11:00 〜 13:00
[AHW23-P07] A mathematical programming approach to stream burning and hydrological adjustment
Stream burning and hydrological adjustment are important pre-processing techniques in hydrological modeling based on digital elevation models (DEMs). To obtain well-represented streams and local drainage directions in a raster form, these techniques adjust the raw elevation values to resolve the deviation of the derived streams as cells having high flow accumulation according to a known hydrographic map, and to remove depressions and flat areas hydrologically unnatural.
However, as most of the existing approaches adjust the elevation values in an ad hoc manner, the modification could largely alter the characteristics of the region, and consequently the results of the subsequent analysis, such as flood impact modeling. We should employ a better strategy of elevation filling and breaching as well as appropriate offset values in order to reduce the undesired impact of the pre-processing.
Here, we present a mathematical programming formulation for stream burning and hydrological adjustment that attempts to minimize the total modification applied to the DEMs. It can optimally determine whether filling or breaching strategy is appropriate for each subregion. The constraint conditions are defined to enforce monotonically decreasing elevations along the streams, the equivalence of stream and drainage directions, and the existence of a lower adjacent cell for land. The derived stream locations in the modified DEMs are consistent with the given hydrographical map, and the non-stream locations do not include depressions or flat areas.
Our preliminary experimental results show the effectiveness of our proposed approach compared to ordinary methods. We provided preliminary experiments using the Shuttle Radar Topography Mission (SRTM) dataset and the National Hydrography Dataset (NHD). We compare the terrain attributes obtained from our approach and existing methods.
To the best of our knowledge, the proposed approach is the first attempt to formulate the stream burning and hydrological adjustment in a mathematical programming problem. As our method can produce minimally modified DEMs for hydrological modeling, we believe that the methodology provides a promising alternative to the existing methods particularly when the impact of the adjustment to the terrain attributes would severely affect subsequent hydrological analyses.
However, as most of the existing approaches adjust the elevation values in an ad hoc manner, the modification could largely alter the characteristics of the region, and consequently the results of the subsequent analysis, such as flood impact modeling. We should employ a better strategy of elevation filling and breaching as well as appropriate offset values in order to reduce the undesired impact of the pre-processing.
Here, we present a mathematical programming formulation for stream burning and hydrological adjustment that attempts to minimize the total modification applied to the DEMs. It can optimally determine whether filling or breaching strategy is appropriate for each subregion. The constraint conditions are defined to enforce monotonically decreasing elevations along the streams, the equivalence of stream and drainage directions, and the existence of a lower adjacent cell for land. The derived stream locations in the modified DEMs are consistent with the given hydrographical map, and the non-stream locations do not include depressions or flat areas.
Our preliminary experimental results show the effectiveness of our proposed approach compared to ordinary methods. We provided preliminary experiments using the Shuttle Radar Topography Mission (SRTM) dataset and the National Hydrography Dataset (NHD). We compare the terrain attributes obtained from our approach and existing methods.
To the best of our knowledge, the proposed approach is the first attempt to formulate the stream burning and hydrological adjustment in a mathematical programming problem. As our method can produce minimally modified DEMs for hydrological modeling, we believe that the methodology provides a promising alternative to the existing methods particularly when the impact of the adjustment to the terrain attributes would severely affect subsequent hydrological analyses.