Terrestrial blog/Learning from disaster? (14 September 2013)
LEARNING FROM DISASTER?
As I write this, the Colorado Front Range in my home town of Boulder has experienced record-shattering rainfall and accompanying catastrophic flooding. Although the flood waters have yet to recede, and more rain is in the forecast for the next day and half, it is already abundantly clear that this event has produced major geomorphic and sedimentary impacts. This is of course not the first major flood in Colorado's history; powerful floods struck different parts of the Front Range in 1894, 1965, 1976, and 1997. But in addition to being unusually widespread, the September 2013 floods are sure to be among the most thoroughly monitored flood disasters to date. The region is home to several major universities and government research laboratories, and as a result has an unusually rich infrastructure of environmental sensing technology. In addition, in an age when nearly everyone owns a digital camera with video capability, it is likely that the event will turn out to be among the most thoroughly photo-documented in the region's history so far. All of this presents a valuable opportunity for our community to deepen our understanding of flood-impact dynamics, and to test and hone our modeling tools accordingly.
Let's start with a few quick facts and figures. Between noon on Wednesday 11 September and noon on Friday 13 September, a weather station at the National Center for Atmospheric Research (NCAR) Mesa Lab recorded about 280 mm (about 11 inches) of rain. Meanwhile, there are reports of some nearby areas receiving more than 350 mm (see https://udfcd.onerain.com/home.php). By comparison, the average ANNUAL precipitation for Boulder is about 500 mm, and the average September precipitation is about 47 mm (data from NOAA ESRL: http://www.esrl.noaa.gov/psd/boulder/Boulder.mm.precip.html). The sustained heavy rain clearly saturated most of the terrain, turning it into a virtual parking lot.
The principal stream that runs through the city of Boulder is Boulder Creek, which drains a roughly 300 square mile (780 square km) basin that has its headwaters along the continental divide in the Colorado Front Range. Its normal flow rate at this time of year is a little less than 2 cms (60 to 70 cfs). On Thursday 12 September, an automated USGS gaging station recorded a peak stage equivalent to over 6000 cfs (about 170 cms). Later that day, after the stage had dropped slightly, a heroic USGS crew made a direct measurement that came out to something like 4600 cfs (130 cms), more or less matching the automated stage-discharge calculation for that time (see http://waterdata.usgs.gov/nwis/uv?06730200). Other streams north and south along the Front Range foothills experienced record-shattering peaks, including Saint Vrain Creek (whose floodwaters have inundated the town of Lyons and split the city of Longmont in half), the Big Thompson River, and the Poudre River, among others.
The human impacts have yet to be fully evaluated. As of this writing (Saturday evening, 14 September), there are three confirmed deaths, and more than 200 people are unaccounted for. With canyon roads washed out or blocked by landslides, mountain communities in the county are accessible only by helicopter.
The morphologic impacts are clearly widespread. Debris flows have been reported in mountain canyons (we'll know more as the roads are re-opened). Shallow landslide scars can be seen here and there along the hills formed over sedimentary rocks along the foothills. Based on media reports and my own wanderings my own neighborhood in south Boulder, streams appear to have scoured banks and widened their channels in many locations. Helicopter video of the town of Lyons shows what look to be potential cutoffs and course changes along Saint Vrain Creek (though it may ultimately re-establish its original channel, perhaps with help from local bulldozers, in the coming weeks). The banks of the main stream in my neighborhood, an intermittent drainage called Bear Creek, are now adorned fresh bars of cobble- to small-boulder-sized material. At sharp bends, cut-bank erosion has substantially widened the channel.
The Boulder Creek Critical Zone Observatory (BCCZO) lies at the epicenter of this storm. BCCZO includes three study catchments within the Boulder Creek basin. Each catchment contains one or two meteorological stations and stream gages, as well as soil moisture sensors. Other instrumentation varies somewhat at each site; for example, the Gordon Gulch site has several groundwater wells equipped with water-level sensors. (For more information on the Boulder Creek CZO, see http://criticalzone.org/boulder/)
In 2010, the National Center for Airborne Laser Mapping (NCALM) collected lidar data collected over much of the Boulder Creek drainage basin. Among other things, the lidar data provide a 1-m resolution, bare-ground DEM three years before the flood. The BCCZO team is already discussing the prospect of conducting a repeat lidar survey in order to quantify the event's morphologic impact. In addition to the BCCZO lidar, there are at least two other coverages in the region that could be studied as well.
A high-resolution, lidar-derived map of topographic change would provide an interesting opportunity for modelers to test the ability of models of hydrodynamics and sediment transport to predict the impact of a major rainstorm. Precipitation input could be estimated from weather radar and rain gauges, while lidar coverage would provide the base topography. Surface roughness could potentially be estimated from satellite imagery, aerial photographs, and/or lidar-derived vegetation-canopy estimates. A natural target for hydrodynamic models would be to determine whether they can reproduce the observed stream-flow records (amazingly, many gaging stations seem to have survived). Adding a model of sediment transport to the mix would enable prediction of locations and magnitudes of erosion and deposition, and comparison with changes measured from lidar differencing.
Naturally, such an exercise would face several challenges. Any input space-time precipitation data is bound to be imperfect. Hydrologic properties of soils would have to be estimated over a relatively large area (though perhaps model sensitivity to soil hydrology would be minimal in view of the widespread near-surface saturation).
An additional challenge would surely be the "micro-topography" of the built environment in the more densely populated areas of impact. Even lidar is limited in its ability to capture the complex terrain of buildings, walls, berms, embankments, and similar features. These small-scale details seem to make all the difference to whether or not a particular basement floods (I'm hearing lots of stories that testify to how capricious these house-by-house impacts have been). On the other, urban micro-topography may turn out not to matter much for the overall pattern of inundation, discharge, and shear stress.
Another, perhaps bigger, challenge would be to capture the role of large, irregular debris. The "sediment" load often includes materials such as trees, propane tanks, automobiles, and even the occasional small building (for various photos of the flood damage, see: http://mediacenter.dailycamera.com/2013/09/14/photos-colorado-flood-damage-aerial-views/). These objects block flow and could potentially determine when and where a particular location experiences bank scour or channel avulsion.
Clearly, any modeling study of the flood's impacts would have to involve some element of uncertainty analysis. This need aligns well with CSDMS 2.0's goal of fostering and facilitating quantitative uncertainty analysis within our community.
Despite the challenges, events like the 2013 Front Range floods---especially when they occur in regions with pre-existing lidar data and real-time measurements of rainfall and stream flow---offer us an opportunity to sharpen our ability to model nature's extremes. Better models, in turn, are key to better planning and preparedness.
-Greg Tucker, 14 September 2013