Top 10 Tips for Drainage Mapping
Using high-resolution topo data for drainage analysis presents unique challenges. Our experience shows that automated elevation models (DEMs) simply don’t work for drainage purposes. So here are our “Top Ten Tips” to help you avoid the pitfalls and set your mapping project up for success!
#1 Rasters are King
TINs (Triangular Irregular Networks) are important, particularly if your workflow involves manipulating mass points and breaklines. In fact, we use TINs extensively in the surface development process for both visualization and surface editing. But the flow paths and drainage catchment area calculations are raster-based (grid) processes that rely on the export and analysis of regularly spaced cells to determine things like slope and downstream flow accumulation.
#2 Depressionless DEMs are the Enemy
Anyone that has ever taken a class or done a tutorial on GIS-based hydrology and flow accumulation knows that Step #1 in the process is…creating a depressionless DEM. For national scale datasets designed for use of 1:24,000 scale, that may work fine. But for local scales looking to understand site specific issues, globally “filling in” depressions (or “sinks”) can royally screw up your surface.
Sinks are an important part of Hydrology!
Filling in sinks arbitrarily can artificially route flows in the wrong direction or away from their actual destination. And in some cases, sinks may be an important part of the hydrology, particularly when associated with man-made drainage structures, so you might actually want to leave some of them in! The key takeaway is that while “filling sinks” is important, it has to be done judiciously and carefully, potentially using masks and adjusting thresholds across your project area as you iterate toward that final “drainage surface”. Simply filling in all sinks to create a depressionless DEM will guarantee you will get false results.
#3 This is not your father’s elevation model
My dad taught me how to read topo maps as a kid, which led to my later fascination with digital surface models. The digital surface models (elevation models) used to support contour and stream mapping are the critical source for creating drainage surfaces, but they are not the same thing. Hydro-flattening is a relatively recent term brought into use by the USGS to support LiDAR data processing around large water bodies. Hydro-enforcement goes one step further to remove digital barriers to stream flows along the stream paths. Hydro-conditioning takes the process even yet further by accounting for how flow is routed across the landscape.
“It’s about the surface, stupid!”
In some cases, artifacts and voids in the LIDAR point cloud can influence how the digital surface will route water and must be manipulated and modified through selective classification and use of breaklines. The result is a drainage surface that actually works and reflects how water flows in the real world. But it’s important to understand that it’s not the same surface that you would use to make contour maps.
#4 The Process is Iterative
You don’t just cue it up and let it rip. Automation has its role, for example in stacking geo-processes and giving the user freedom to adjust parameters, but like with any artisanally crafted product, there is a certain amount of grinding out the rough edges and cleaning up the noise, looking for and fixing errors, that is necessary to produce a good drainage surface. And not all surfaces are made equal. It would be inappropriate, for example, to apply the same parameters and rules to urban Philadelphia as it would to the rural areas surrounding Olympia or the suburbs of Phoenix.
#5 Scale Matters
Lewis Carroll, author of “Alice in Wonderland” once penned a wonderful argument between fictional characters arguing about the scale of maps. At a rational level, we all understand that maps are just a generalization of the real world, but with the ability to zoom in and out indefinitely at will, somehow our minds secretly expect the results to be…perfect.
We zoom in on our screens past the point of blur and feel inherently disappointed when we can’t see EVERY DETAIL. At some point, we have to zoom back out and say, “Good enough”. We don’t have to fix every cell in the drainage raster, we just need the flow to be routed to the outlets (or inlets) properly, and the drainage area to each point to be correct. The scale of our analysis will ultimately dictate how much “cleaning” needs to go into the surface model preparation and how many cranks of the iterative wheel we have to churn.
“And then came the grandest idea of all! We actually made a map of the country, on the scale of a mile to the mile!”
