Calibration of paired watersheds: Utility of moving sums in presence of externalities

Abstract

Historically, paired watershed studies have been used to quantify the hydrological effects of land use and management practices by concurrently monitoring 2 similar watersheds during calibration (pretreatment) and post‐treatment periods. This study characterizes seasonal water table and flow response to rainfall during the calibration period and tests a change detection technique of moving sums of recursive residuals (MOSUM) to select calibration periods for each control–treatment watershed pair when the regression coefficients for daily water table elevation were most stable to minimize regression model uncertainty. The control and treatment watersheds were 1 watershed of 3–4‐year‐old intensely managed loblolly pine (Pinus taeda L.) with natural understory, 1 watershed of 3–4‐year‐old loblolly pine intercropped with switchgrass (Panicum virgatum), 1 watershed of 14–15‐year‐old thinned loblolly pine with natural understory (control), and 1 watershed of switchgrass only. The study period spanned from 2009 to 2012. Silvicultural operational practices during this period acted as external factors, potentially shifting hydrologic calibration relationships between control and treatment watersheds. MOSUM results indicated significant changes in regression parameters due to silvicultural operations and were used to identify stable relationships for water table elevation. None of the calibration relationships developed using this method were significantly different from the classical calibration relationship based on published historical data. We attribute that to the similarity of historical and 2010–2012 leaf area index on control and treatment watersheds as moderated by the emergent vegetation. Although the MOSUM approach does not eliminate the need for true calibration data or replace the classic paired watershed approach, our results show that it may be an effective alternative approach when true data are unavailable, as it minimizes the impacts of external disturbances other than the treatment of interest.

Hydrologic Relationships of a Paired Coastal Watershed on the Santee Experimental Forest for Evaluating Watershed-scale Effects of Longleaf Pine Restoration on Water Yield

“…With both the calibration periods spanning just two years, the authors attributed the observed differences in the flow relationships between these periods to modification of soil properties during the operational silvicultural activities during clear-cutting and logging. In another study at the same NC site, Ssegane et al (2017) found statistically significant pre-treatment relationships, consistent with those from the 1988-1990 pre-disturbance level, using only 762 days and 608 days (for two treatment watersheds) of stable calibration period from 2009 to 2012 period that included some disturbances in the watersheds. Bren and Lane (2014) found a rapid increase in quality of calibration relationship as the record length increases but with no increase after 3-year length for all temporal scales of flow.…”

“…Ordinary least squares regression (OLS) was used to develop a calibration equation between the control and treatment watersheds and its significance test (Warton, Wright, Falster, & Westoby, 2006). However, since Durbin-Watson test (SAS, 2006) showed a positive autocorrelation of the monthly runoff on both the watersheds, we compared for differences between regression relationships using OLS versus using a resampling technique that accounts for serial correlation in time series and geometric mean (GM) regression, following (Ssegane et al, 2017), using the ts andlmodle2 R-statistical package (R Development Core Team, 2015). In this approach, the original data were resampled in pre-determined blocks 1000 times to estimate regression coefficients.…”

Response of Drainage Water Quality to Fertilizer Applications on a Switchgrass Intercropped Coastal Pine Forest

“…A large pump installed at the main roadside ditch (downstream of all weir outlets) minimized weir submergence during large storm events [9]. Other details of hydro-meteorological measurements were documented [12].…”

Untitled

“…Because the calibration Eq. 6 obtained by a bootstrap regression (Ssegane et al, 2017; that potentially reduces the effects of autocorrelation in daily values) was significant, we chose it to correct the bias in daily solar radiation data (Mj m -2 d -1 ) from the downloaded hourly data for all the stations.…”