Citation:

Detenbeck, N. Plans for fine-resolution stream/river temperature modeling in the Penobscot watershed. Maine Temperature Monitoring Workgroup Annual Meeting, Augusta, ME, May 17, 2023.

Impact/Purpose:

This informal presentation is being shared with the Maine Temperature Monitoring Workgroup which has been developing a database of continuous monitoring data for temperature across the state with input from Federal, state, tribal, and NGO entities.  Spatial statistical network models are one approach for predicting stream thermal regimes across a stream/river network.  They can be used to assess the distribution of existing coldwater habitat for fish such as salmonids and brook trout and to map the occurrence of cold water refuge areas, important during times of low flow and temperature stress.  They can also be used to prioritize restoration of forested buffers to protect and increase coldwater habitat.  We will discuss development of a fine-resolution SSN model for the Penobscot River Basin in Maine, using data contributed by the Maine Temperature Monitoring Workgroup.

Description:

Spatial statistical network (SSN) models can be used to model temperature throughout stream and river networks and can account for spatial correlation.  For example, this approach allows us to determine the distance downstream that forested riparian zones will influence and protect cold water thermal regimes.  Taking spatial autocorrelation into account also prevents us from underestimating variance and including too many predictors in a model or including the wrong predictors because of spurious correlations.  SSN models can account for effects of intra- and interannual flow regime variability, groundwater inputs, topographic and vegetation shading, air temperature, and interactions of stream morphometry with these factors.  They can also account for temporal variation in thermal regimes by developing separate models for monthly metrics, by predicting more integrated measures (seven day average daily maximum (7DADM) or growing season maximum), or through integration with hydrologic and climate change models.  SSN models have been used to predict and assess intra- and interannual variation in cold water habitat, the distribution of cold water refuge areas during stressful periods of low flow and high solar radiation, the effect of riparian forest structure (length and width of forested buffers) on thermal regimes, and to forecast the interaction of vegetation management with climate change effects.  We developed new R programs available to run in the cloud to efficiently calculate shade on high resolution networks across large regions.  We will be developing a fine-resolution SSN temperature model for the Penobscot River Basin using continuous monitoring data collected by the Maine temperature monitoring network.  The success of an SSN model depends in part on the distribution of existing monitoring stations.  Do they represent the full range of covariates and covariate interactions expected to influence stream temperature?  Do points cover a range of sample densities to allow us to model spatial autocovariance patterns?  Do we have monitoring data downstream of dams or large lakes that disrupt downstream patterns of influence?  Success of SSN model development also depends on the quality of input data, e.g., accuracy of station locations on the stream network, screening of time series for outliers related to logger burial or “out of water” events, and removal of potential biases related to short data gaps by infilling. 

Record Details: