Forest Fish Project

Lake Shoreline Contours

The Forest Fish project has made an important improvement to Ontario’s provincial DEM by incorporating an additional contour line during the DEM interpolation procedure.  This additional shoreline contour information helps to further refine the DEM and topographic indices calculated at the shoreline.  For hydro-ecological applications the land-water interface is of special importance.  Figures 1 demonstrates how flow can be erroneously routed across the terrestrial portion of the near-shore zone when shoreline contours are not used.  Figure 2 demonstrates the improvement to flow routing in low relief, near shore zones, especially for complex shorelines.

Figure 1. North Grace Lake – no shoreline contour used causing flow routing error

Figure 2. McKaskill Lake – with shoreline contour used and improved flow routing

Stripe Filtering

The second improvement made to the provincial DEM was the significant reduction of a regular striping error.  This error is commonly found in many elevation data sets and is a regular striping pattern produced from orthophotography using semi-automated profiling devices.  Many authors have recognized this problem and several have provided filtering/conditioning solutions (O’Callaghan & Mark 1984; Garbrecht & Starks 1995; Mussakowski et al. 1993; Gesch et al. 2002; Albani & Klinkenberg 2003).  The Forest Fish project implemented a technique published by the USGS (Oimoen 1997 & 2000) and found it to be most effective for resolving flow routing problems in small catchments (< 5ha.)  This is important for two reasons.  First, because the topographic indices used to predict brook trout habitat rely on proper flow routing and in-turn calculation of upslope contributing area.  Second, because small catchments around the shorelines of lakes are often used as brook trout habitat.

A sensitivity analysis was performed to determine the most appropriate size of vertical and horizontal kernels for the stripe-removing filter.  The filter quality was based on three measures.  The first criterion was a visual assessment of the resulting DEM to ensure enough of the striping was removed but also to avoid excessive generalization/blurring of the topography.  The second assessment was performed by examining five topographic profile lines in areas of the striping.  The third criterion was an analysis of flow routing and upslope contributing areas at each of the 18 brook trout habitat sites.

Visual Assessment

An example of vertical striping problems in the DEM around Dickson Lake.

The DEM after the most suitable filter is applied.

Topographic Profile Line Analysis

The graph below shows an example of the reduction in striping error from an elevation profile in the Dickson Lake catchment.  The profile line is 1200m in length and traverses terrain where vertical striping is visually evident in plan form.  Three bands of striping are evident at the height of land between horizontal distances 400m to 900m.  The blue profile line represents the pre-filtered, striped surface.  Post-filter changes to the topography are specific to areas that contained striping while the filter does not affect areas without striping.  The red line represents the post-filtered surface.

Flow Routing and Upslope Contributing Area Analysis

The two figures below show improvements to the flow routing and upslope contributing area calculations at a small brook trout catchment on Dickson Lake.  Improvements are determined by the increase in catchment size, which translates to a higher calculated value of topographic index at this site.  Prior to the filtering improvement the topographic index model would not have identified this brook trout site.  After the filter is applied the topographic index value is increased and is subsequently identified as a potential brook trout habitat.  The black arrows indicate the catchment outlet where YOY brook trout were observed in the field.