Sediment Models
SSCs measured in Carnivore and Chamberlin Creeks were modeled using a combination of hydrologic, meteorological, and temporal variables. The modeling periods are the 2015 and 2016 open-channel seasons, which encompass the vast majority of annual sediment transfer to Lake Peters. To develop multiple-regression models of SSC for Carnivore and Chamberlin Creeks, a similar approach to Hodgkins (1999) and Schiefer et al. (2017) was used. We assessed 60 potential predictors: a range of hydrological, meteorological, and temporal explanatory variables, all linearly interpolated to match times of SSC sampling (Table 1). The frequency of SSC sampling was sufficiently discrete that when we tested for serial autocorrelation, a negative result was returned, thus no adjustment was necessary. Correlations among predictor variables were calculated in R software (Thurston, 2017). Correlated variables (p < 0.05), which could introduce spurious relations if input as covariates, were grouped to ensure that they would not be selected in the same model. A for-loop was constructed to cycle through the correlated groups, applying the ‘glmulti’ function for exhaustive candidate testing (Calcagno & Mazancourt, 2010) for each sub-catchment separately (Thurston, 2017). Akaike’s (1977) information criterion (AIC) was used to assess the relative goodness of fit for each candidate model, while avoiding overfitting (Burnham & Anderson, 2002), and statistics for the best models with similarly low AICs were compared.