Implications of Uncertainty in a Pre-treatment Dataset when Estimating Treatment Effects in Paired Catchment Studies: Phosphorus Loads from Forest Clear-cuts

Abstract

Estimates of increased nutrient export caused by forest clear-cuttings are mostly based on long-term paired catchment studies, where the treatment effect is calculated using pre-treatment dataset regression and post-treatment records of nutrient loads. In these studies uncertainty in the regression between the pre-treatment loads from the control and from the treatment catchments is typically neglected, even though it affects determination of the magnitude and duration of the treatment effect. This uncertainty is described in terms of variance in regression coefficients and residuals. The aim was to study how uncertainty in a pre-treatment dataset is propagated to estimates of the treatment effects using two sets of paired catchment data in eastern Finland, where the total phosphorus loads following forest clear-cutting and site preparation were investigated. The results showed that neglecting the uncertainty in the pre-treatment data did not change the conclusions when the treatment effect was strong enough. However, when the treatment effect was small, neglecting the uncertainty in the pre-treatment dataset easily leads to over-interpretation of the results. The uncertainty in a pre-treatment dataset should be taken into account in the paired catchment studies in order to avoid bias in the estimation of management effects on the loads.