Use of open-source GIS for the preprocessing of distributed hydrological. models

Transcription

1 Use of open-source GIS for the preprocessing of distributed hydrological models F. Branger, I. Braud, S. Debionne, J. Dehotin, S. Jankowfsky, O. Vannier, P. Viallet

2 Who are we? Cemagref Hydrology-Hydraulics Research Unit Researchers in hydrological modelling Development and application of distributed hydrological models Hydrowide Consulting in hydro-informatics Use of GIS data as model inputs Objective of this presentation Share our non-specialist experience with open-source GIS From 2005 to 2009 Non exhaustive 2

3 Outline Introduction Principles of the hydrological modelling approach Input data requirements Examples of pre-processing operations and problems 1. DEM analysis and subcatchment delineation 2. Topology of vector layers 1. One data layer 2. Several data layers 3. Advanced editing of vector layers Conclusion and perspectives 3

4 1. Introduction: Hydrological models At the watershed scale, simulation of the water cycle Distributed hydrological model : Watersheds are discretized into homogeneous spatial units Application of equations on these units Conceptual / physically based models Parameters and boundary conditions are spatialized Land use, soil properties, DEM 4

5 1. Introduction: The LIQUID modelling platform Framework for developing hydrological models (Viallet et al., 2006) Accelerates development of integrated and distributed models Increases reusability (models are a combination of modules) Selection of consistent 3rd party libraries (I/O, LinAlg, DB ) Open source (but not free) software, proprietary of Hydrowide The use of the platform at Cemagref Model concept follows object-oriented principles Irregular hydro-landscapes (hydrological functional units) 1 hydro-landscape = 1 module 5

6 1. Introduction: Spatial discretization Model mesh in LIQUID = vector descriptions of hydrolandscapes PostgreSQL/PostGIS data base Polygons, lines, points Requirements : Model mesh = jigsaw puzzle Perfect fit of all the pieces (model water balance): No overlapping, no holes in and between different polygons 6

7 1. Introduction Application cases Flash floods on mediumsize catchments Influence of landscape management on hydrology on small catchments (0-10 km²) AVuPUR project : 7

8 Outline Introduction Principles of the hydrological modelling approach Input data requirements Examples of pre-processing operations and problems 1. DEM analysis and subcatchment delineation 1. Topology of vector layers 1. One data layer 2. Several data layers 2. Advanced editing of vector layers Conclusion and perspectives 8

9 1. DEM analysis and delineation of subcatchments Available data : Raster DEM River network vector map (optional) SAGA GIS 1.2 / MapWindow TauDEM Operations : sink removal flow directions extraction of river network delineation of subcatchments conversion to shape file 9

10 1. DEM analysis : problems with SAGA GIS Self- intersecting polygons or multi-polygons 10

11 1. DEM analysis : problems with SAGA GIS Self- intersecting polygons or multi-polygons Overlapping polygons 11

12 1. DEM analysis : problems with SAGA GIS Self- intersecting polygons or multi-polygons Overlapping polygons «Physically incorrect» polygons 1-pixel subcatchments (independent of the min. pixel threshold) Unrealistic catchment shapes at river reach confluences 12

13 1. DEM analysis : solutions? Manual editing (OpenJump, QGIS) Use of MapWindow TauDEM TauDEM/MapWindow SAGA GIS 13

14 Outline Introduction Principles of the hydrological modelling approach Input data requirements Examples of pre-processing operations and problems 1. DEM analysis and subcatchment delineation 1. Topology of vector layers 1. One data layer 2. Several data layers 1. Advanced editing of vector layers Conclusion and perspectives 14

15 2. Topology of vector layers Requirements : topologically correct final mesh of polygons No overlapping of different polygons Adjacent polygons need to have the same vertexes No holes in and between polygons (water balance) Minimal size for polygons in order to limit their number No multi-polygons 15

16 2. Topology of vector layers Requirements : topologically correct final mesh of polygons No overlapping of different polygons Adjacent polygons need to have the same vertexes No holes in and between polygons (water balance) Minimal size for polygons in order to limit their number No multi-polygons Test of topological functions of different GIS: OrbisGIS, OpenJump and GRASS 16

17 2.1 Topology of vector layers: one data layer Test data set of polygons, topologically correct 17

