Alan Kasprak - Current Research

Improving Techniques for Monitoring and Assessing River Dynamics

In addition to more fundamental research on sediment transport and modeling channel evolution, I have additional interests in applied topics, including how sediment and wood enters and moves through river systems, along with how classification frameworks can be used to understand channel processes through the lens of channel form. This work falls into three broad areas:

  • Leveraging large-scale data to understand wood dynamics in rivers
  • Understanding sediment delivery to channels and its influence on channel form
  • Using channel form to infer process through stream classification

This page describes each of these projects in more detail, along with any future directions I'd like to pursue for each.

Leveraging Large-Scale Data to Understand Wood Dynamics in Rivers

The advent of watershed-scale lidar data collection, and more recently, basin-scale structure-from-motion mean that we can collect data with sufficient resolution to not just study the form of landscapes, but also the habitat of the organisms that inhabit them. Add to this the availability of network-scale stream monitoring programs, and we have a starting point for drawing inferences regarding the occurrence of life as a function of its landscape.

In this case, I've built on work I began in my master's degree examining large wood in streams, which is vital in creating and promoting habitat for a host of aquatic organisms including fish and invertebrates. Working with, among others, Nate Hough-Snee and Joe Wheaton (Utah State University) and Brett Roper (US Forest Service), we've leveraged landscape-scale monitoring and remote sensing data to understand the natural and anthropogenic factors that cause wood in streams to exhibit significant variability across and between watersheds.

Examples of variability in wood frequency from monitoring across five basins in the Pacific Northwest

Examples of variability in wood frequency from monitoring across five basins in the Pacific Northwest

Using ordination, we examined the effects of climatic variables (e.g. annual precipitation), disturbance variables (e.g. recent timber harvest, road construction, grazing) and geomorphic variables (e.g. stream power) on the amount and volume of large wood in streams.

Factors influencing instream wood frequency and volume across five study basins shown above. Direction of lines indicate strength and nature (e.g. positive, negative) of correlation.

At the basin-scale, our results indicate that the most powerful predictions for wood loading in streams can be derived from large-scale variables such as climate and precipitation. A recently-published article in River Research and Applications details this finding. Within an individual watershed, however, variables that can influence reach-scale wood dynamics, including recent wildfire and forest composition played a larger role in determining the prevalence of large wood at monitoring sites. Stream power or transport competence was important on a site-by-site basis (detailed in a recent paper in Riparian Ecology and Conservation), but in terms of whole basins, it appears the ability of many trees to grow to large sizes plays the predominant role in determining wood abundance and volume.

Factors influencing instream wood frequency and volume across five study basins shown above. Direction of lines indicate strength and nature (e.g. positive, negative) of correlation.

I am interested in continuing this work, particularly by leveraging spatially-extensive, high-resolution datasets that can provide more accurate measurements of wood transport competence (e.g. channel width, stream slope, stream power) in concert with basin-scale flow and hydrologic modeling to further investigate the large-scale importance of wood transport on patterns of abundance across and between watersheds.

Check out some of our recent work on the drivers of in-channel wood across the American West in Riparian Ecology and Conservation and River Research and Applications!

Using Channel Form to Infer Process through Stream Classification

Classifying streams is a widespread practice in watershed management and river restoration, because we often want to know what a river "should" look like, or we're trying to understand the distribution of stream types (and their conditions) across a region.

There are a lot of different classification frameworks that can help accomplish these things, but how they compare - and the commitments of time, data, and expertise required by each one - hasn't been explored very much at all. What's more, frameworks typically contain varying degrees of measurements of either form (the width of a channel, for example) or process (how much sediment is moving through that channel). Frameworks that measure form in lieu of process are often criticized for being overly simplistic, or not capturing the drivers behind a channel's appearance.

Along with a group of collaborators from universities and agencies worldwide, I set out to understand when and why the outputs from four different classification frameworks varied across the Middle Fork of the John Day River, a large basin in the Pacific Northwest.

What we found was quite surprising. When we compared reach-scale outputs of the River Styles Framework, Natural Channel Classification, the Rosgen Classification System, and an automated, statistical clustering-based approach, the frameworks produced comparable outputs much of the time.

Clockwise from the top left, (1) River Styles, (2) Natural Channel Classification, (3) Rosgen Classification Framework, and (4) Statistical Clustering applied across the Middle Fork John Day basin.

Clockwise from the top left, (1) River Styles, (2) Natural Channel Classification, (3) Rosgen Classification Framework, and (4) Statistical Clustering applied across the Middle Fork John Day basin.

Of course, while the names each framework assigned to a particular reach were different, the geomorphic attributes (i.e. the units, planform, confinement, etc.) of those classes were strikingly similar. Here's an example of four sites, and how each framework classified them.

How do the frameworks compare across four selected sites? Examples of good, moderate, and poor agreement.

What's more, this agreement came about even though the four frameworks measured vastly different components of form and process, which suggested to us that simply characterizing a framework based on what it measures may not be indicative of its inherent quality. That's because much of the time, form reflects, and is derived directly from, process in rivers. Rather that that simple distinction being the driving factor in a manager or scientist's choice of a framework, the decision ought to ultimately rest on whether the data, time, or expertise required to execute the framework's procedure aligns with the goals of a particular project.

Check out the manuscript in PLOS ONE here!

This is a collaboration with: Nick Bouwes, Nate Hough-Snee, Gary O'Brien, Reid Camp, and Joe Wheaton (Utah State University) Gary Brierley (University of Auckland) Kirstie Fryirs (Macquarie University) Tim Beechie and Hiroo Imaki (NOAA) Dave Rosgen (Wildland Hydrology)