Thermal regime of a headwater stream within a clear-cut, coastal British Columbia, Canada

Abstract

Abstract:This study examined the thermal regime of a headwater stream within a clear-cut. The stream had a complex morphology dominated by step-pool features, many formed by sediment accumulation upstream of woody debris. Maximum daily temperatures increased up to 5°C after logging, and were positively associated with maximum daily air temperature and negatively with discharge. Maximum daily temperatures generally increased with downstream distance through the cut block, but decreased with distance in two segments over distances of tens of metres, where the topography indicated relatively concentrated lateral inflow. Localized cool areas within a step-pool unit were associated with zones of concentrated upwelling. Bed temperatures tended to be higher and have greater ranges in areas of downwelling flow into the bed. Heat budget estimates were made using meteorological measurements over the water surface and a model of net radiation using canopy characteristics derived from fisheye photography. Heat exchange driven by hyporheic flow through the channel step was a cooling effect during daytime, with a magnitude up to approximately 25% that of net radiation during the period of maximum daytime warming. Heat budget calculations in these headwater streams are complicated by the heterogeneity of incident solar radiation and channel geometry, as well as uncertainty in estimating heat and water exchanges between the stream and the subsurface via hyporheic exchange and heat conduction.

Corrected prediction intervals for change detection in paired watershed studies

“…Though Harr et al (1979) correctly produced prediction intervals in the OLS setting and the above studies are significant developments for the use of temporally autocorrelated hydrological data, the intervals created by Watson et al (2001), Moore et al (2005) and Gomi et al (2006) are not prediction intervals. Their intervals are based on an estimate of the innovations variance, and do not account for prediction variance that includes variation due to the estimation of linear model and autoregressive function parameters, and covariance between observations (autocorrelation).…”

“…To address the limitations that annual or stormbased data pose for change detection in watershed studies, researchers began data collection on daily and monthly time scales (Scott and Lesch 1997, Watson et al 2001, Moore et al 2005, Gomi et al 2006. Seasonal periodicity and serial autocorrelation can be present in data of measurements collected over finer time scales.…”

“…The percentage of these posttreatmentγ j values that exceeded the prediction limits was computed, and a significant treatment effect was declared if more than 5% of these values exceeded the 95% prediction limits (Watson et al 2001). Moore et al (2005) and Gomi et al (2006) modified the estimation and analysis method proposed by Watson et al (2001) using daily stream temperature data to evaluate the headwater catchment responses to clear-cut harvesting in coastal British Columbia, Canada. These researchers correctly recognized that autocorrelated data not only require attention in the analysis of residuals, but additionally in the estimation of linear model parameters from equation (1).…”

“…In addition to recognizing that the covariance structure should be incorporated into the linear model parameters estimator, Moore et al (2005) and Gomi et al (2006) also recognized that using the partial autocorrelation function (PACF) can aid in determining the appropriate order of the autoregressive process. After determining the autoregressive structure and estimating the linear model and autoregressive parameters for the calibration period data, the residuals, innovations, prediction intervals, and assessment of treatment effects were computed following the methods of Watson et al (2001) described above.…”

“…To continue the prediction interval change detection methodology in the presence of autocorrelated data, researchers began to filter the model residuals via autoregressive (AR) analysis and use the estimated variance of the AR disturbances to construct prediction intervals and assess treatment effects (Watson et al 2001, Moore et al 2005, Gomi et al 2006. This method ignores variation attributed to the estimation of autocorrelation model parameters, and changes in prediction precision due to where the control watershed response values lie in relation to their range during the calibration period.…”