quality assessment of drainage networks extraction from a...

1

Click here to load reader

Upload: truongtuyen

Post on 22-May-2019

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Quality assessment of drainage networks extraction from a ...engees.unistra.fr/fileadmin/user_upload/pdf/cevh/Payraudeau_poster_montpellier2000... · Digital Elevation Model (DEM)

Laboratory Spatial

Structures & SystemsCemagref -ENGREF

Conclusion

Determining flow directions with the D8 procedure : advantages & weaknesses

Simple & robust method.But the flow direction discretisation in 8 possible directions separated by 45°introduces a bias on flow directions ;this bias depends on the orientation of the numerical grid.

Illustrative example of this grid bias for an area of limited relief : differences between drainage networks extracted from a DEM with different orientations

Main aim : to propose a method for evaluating the quality of drainage networks extracted

from a grid DEM with the D8 procedure

The quality assessment method proposed enables hydrological users to quantify and spatialise the quality of the drainage networks extracted from a grid DEM with the D8 procedure.This approach represents a simple and efficient means of comparing

different solutions developed to improve the extraction results, based on a modification of the DEM ; in this study, considering a reference drainage network.

Glossary :

D8 : procedure for flow directions determination, with flow assigned to one of the 8 neighbours in the direction with the steepest downward slope.DEM : Digital Elevation Model -numerical description of relief.

Contact :

[email protected]

[email protected]

[email protected]

Quality assessment of drainage networks extraction from a Digital Elevation Model (DEM)

Julie Charleux-Demargne, Indarto, Sylvain Payraudeau

UMR Spatial Structures & Systems, Cemagref – ENGREF, 500 rue J.F. Breton, 34093 Montpellier Cedex 5

Digital Elevation Model (DEM) interest for watershed management

Information extracted from DEM

Water resource integrated management requires a spatial knowledge of watersheds based on the structure given by drainage networks.

Watershed outlet

Watershedboundaries

Drainage network

Rotation 0°

2 Km11 0

Rotation 18°

211 0 Km

Rotation 36°

211 0 Km

Drainage networks extraction1. DEM (elevation) 2. Flow directions 3. Drainage areas

8 1 23

4567

8 possible directions : choice of the steepest downward slope

Drainage network : cellswith drainage area > 5

Drainage network No drainage network

19 26

17 21

22 30

28 32

37 41

32 39

42 45

45 48

11 10

9 8

17 16

20 20

10 12

6 9

15 18

20

42

23

30 30

24 25

35 35

40 40

31 33

27 29

38

41

40

2 1

3 1

2 1

2 1

2 1

2 1

2 1

1 1

1 2

1 3

1 12

11 9

33 3

48 6

17 3

4

1

3

3 6

10 4

2 3

1 1

3 3

4 3

2

1

2

Quality assessment method : probabilistic measure of drainage networks

Application of the method : optimising a procedure to improve the drainage networks extraction

The grid DEM is rotated through successive rotations to extract drainage networks relative to different angles θ. After inverse rotation (by - θ), those realisations of drainage networks are summed together to obtain a map giving the the rate of occurrence of each cell belonging to the drainage network.

P : probability for a cell to belong to the

drainage networks, in %.

The Pallas watershed : 52 km²

Examples on 3 watersheds of the Herault regionThe Vène watershed : 66 km² The Orb watershed : 1550 km²

The critical areas of highest uncertainty for drainage networks extraction are those with very low relief and those with river confluence.

DEM modification integrating a reference drainage network

Network used : BD Carto database (IGN)

Comparison of probabilistic measures to determine the optimal value dz

The critical areas persistent whatever the dz value correspond to areas where rivers are not represented in the BD Carto database.

The optimal dz value depends on the processed DEM & on the terrain within the watersheds : Pallas 8m ; Vène 4m ; Orb 8m

Grid DEM Drainage networks

Rotationangle θ

θ

Rotation angle -θ

2 realisations of drainage networks

θ

Quality assessment method based on the rotation of the grid DEM

1 ≤ P < 55 ≤ P < 10

10 ≤ P < 5050 ≤ P < 8080 ≤ P ≤ 100

Re-assignment of the DEM elevations for the cells belonging to the reference network

Before modifying After modifying with the dz value

dz

Elevation(m)

Comparison of % of cells belonging to drainage networks (P≥ 1) for different values dz :

Principe

θ varies in [0°, 90°[, incremented by 1° ; consequently, the map gives us the probability for a cell to belong to the drainage network.

Proposed solution

When dz increases, flow directions are more and more independent of the grid orientation, hence the extracted drainage networks are more stable

Probabilistic measure of drainage networksfor the downstream area of the Orb watershed

The more stablethe drainage network, the smaller the number of cells belonging to drainage networks for the 90 angles θAfter modifying

Critical areas

Reference drainage network

50

60

70

80

90

100

0m 2m 4m 6m 8m 10m

% c

ells

(P ≥

1) /

the

initi

al D

EM

Value dz

Orb watershed

Before modifying

Drainage networkGrid orientation

ConferenceHydrology of mediterranean regions

October 11th-13th, 2000Montpellier, France