Hydrological Network Modelling GEOG1002 Dr P. Lewis.

Hydrological Network Modelling

GEOG1002

Dr P. Lewis

Interest in part of hydrological cycle

Precipitation:clouds to ground

Flow: vertical - infiltrationhorizontal - surface runoff

Components of river forecast model:

• Baseflow: – the amount of water coming from groundwater.

• Runoff: – the amount of water coming from surface

runoff.

• Routed Flow: – the amount of water coming from an upstream

point.

Examine here hydrological network model

• required output:required output: – flow at a point / time

• inputs: inputs: – hydrological network – runoff

• require for: require for: – flood modelling/prediction (water flow through

network) – routing of water in Global Climate Models

GCMs

• Models of global energy – inputs (solar radiation) – outputs (longwave radiation) – transfers (e.g., atmosphere, ocean fluxes) – state (prediction)

GCMs

• Models concentrate on vertical fluxes

• poor modelling of hydrological routing– essentially dump excess in

nearest ocean grid cell – doesn’t give time lag for travel

• cant easily relate to measurement

Model for hydrological routing in GCM:

• simple

• fast

• validated

• Describe model of Naden (1992)

• Information sources for networks

Naden (1992)

• Model grid to point via river channel network

• can validate at series of points

• Require data on river network – topology – cross-section – slope/speed

Network Response Function:

• define Network Width Function (NWF) – no. of 'links' upstream from a point

DISTANCE

LIN

KS

define Routing Function

• function of stream velocity (slope) and cross-section

• describes time lag due to friction in channel

• RF introduces delay into system through convolution with NWF

Routing Functionre

spon

se

time

Routing Function

Network Response Function (NRF)

• 'flash' (impulse) of water onto system - measure NRF

• essentially same 'shape' and NWF, but smoother

• depends strongly on time lag in RF

CALCULATION OF NETWORK RESPONSE FUNCTION FROM NWF

• NRF calculated from NWF and RF using velocity (A m/s) and diffusion (D m2/s) coefficients

RoutingFunction

(parametersA and D)

TIME

RE

SP

ON

SE

=

NRF

*LIN

KS

DISTANCE

NWF

Convolution

• powerful mathematical tool for linear systems

• sum of weighted contributions over a moving window

convolve

distance

alti

tude

with

distance

alti

tude

Has

effect of

SMOOTHING

CALCULATION OF PREDICTED RIVER FLOWS• Disaggregated precipitation generated by a mesoscale

model

• Hydrological model used to produce ‘generated runoff’ from precipitation, where ‘generated runoff’ is that portion of precipitation which enters the channel network

*

TIME

RE

SP

ON

SE

NRF

TIME

FL

OW

PREDICTEDFLOW

=

TIME

RUNOFF

RU

NO

FF

• now have model of predicted flow based on:– network topology (NWF) – cross-section. speed/slope (RF) – runoff

• can 'validate' model by comparing predicted/modelled flows at point

• model is simple enough for GCMs

• requires data to define NRF

Consider data sources

• need to be global

• range of sources available: – digital network data ('blue line')

• topological structuring

• network links structured to flow downhill

• removal of braids and lakes

– derive from DEMs • models to derive networks from DEM e.g. GIS model used

in ARC/INFO

• can use to define catchment

Data sets

DATA SET SCALE SOURCEDCW Blue Line 1:1 000 000 USDMA/ESRI1:50k Blue Line 1:50 000 Ordnance Survey/IHDCW DEM 1km UCL (ANUDEM)EDC DEM 1km EDCDTED DEM 1 1km Military (MCE Feltham)DTED DEM 2 100m Military (MCE Feltham)OS/IH DEM 50m Ordnance Survey/IHIFSAR DEM 30m UCL 3DIMGTOPO30 DEM 1km USGS EDCHYDRO1K Blue Line 1km USGS EDCRIVER REACH 1:250 000 EPA

Case studies

Severn and Thames ~10 000 km2 each

Data Set Scale /Resolution

Thames Severn

1:50k Blue Line 1: 50 000 0.89 0.87DCW Blue Line 1: 1 000 000 0.88 0.85DTED Level 1 100 m 0.88 0.83DTED Level 2 1 km 0.86 0.74EDC 1 km --- 0.74GTOPO30 1 km 0.86 0.68Summed Runoff (Not routed) 0.72 0.50

(Higher values are better, approaching 1.00)

EFFICIENCY OF FITUK Predicted Flows vs Observed Flows

Efficiency of Fit 1 QO QE 2

QO QO 2QO = Observed Flow

QE = Estimated Flowwhere:

Summary• hydrological flow modelling important for routing

of water – e.g. in GCM – simple, fast, validate – Naden 1992 flow to a point in network – define:

• NWF • RF • NRF = NWF * RF • flow = runoff * NRF

Summary

• model requires network data– various available, DCW etc. – variable quality – can derive network from DEM

• result dependent on quality/resolution of DEM

• need accurate high resolution DEM globally (satellites)