We ran the same storm through two identical models. Only the terrain grid size changed. Here is what we learned.
Every hydrologist and civil engineer has faced the same question at the start of a project: Is my data good enough?
We worry about rainfall records. We worry about land cover maps. We worry about soil data. But one variable often escapes scrutiny the Digital Elevation Model (DEM) grid size that defines our watershed boundaries and terrain slopes.
With high-resolution LiDAR data becoming more accessible, the pressure to use finer grids is growing. But is finer always better? And more practically: Does a 1m DEM actually change your design discharge enough to matter?
To find out, we ran a controlled experiment.
The Experiment: One Storm, Two DEMs, Everything Else Identical
We selected a small catchment—approximately 2 square kilometers, representative of a typical rural bridge or culvert drainage area. We then built two separate HEC-HMS models:
| Model | DEM Resolution | Data Source |
|---|---|---|
| Model A | 10 meters | Standard publicly available DEM |
| Model B | 1 meter | High-resolution LiDAR |
Every other parameter was held constant:
Same meteorologic model (same storm event)
Same SCS Curve Number (70)
Same impervious area (5%)
Same lag time (40 minutes)
Same baseflow recession parameters (0.7 recession factor, 0.05 initial baseflow, 0.04 threshold)
The only variable was the DEM resolution used to delineate the watershed and calculate slopes.
The Results: Less Difference Than Expected
When we compared the outflow hydrographs at the catchment outlet, the differences were surprisingly small:
Peak Discharge: The 1m DEM produced 0.85 m³/s compared to 0.81 m³/s from the 10m DEM — a difference of less than 5 percent.
Time of Peak: Both models showed the peak occurring at exactly the same time (03May2023 at 00:56). The finer terrain detail did not speed up or delay the flow.
Total Runoff Volume: The 10m model produced 6.91 mm of runoff; the 1m model produced 6.79 mm — a difference of less than 2 percent.
What Caused the Difference?
When we dug deeper, the primary driver of the peak flow difference was not slope precision—it was drainage area delineation.
| Model | Drainage Area |
|---|---|
| 10m DEM | 2.05 km² |
| 1m DEM | 2.16 km² |
The high-resolution DEM captured slightly more contributing area at the edges of the basin. These small fringe areas subtle swales and marginal drainage paths are smoothed over or completely missed by the coarser 10m grid. The refined slope representation alone had almost no independent effect on the outflow.
What Does This Mean for Engineering Practice?
For a bridge or culvert designer, the key question is not statistical significance it is practical significance.
A 5 percent difference in peak discharge falls well within the safety factors we already apply in hydraulic design. Typical bridge design safety margins range from 25 to 50 percent. The uncertainty in rainfall intensity-duration-frequency curves alone is often ±10 to 20 percent.
In other words: The difference between a 10m and 1m DEM is smaller than the uncertainty you already accept in your rainfall data.
When Does High-Resolution DEM Actually Matter?
This experiment tested a specific scenario: a rural 2 km² catchment. The answer may change with context.
High-resolution DEM (1m or better) IS valuable when:
| Application | Why |
|---|---|
| Urban drainage design | Curbs, inlets, and small swales matter at sub-meter scale |
| Floodplain mapping | FEMA requires high-resolution terrain for accurate flood boundaries |
| Bridge scour analysis | Localized flow convergence needs fine topography |
| Dam breach modeling | Downstream wave propagation is terrain-sensitive |
Standard-resolution DEM (10m or 30m) IS sufficient when:
| Application | Why |
|---|---|
| Rural bridge design | 5% peak flow difference is within safety margins |
| Regional watershed planning | Coarser grids run faster and are often free |
| Climate change impact studies | Uncertainty in future rainfall dwarfs DEM error |
| First-pass screening | Identify problem areas before investing in LiDAR |
Practical Recommendations
Based on this analysis, here is a simple decision framework:
Start with available data — Do not automatically demand LiDAR. A 10m DEM may be perfectly adequate.
Match resolution to risk — High consequence infrastructure near dense development may justify high-resolution data. A rural crossing likely does not.
Test sensitivity yourself — Run your model with coarser and finer DEMs. You may find, as we did, that the difference is negligible.
Spend your budget where it matters — If you have limited resources, invest in better rainfall data or site-specific runoff curve numbers before upgrading your DEM.
Final Thoughts
More data is not always better data. Better decisions come from understanding what a given data resolution actually buys you.
For a small rural catchment, a 1m DEM produced a peak flow less than 5 percent higher than a 10m DEM—a difference that disappears within standard engineering safety factors and rainfall uncertainty.
That does not mean high-resolution DEMs are useless. It means they are not always necessary.
What Do You Think?
Have you compared DEM resolutions in your own modeling work? At what catchment scale did you start seeing meaningful differences? Share your experience in the comments below.
No comments:
Post a Comment