Modelling the Origin of Channel Flow using Autonomous Software Agents Sim Reaney University of Leeds.

Modelling the Origin of Channel Flow using

Autonomous Software Agents

Sim ReaneyUniversity of Leeds

The Problem

Most models will tell you how much water leaves a catchment as discharge

It is possible to have the same discharge with water being sourced in very different places Spatial equifinality

GIS analysis can locate areas of high runoff generation

What is not know is the fate of the generated runoff or the source of the channel water

The Need

What is needed is a model that can: Trace the flow of water from source to fate Operate at hydrogically and management scales

of interest Temporal scale relating to storms Spatial scale of a catchment

Published Approaches

There have been a number of models published that trace flow

These have been at various scales Rill growth in plots River channels Landscapes

RillGrow

The RillGrow models of Favis-Mortlock (1998) and Favis-Mortlock et al. (2000) simulates the flow paths of raindrops across a soil surface.

Consider the generation of rills at the spatial scale of up to 2.2 metres by 2.2 metres.

Raindrops are inserted into the model and their flow path is determined by the surface energy gradient the location of surface depressions

Ephemeral Channels

Shannon et al. (2002) have developed a model of ephemeral channel flow based on parcels of water.

Water routing is determined either by either the channel bed topography by a stochastic routing algorithm

This model is able to simulate the flow within a channel reach This sets it apart from other approaches due to

the far greater spatial scale

‘Precipitons’ Chase (1992) developed a landscape evolution model of:

fluvial erosion transportation deposition

Operated at the spatial scale of hillslopes to mountain ranges. The surface is subjected to a number of localised storms, termed

‘precipitons’, routed across the surface. erode the surface and transport a slope limited amount of material.

It was found that this approach could simulate complex landscapes.

Problems… All of these models are limited in resolution

Spatial or Temporal Both

In area of extent This is due to the large memory requirements

of the simulations To model the origin of discharge, detail is

required on both temporal and spatial scales

The Solution… Software agents that follow hydrological rules Agents are able to move over a surface The agent then uses probability theory to

decide its movement The life history of each agent is stored and

can be analysed using GIS

hydroAgents

Memory Demands Solution

Separation of the environment generator and the agent population

The environment for the agent is generated by a hydrological model

Each agent has zero mass The agent does not influence with the mass

balance of the main model

System structure

CRUM-2DHydrological Model

hydroAgentpopulation

Puddle depth

Flow routing

Flow velocity

Infiltration rate

Depression store depth

Output flow paths andlife details of the

agents

hydroAgent Rule Set

The decisions are based on probability theory conditioned on the local hydrogical environment

At each model iteration, the agent follows the following rules: Does the agent infiltrate?

If not, Does the agent leave the cell?

If so, Which direction does the agent move in?

Agents in action

Investigation of flow paths for a simple slope and storm Rainfall of 75 mm hr-1 for 5 minutes Slope of 50 by 50 metres at a gradient of 6° Infiltration characteristics of semi-arid bare scrub One metre spatial resolution One second time step

5,000 agents introduced per minute Initial position of the slope chosen at random

Products

Distribution of flow path lengths for the storm

Area directly connected to the outflow

flow path length (metres)

0 2 4 6 8 10 12 14 16

freq

uenc

y

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

cum

mul

ativ

e fr

eque

ncy

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Contributing Area

Flow Convergence

0 20 Meters

Real Catchments, Real Rainfall

In September 1997 there was a high magnitude storm event in the Rambla de Nogalte catchment, SE Spain

Large spatial differences in the amount of channel flow

hydroAgents can be used to determine the origin of the channel flow

Considered two contrasting catchments Cardena Sub7

Cardena

Sub 7

Sub7 Catchment

Lower Cardena Catchment

September 1997 Storm Event

date

27/09/97 28/09/97 29/09/97 30/09/97 01/10/97

rain

fall

inte

nsity

(m

m h

r-1

)

020406080

100120140160180200

time

09:30:00 09:50:00 10:10:00 10:30:00

rain

fall

inte

nsity

(m

m h

r-1

)

020406080

100120140160180200

Discharge Hydrograph

time (seconds)

0 2000 4000 6000 8000

rain

fall

(mm

hr

-1)

0

50

100

150

200

disc

harg

e (m

3 s-1

)

0

2

4

6

8

rainfalllower Cardena Qsub 7 Q

Lower Cardena – Expansion

Lower Cardena - Contraction

Sub 7 - Expansion

Sub 7 - Contraction

Results

The simulations predict that the main contributing areas for the lower Cardena are: the steep slopes at the lower end of the

catchment the upper sections of a number of gullies.

In Sub 7, the channel flow was predicted to be sourced from: the channels and the lower sections of the slopes directly adjacent

to the channel.

Extensions

Improved spatial representation Improved DEM

Lidar Development of different rule sets for

Chemical modelling Sediment tracing

Other Applications

Source areas for water pollutants Model movement of nitrates through a catchment Rather than inserting the hydroAgents uniformly

across the landscape, relate to the pattern of fertiliser application and deposition

sheepAgents Model the movement of livestock and the

production of fecal material Then able to model the movement of the fecal

material in the hydrological environment

Conclusions

Source areas for channel flow are important for management These can be mapped with the agent based

approach presented The application of agent based modelling

techniques to hydrological problems can give insights not possible with traditional, continuous representation hydrological models