A Parallel Algorithm for Removing Spurious Pits from Terrain Data by Carving
Location
ECC 216
Event Website
http://water.usu.edu/
Start Date
4-3-2012 3:45 PM
End Date
4-3-2012 3:50 PM
Description
This presentation will describe a parallel algorithm that has been adapted from the geomorphometric method of carving to reduce computation time for eliminating sinks from a grid-based digital elevation model. Grid-based Digital Elevation Models (DEM) are complete electronic representations of the land surface depicted as rectangular data arrays. A grid DEM is a continuous surface, where each pixel represents a value of elevation at a geographic location and the size of the cell indicates the resolution. Hydrologic modeling requires pre-processing of land surface datasets to remove local minima that impede the flow of water and allow accurate hydrological modeling. These sinks can be true artifacts of the land surface or a result of the creation of the DEM and must be removed so that a fully connected river network can be generated. Soille (2003) proposed carving as an alternative to sink filling and functions by generating a descending path from the bottom of the sink. Terrain elevations are successively lowered to that of the sink, following along a flow path until a lower elevation is reached. This approach has a benefit to the commonly used alternative, fill, which can produce large flat areas making it difficult to accurately determine flow direction over the area. Advancements in technology have significantly increased the resolution of DEMs, resulting in improved accuracy, greater areal coverage, and subsequently larger file sizes. The amount of time required to perform hydrologic computations on the larger datasets has also considerably increased. Enhancements in computer hardware have helped reduce computation time, but there is a need to process hydrologic data more efficiently. The Message Passing Interface Standard (MPI) is a message passing library standard of communications protocols to program parallel computers. Implementations of parallel algorithms have shown to be one successful approach to reducing computation time for hydrologic modeling. The approach discussed by Wallace et al. (2010) decomposes the dataset into strips that are distributed between processors, reducing processing load and as a result, processing time. This method was implemented in adapting carve algorithm to a parallel version. A main processor distributes strips of the dataset, each processor implements the carve algorithm on their dataset and returns a modified strip to the main processor that generates the final, modified dataset. Soille, P., J. Vogt, and R. Colombo, Carving and adaptive drainage enforcement of grid digital elevation models, Water Resour. Res., 39(12), 1366, doi:10.1029/2002WR001879, 2003 Wallace, R. M., D. G. Tarboton, D. W. Watson, K. A. T. Schreuders and T. K. Tesfa, (2010), ‘Parallel Algorithms for Processing Hydrologic Properties from Digital Terrain,’ GIScience 2010, Sixth international conference on Geographic Information Science, Zurich, Switzerland, September 14-17, http://www. giscience2010.org/pdfs/paper_229.pdf
A Parallel Algorithm for Removing Spurious Pits from Terrain Data by Carving
ECC 216
This presentation will describe a parallel algorithm that has been adapted from the geomorphometric method of carving to reduce computation time for eliminating sinks from a grid-based digital elevation model. Grid-based Digital Elevation Models (DEM) are complete electronic representations of the land surface depicted as rectangular data arrays. A grid DEM is a continuous surface, where each pixel represents a value of elevation at a geographic location and the size of the cell indicates the resolution. Hydrologic modeling requires pre-processing of land surface datasets to remove local minima that impede the flow of water and allow accurate hydrological modeling. These sinks can be true artifacts of the land surface or a result of the creation of the DEM and must be removed so that a fully connected river network can be generated. Soille (2003) proposed carving as an alternative to sink filling and functions by generating a descending path from the bottom of the sink. Terrain elevations are successively lowered to that of the sink, following along a flow path until a lower elevation is reached. This approach has a benefit to the commonly used alternative, fill, which can produce large flat areas making it difficult to accurately determine flow direction over the area. Advancements in technology have significantly increased the resolution of DEMs, resulting in improved accuracy, greater areal coverage, and subsequently larger file sizes. The amount of time required to perform hydrologic computations on the larger datasets has also considerably increased. Enhancements in computer hardware have helped reduce computation time, but there is a need to process hydrologic data more efficiently. The Message Passing Interface Standard (MPI) is a message passing library standard of communications protocols to program parallel computers. Implementations of parallel algorithms have shown to be one successful approach to reducing computation time for hydrologic modeling. The approach discussed by Wallace et al. (2010) decomposes the dataset into strips that are distributed between processors, reducing processing load and as a result, processing time. This method was implemented in adapting carve algorithm to a parallel version. A main processor distributes strips of the dataset, each processor implements the carve algorithm on their dataset and returns a modified strip to the main processor that generates the final, modified dataset. Soille, P., J. Vogt, and R. Colombo, Carving and adaptive drainage enforcement of grid digital elevation models, Water Resour. Res., 39(12), 1366, doi:10.1029/2002WR001879, 2003 Wallace, R. M., D. G. Tarboton, D. W. Watson, K. A. T. Schreuders and T. K. Tesfa, (2010), ‘Parallel Algorithms for Processing Hydrologic Properties from Digital Terrain,’ GIScience 2010, Sixth international conference on Geographic Information Science, Zurich, Switzerland, September 14-17, http://www. giscience2010.org/pdfs/paper_229.pdf
https://digitalcommons.usu.edu/runoff/2012/Posters/27