18 2.1 Topology of vector layers: one data layer Test of function OrbisGIS/Find overlapping features: SELECT intersection(a.the_geom, b.the_geom) as the_geom FROM polygones a, polygones b where intersects(a.the_geom, b.the_geom) and a.id <> b.id; Results in three output files: Overlapping polygones Overlapping lines Overlapping points 18

19 2.1 Topology of vector layers: one data layer Results 19

20 2.1 Topology of vector layers: one data layer 20

21 2.1 Topology of vector layers: one data layer Test of function Detect/Overlaps of OpenJump The merge: 21

22 2.1 Topology of vector layers: one data layer Methodology Topology is automatically built while loading data to GRASS GIS (v.build) gaps are transformed to polygons Use of v.clean/rmarea to merge the new small polygons to adjacent bigger polygons Last manual corrections with OpenJump editor tools (cut and union polygons) 22

23 2.1 Topology of vector layers: one data layer Original data GRASS output 23

24 2.2 Topology of vector layers: several data layers Objectives: No overlaps, no holes and not too many polygons Original data Intersection with road Union of polygons Information loss 24

25 2.2 Topology of vector layers: several data layers A better solution would be to: divide polygons with holes fusion small polygons with the same attributes, which are close together fusion all holes inside one polygon to a single hole with same surface At the moment only manually! 25

26 Outline Introduction Principles of the hydrological modelling approach Input data requirements Examples of pre-processing operations and problems 1. DEM analysis and subcatchment delineation 2. Topology of vector layers 1. One data layer 2. Several data layers 3. Advanced editing of vector layers Conclusion and perspectives 26

27 3.1 Intersection of lines and polygons Need to divide polygons according to the river network (one polygon on each border) The polygon has to be divided at the source of the river At the moment no automatic function exists 27

28 3.1 Intersection of lines and polygons Need to divide polygons according to the river network (one polygon on each border) The polygon has to be divided at the source of the river At the moment no automatic function exists 28

29 3.1 Intersection of lines and polygons Need to divide polygons according to the river network (one polygon on each border) The polygon has to be divided at the source of the river At the moment no automatic function exists 29

30 3.2 Preparation of the river network The river needs to be subdivided at each intersection with an adjacent polygon The direction of each river reach must be from downstream to upstream 30

31 3.3 Transformations lines -> polygons Transform hedgerows from Linestrings to Polygons Change the boundaries of the adjacent fields Processed in a C program embedding SQL queries (PostgreSQL/PostGIS) (Spatola, 2005) Libpq library PostGIS functions involved : buffer, intersection, difference, geomunion 31

32 3.3 Transformations lines -> polygons Program crashes Typical GEOS errors : «side location conflict», «unable to assign hole to a shell» Reasons not very clear Our observations: Generation of «geometry collections» at intersection step => crash at the following step Slight changes in the coordinates of geometries between 2 steps : rounding errors? 32

33 3.3 Transformations lines -> polygons Program crashes Generation of «dirty» geometries improper to hydrological computation 33

34 3.4 Determination of interfaces Determination of boundaries between different elements Necessary for calculation of lateral fluxes (interfaces) The format «Linestring» is needed Calculation of Interfaces length distances to center PostgreSQL/PostGIS ; SQL queries (intersection) 34

35 3.4 Determination of interfaces Field/hedgerows borders Polygon/polygon intersections Generation of multipolygons/ geometry collections narrow polygons, self intersection, self tangency unconnected Automatic editing unsuccessful Manual editing for each element 35

36 Conclusions Many data pre-processing operations are required for distributed hydrological models Our current experience : combinations of several opensource GIS SAGA GIS Taudem/MapWindow OpenJump (The Merge) QGIS OrbisGIS PostgreSQL/PostGIS GRASS GIS Our objective : reduce the number of softwares used and automate as far as possible the preprocessing Dysfunction of GEOS/JTS Intersection causes problems at different preprocessing steps 36

37 Conclusions Required functionalities: Raster processing (extraction of river network, calculation of sub-basins, vectorisation) Topological corrections (overlaps, gaps,...) Creation of planar topology (edges, nodes, faces) Flip lines Intersection of polygons with lines, automatic division of polygons with holes Advanced manual edition tools: cut/merge polygons, move vertex to vertex of another layer Script language for automatisation of processes GRASS GIS = a promising candidate Still a lot of work to be done 37

38 Many thanks for the attention! Questions? 38