble areas

Parameter Estimation Module

The goal of Parameter Estimation


 to estimate surface runoff parameters in subbasins  to estimate flow velocity and attenuation parameters  to summarize parameters for each subbasin

Estimating Surface Runoff Characteristics


 Initially computed on a cell by cell basis  Now moving towards generalizing land cover and soil class over subbasin first

(Maidment (Ed.), 1993, Handbook of Hydrology) (Chow et al, 1988, Applied Hydrology)

Overland Velocity with Mannings Equation


 Initially computed on a cell by cell basis  Now moving towards generalizing land cover and slopes class over subbasin first

V = (1/n) * R2/3 * S1/2

Weighted flow length and aggregation algorithm to create Unit Hydrographs

Overland Velocity, Flow Time

Aggregate cells at basin outlet During each routing interval

Flow Path, Flow Length

li t ! i !1 vi

Key Lessons in Parameterization


 While GIS routines work well, existing parameter tables in hydrology textbooks are only of limited utility  There is no on-going effort to document parameters from previous studies though these are often extremely useful  Uniform parameter estimates are often at least as good spatially distributed parameters; simpler is better  Field observations and local estimates are invaluable

Data Preprocessing Module

The goal of Data Processing


 to convert available station & satellite rainfall estimates into a common format  to set up ascii files for water balance and flow routing models to ingest

Interpolation routines to grid point rainfall data


Grids adhere to a naming convention which allows for subsequent automation

Gage Data

Daily Grids

Zonal algorithms to compute subbasin mean values and export to an ASCII files
Rain / Evap Grid Output to ASCII File

Subbasins

Key Lessons in Data Preprocessing


 Using a single rainfall value for each subbasin is consistent with the resolution/precision of the satellite rainfall estimates  Saving data values in ASCII files (instead of directly assessing the grids) speeds up subsequent flow routing computations considerably

Water Balance Module

The goal of Water Balance


 to separate input rainfall into evapotranspiration, surface, interflow, baseflow and ground water components  to maintain an accounting of water in storage (soil moisture content) at the end of each simulation time step

Conceptual Model of Water Balance

Two Water Balance Options


 Single layered soil with
Hortonian bucket with partial contributing areas Single subsurface reservoir but different residence times for interflow and baseflow

 Two layered soil with


SCS Curve Number Method Separate reservoirs and residence times for interflow and baseflow

Partitioning Fluxes in single layered model


Rainfall

Soil layer

Hortonian with Partial Contributing Areas Saturated Hydraulic Conductivity

Ground Water

Transferring Fluxes in single layered model


Rainfall

Surface Runoff

Unit Hydrograph
Soil layer

Interflow Linear Reservoir + Baseflow Linear Reservoir

Ground Water

Partitioning Fluxes in two layered model


Rainfall

Upper layer Lower layer

SCS Curve Number Method

Green Ampt Based Parameterization

Ground Water

Transferring Fluxes in two layered model


Rainfall Surface Runoff Unit Hydrograph Upper layer Lower layer Baseflow Conceptual Linear Reservoir Interflow Conceptual Linear Reservoir

Ground Water

Key Lessons in Water Balance


 SCS Curve number classes dont correspond directly with mapped land cover / vegetation classes  Hortonian with partial areas performs at least as well and is easier to parameterize than SCS method for runoff generation  Recession portion of the hydrograph has been the most difficult to model correctly

Flow routing Module

The goal of Flow Routing


 to aggregate the runoff contributions of each subbasin at the subbasin outlet  to move the runoff from one subbasin to the next, through the river network to the basin outlet

Within subbasin routing


Apply unit hydrograph to excess runoff to obtain runoff at subbasin outlet
Runoff Water Balance

Unit Hydrograph

Channel Routing Overview


Sub-basin 1

+
Main channel Sub-basin 3

Sub-basin 2

+
Main channel

Sub-basin 4

+ Outlet

Three Channel Routing Options


 Pure Translation Routing  Diffusion Analog Routing  Muskingum Cunge Routing

Pure Translation Routing


Only parameter required is lag time or celerity Simple but surprising effective in large basins

Input

Output

Time

Time

Diffusion Analog Routing


Linear routing method Requires two parameters
Velocity for translation Diffusion coefficient for attenuation Input Output

Time

Time

Muskingum-Cunge Routing
Non-Linear, Variable Parameter routing method Accounts for both translation and dispersion

Conceptual reach sections with time varying storage

Distance along river reach

Key Lessons in Flow Routing


 The fewer parameters you have to estimate, the easier it is to obtain a representative model  The ease of developing a representative model often determines readiness of users adopt the model as much as precision of the model  Recommend the diffusion analog model for large scale applications; it achieves a reasonable balance between simplicity and process representation

Post-processing Module

The goal of Postprocessing


 to compute flow statistics (max, min, mean, 25, 75, 33, 66 and 50 percentile flow)  to rank and display current flows relative to percentile flows (high, low, medium)  to perform preliminary inundation mapping (based on uniform flow depths within each reach)  to display hydrographs where needed

Characterizing Flood Risk


Generate Daily Historical Rainfall (1961-96) by reanalysis Produce a synthetic streamflow record

Determine locations where bankfull storage Is exceeded

Compute Bankfull storage

Color-coded maps to indicate level of hazard

Hydrographs with their historical context

Nzoia Basin, Kenya

Nzoia Basin, Modeled vs Observed Streamflow

Limpopo River Basin

Olifants, Kruger National Park - Mamba

FEWS Flood Risk Monitoring System Flow Diagram


Data
RFE

Stream Flow Model Preprocessing MAP

Output / Decision Support System

QPF

Water Balance

MAE Basin Linkage


Lumped Routing

PET

Soil
Routing Parameters
LU/LC

Soil Parameters DEM

Dist. Routing

Flood Inundation Mapping

Updating

Landsat 7

SPOT

FEWS DMIP application Flow Diagram


Data Preprocessing
Radar

Stream Flow Model MAP


Water Balance

Statistical Results and Comparison

Pan Data PET Statsgo Soil

MAE Basin Linkage


Lumped Routing

LDAS IGBP

Routing Parameters Soil Parameters Dist. Routing

DEM

Model Results

DMIP basins

^
P ! P !

P !

P !
0 15 30 km
Calibration Stream Gage Interior Stream Gage Gage at Watts Blue River - 1233 sq. km Elk River- 2251 sq. km Baron Fork- 795 sq. km Illinois at Watts- 1645 sq. km Illinois at Tahlequah - 2484 sq. km
Rivers
0 40

P !

P ! P ! P !

^
! P "

^ ^

P !

80 Km

Results
Mean Monthly Flows: Baron Fork, 06/1993-07/2000
Observed Fews HRC OHD USU UWO EMC ARS 30 CMS 0 5 10 15 20 25

6 month

10

12

Key Lessons in Postprocessing


 The importance of hydrographs to decision makers is often overrated  The most important questions decision makers want answered are how many people were/will be affected, and where are they?  Hazard maps and flood maps are far better methods of commuting to decision makers than hydrographs  Estimates of affected/at risk populations and their locations are the most useful outputs of the hydrologic analysis

Conclusions
 The Geospatial Stream Flow Model (GeoSFM) is a semidistributed hydrologic model for wide-area hydrologic analysis  It uses globally available terrain, soil and land cover data, and satellite derived estimates of daily rainfall and PET  The model stream flow outputs include stream flow and flood hazard maps  Model validation in the Nzoia and Limpopo river basins were satisfactory  DMIP results in uncalibrated mode compare with results of other modeling groups

Вам также может понравиться