Ideas

Parent Previous Next


Using the node layer for error checking


Once you have extracted the node layer it would be well worth your while to check the output. A mouth node typically has 1 polyline flowing into it although it is not unusual to have 2 or 3 polylines flowing into it. Thus filtering by type and sorting by valency this will help you identify unusual mouth nodes. For example a mouth with 10 polylines flowing into it would be a candidate for review.


The same process can be applied to source nodes. One would typically expect a single polyline flowing out (but 2 would not be unusual) of the source. Therefore a source node with a valency of 3 or greater would be a candidate for review.


Mouth nodes should only exist at the outlet of the network, so if they are seen in the middle of the network this would suggest a disconnection.


Sources within a network could be a saddle but they may represent polylines flowing in an upstream direction.


Polyline length


The origin of your network may be hand digitized, DEM or a centreline extracted from another dataset. Search for very small polylines (e.g. less than 1m) and you may find erroneous polylines. Such polylines may not actually be causing topological errors but equally they are contributing nothing to the network, so could be removed. Such polylines need to be reviewed on an individual basis.


Polyline attribution


Check the attribution of your network for logical errors. Fields should have values, even default values like -1 or -9999, if not why? Name fields should only contain valid characters. You cannot have a strahler order of zero and it its unlikely to go much above 8 so if you have a strahler value of 12 then its worth checking why. The maximum a Shreve order can be is the number of sources for that catchment. These simple tests can often be done by simply sorting the table by the field.


Networks encoded with a value that you expect to have continuity along its length are worth checking. For example river name should not be missing along the length of a river with a known name.


Does your network have a field that encodes the network into river/lake centreline? If so then you would expect the polylines that represent the lake centreline to be labelled as such.


If you have used RivEX to encoded catchment ID then colour code your network by it and just look at it. If you see a change of colour within a catchment then this would indicate a break in the network or possibly a small catchment draining into a sink hole / canal system.