– Sylvie & Bruno Concluded, by Lewis Carroll (1983)
#6 It’s About the Points, Stupid
A typical LIDAR deliverable is a hydro-flattened raster DEM (USGS 2018). It’s important to understand that the process of exporting a raster from the LiDAR point cloud itself is a smoothing process. It removes data and voids disappear. Cells with multiple point values are averaged. Cells with none are interpolated. Artifacts are difficult to identify and resolve. Using the originally classified LiDAR point cloud as the starting point is the best way to ensure high-quality drainage data comes out at the end.
#7 More is Not Necessarily Better
The good news is that more points mean more detail. The bad news is you may not need it. For example, in an urban environment, we want to make sure surface flow in the streets is routed down the gutter. We still have to deal with artifacts due to reflected shots off windows and near-ground feature resolution (i.e., low things that are hard to distinguish from ground during point classification). We still have tree overhangs and voids due to cars and other “non–ground” objects. More points aren’t going to change any of that.
The raster supporting your drainage analysis is what drives the required point density, not the other way around.
Our experience has shown that for drainage purposes, rasters at a higher resolution than 1m just add more overhead in terms of processing, file size, and yes, noise. So QL2 LIDAR (2ppsm) seems to be an ideal point density to support surface models for drainage. That being said, sensor technology continues to evolve and improve, and higher and higher density point clouds are being produced all the time. Just remember that the raster size supporting your drainage analysis is what drives the required point density. And for that, we recommend an ideal raster cell resolution of about 1m.
#9 GPS Point Data Needs to be Validated
What does every self-respecting municipality do when they want to do a culvert inventory and assessment project? They hire summer interns and arm them with GPS equipment. But while GPS has been an amazing and transformational technology, we still have to know what points we are measuring and how accurate they need to be. A culvert or catch basin inventory done from GPS needs to be scrubbed and validated against orthoimagery and then updated to reflect the intention of the drainage mapping initiative. For example, culvert GPS points are often captured at the crown of the road under which the culvert passes. For drainage mapping, those points need to be at the base of the inlet side in the ditch. Catch basin points are often captured on the manhole centered over the top of the structure. For drainage mapping, those points need to be moved into the street in the gutter (See “Put Your Mind in the Gutter”). It’s OK to artificially adjust culvert or catch basin inventory positions (e.g., as a copy) for drainage mapping purposes, but take a look at your data before you start, just so you know what you may be in for!
#8 Put Your Mind in the Gutter
Literally. Think about where water flows in developed areas. It’s not at the top of the curb or toward the catch basin manhole, which is behind the curb next to the sidewalk! It’s in the street, along the gutter. The iterative process required to create good drainage surfaces means following each flow accumulation pathway along its route to determine if it is a reasonable representation of how water should actually flow in the real world. This is an important thing to remember when it comes to how much detail you need in the point cloud, and how important breaklines are in identifying curbs and gutters.
Remember that the primary goal of the drainage surface is to help create accurate drainage areas at the scale needed (e.g., analyzing culverts, catch basins, MS4 outfalls, etc.). So long as the artificial flow paths inside that drainage area result in a reasonable representation of how water is routed, then derived metrics for things like slope and travel time will still likely be orders of magnitude more accurate than using field surveys, and at a fraction of the cost.
#10 Culverts & Catch Basin Catchments
Ultimately, the primary goal of a drainage surface editing project is to accurately depict the contributing drainage areas (drainage catchments) associated with key points of interest, like culverts and catch basins. Unfortunately, traditional DEMs, even those that meet USGS requirements for hydro-flattening of water bodies, are insufficient for the task. Said differently, “out of the box” DEMs simply won’t work for drainage analysis. Ensuring that artifacts and voids in the LiDAR data are not artificially influencing flow paths across the surface can be crucial. Hydro-enforcing and hydro-conditioning the surface by filling spurious sinks and adding breaklines to wall up or trench through barriers will remove “errors” in the drainage surface and improve the accuracy of how water is routed. But the result will be an “artificial” DEM that is no longer valid for topographic mapping or other purposes. Finally, placing your catch basin or culvert points accurately on the surface is critical. Often times erroneous results are as much about erroneously placed points as much as about errors on the surface. As such, the entire process is necessarily iterative to achieve accuracy at the scale you need.