Temporal variation in stream embeddedness and aquatic habitat

Abstract

Sediment is as a leading pollutant to streams and rivers in the United States and globally. An estimated 0.5 to 5 billion tons of soil is delivered to streams in the United States annually leading to sediment impairments in over 250,000 stream kilometers and billions of dollars in environmental damage. Excess fine sediment (< 2 mm diameter) can degrade habitat quality for aquatic species leading to declines in biodiversity and loss of ecosystem services. Embeddedness, a metric that quantifies the degree of fine sediment accumulation around coarser particles, has been used by ecologists and engineers to assess habitat quality. Typical embeddedness surveys are time intensive, spatially limited and often do not consider temporal variability. Recent studies have developed models for remotely predicting the average embeddedness at the network scale via its relationship with bankfull shear velocity. However, stream flow conditions and environmental factors are never constant; thus, embeddedness is expected to vary temporally. The primary focus of this study was to determine the environmental factors that influence embeddedness fluctuations and assess the accuracy of the Ridge and Valley (RV) embeddedness prediction model at Virginia Tech's StREAM Lab and 12 sites in the upper Roanoke River basin. Repeated embeddedness sampling was used to create a time series of incremental changes and examine how flow regime and temperature influence embeddedness. We found that embeddedness followed substantial temporal patterns at StREAM lab with peak values in late Winter early spring and minimum values in Summer. The strongest factors driving temporal variation were average air temperature, freeze-thaw cycles, and maximum stage between sampling events. Variation was present within the Upper Roanoke basin though no generalizable temporal patterns could be determined. However, the RV prediction curve produced a reliable and accurate estimate with a mean absolute error between 3 and 16% even in the presence of substantial variation for both the StREAM Lab and the Upper Roanoke. These results indicate that embeddedness can be predicted temporally to some extent. This could potentially increase the accuracy of remote predictions by accounting for seasonal fluctuations and help determine when embeddedness could cause suitable habitat to become unsuitable habitat.