pyDEM: Global Digital Elevation Model Analysis

Abstract

Hydrological terrain analysis is important for applications such as environmental resource, agriculture, and flood risk management. It is based on processing of high-resolution, tiled digital elevation model (DEM) data for geographic regions of interest. A major challenge in global hydrological terrain analysis is addressing cross-tile dependencies that arise from the tiled nature of the underlying DEM data, which is too large to hold in memory as a single array. We are not aware of existing tools that can accurately and efficiently perform global terrain analysis within current memory and computational constraints. We solved this problem by implementing a new algorithm in Python, which uses a simple but robust file-based locking mechanism to coordinate the work flow between an arbitrary number of independent processes operating on separate DEM tiles.We used this system to analyze the conterminous US's terrain at 1 arcsecond resolution in under 3 days on a single compute node, and global terrain at 3 arc-second resolution in under 4 days. Our solution is implemented and made available as pyDEM, an open source Python/Cython library that enables global geospatial terrain analysis. We will describe our algorithm for calculating various terrain analysis parameters of interest, our file-based locking mechanism to coordinate the work between processors, and optimization using Cython. We will demonstrate pyDEM on a few example test cases, as well as real DEM data.

Data-Driven, Multi-Model Workflow Suggests Strong Influence from Hurricanes on the Generation of Turbidity Currents in the Gulf of Mexico

“…To assist the modeling of the generation and then the fate of the turbidity currents, derivative layers were computed from the SRTM30+ bathymetry, including slope gradients, and location and dimensions of the channelizations. Channel locations are well-discriminated using integration methods such as contributing area; here, the PyDEM package [53] was used for such procedure (see background on Figure 6B). On those features, channel-floor dimensions like widths and gradients were computed using operations on the original gridded bathymetry [54], and their values were mainly supplied to the RANS/TURBINS turbidity current models.…”

Cold-Air Pooling in the Upper Douro Valley: An Observational Study

“…To calculate the topographical amplification factor along the Douro Valley and tributaries, the pixels in the SRTM data corresponding to the floors of the valley and its tributaries need to be identified in the down-valley direction. These pixels were found using the pyDEM package (Ueckermann et al 2015; see the appendix). First, the magnitudes of the slopes of the SRTM data and their directions (or aspects) were calculated (Fig.…”

“…A connectivity (or adjacency) matrix is constructed as part of this calculation. The value in row i and column j of this matrix represents the fraction of the area of pixel j that drains into pixel i (Ueckermann et al 2015). Hence, by starting at a pixel located at the head of a tributary and following the main drainage route in the connectivity matrix, the pixels corresponding to the valley floor can be identified in the down-valley direction.…”

“…These areas are treated as flat regions (or "flats") within pyDEM. The total area contributions that drain into a flat region are collected, and then distributed proportionally to pixels at the edge of the flat, depending on the difference in elevation between the flat region and surrounding pixels (Ueckermann et al 2015). The coordinates of pixels within the flat region are not included in the connectivity matrix.…”