Upper Mississippi River Restoration Program Long Term Resource Monitoring

Methods

Sedimentation monitoring was conducted in backwater lakes, as defined by Wilcox (1993), in Navigation Pools 4, 8, and 13 of the UMR (Figure 1). Backwaters are much less prevalent downstream of Pool 13; therefore, pools below Pool 13 were excluded from this study.  Impounded areas within the study pools were excluded from the study area because, while these areas are considered backwaters, these areas are typically more lotic environments and not the focus of this study. Lake Pepin in Pool 4, a large tributary delta lake, was also excluded from the study area because it is atypical of backwater lakes of the UMR. The spatial extent and stratification of the study area were delineated using a Geographic Information System (GIS) database of aquatic areas in the UMR (Owens and Ruhser 1996).

Transects for sampling were aligned perpendicular to flow and allocated across three strata. The large backwater lake stratum included lakes greater than 40 ha. Each pool had six large backwater lakes from which we randomly selected two transect locations in each lake, for a total of 12 transects. The remaining smaller backwater lakes were divided into two additional strata (low and high connectivity) based on connectivity metrics, such as the percent of their perimeter connected to a channel. We randomly selected 13 transects within these two strata based on the size of the strata in the pool. Of the 75 transects selected for survey, 63 transects were surveyed sufficiently to be included in the analysis presented here (Table I). The size of the study area represented by transects was 2,199 ha in Pool 4, 1,976 ha in Pool 8, and 2,780 ha in Pool 13.

We obtained net sedimentation rates annually between 1997 and 2001 by repeated annual measurement of the bed elevation at fixed sites along each transect. Vertical and horizontal benchmarks were established at each transect to allow replication of surveys. To ensure that we did not undersample short transects or oversample long transects, we selected the distance between sample locations along each transect based on the length of each transect. Transects longer than 600 m were sampled at 30.4-m intervals; transects between 150 and 600 m in length had 15.2-m intervals; and transects shorter than 150 m had 7.6-m intervals. Sampling intervals along terrestrial portions of transects were 1.52 m. The terrestrial portion of transects extended 15.2 m shoreward of the posts set near each shore, or ending sooner if a physical obstruction was present (e.g., rock embankment). The terrestrial portions were surveyed using differential leveling (Brinker and Wolf 1984) to determine bed elevation relative to the benchmark.

Most aquatic surveys were conducted over the ice to increase the accuracy of the measurements. Sampling sites along transects were located by tape measure using the established benchmarks and realigning every 90 m with an electronic distance measuring device.  Water depths were measured with a surveyor's tape attached to a somewhat buoyant pole with a 20-cm diameter plate at the end to retard sinking in soft sediments.  Given that soft sediments are uncommon in backwaters of these pools (<2% of the area has sediments with moisture content >80%; Rogala 1996), we assumed only minor problems associated with sinking of the pole. These problems could result in underestimating sedimentation rates of unconsolidated sediments at some locations, but given the transient nature of unconsolidated sediments in a river system, we further assume minor impacts to our estimates of sedimentation. The water elevation relative to the benchmark was determined each year of survey by differential leveling. The aquatic portions of some transects were sampled during ice-free conditions using the differential leveling method for shallow or short transects. Maps illustrating the location of transects are included in Appendix A.

Bed elevation relative to a reference water surface elevation was calculated for each sampling of selected locations to allow comparisons among transects. Using U.S. Army Corps of Engineers daily discharge records (U.S. Army Corps of Engineers 2002a,b), a low water condition that is exceeded 90% of the time was used as the reference elevation for each pool. Sedimentation rates at each sampling location were calculated as the difference in bed elevation over two surveys, divided by the period between the two surveys. These periods could be individual years or multiple years.  When comparing changes over selected periods, rates were determined using only sampling locations where data were present for all periods of comparison. We excluded data from recently dredged areas in Brown's Lake, Pool 13, when estimating poolwide mean rates because of abnormal sedimentation processes in these areas. The data from dredged areas were only used to estimate rates in the dredged areas themselves. 

The rates over an entire transect are of little interest because the ecological significance of sedimentation would generally be dependent on whether the sedimentation was in terrestrial or aquatic areas. Therefore, we also determined mean sedimentation rates for aquatic and terrestrial portions of transects by post-stratifying on the basis of positive or negative bed elevation (relative to the reference condition) in the first year of the period of comparison. As a result, for annual change detection, a given site might be classified as terrestrial one year and aquatic the next. Additional post-stratification was used in a similar manner to investigate rates of sedimentation along shorter segments of transects (e.g., shallow near-shore segments based on bed elevation).

Data were analyzed to determine mean sedimentation rates, standard errors, and significant differences between means rates using analyses that addressed both the stratification and clustering within the study design. We used survey software (SAS^® PROC SURVEYMEANS; SAS 2000) to adjust for potential correlation within transects.  This software assumed equal (rather than spatial) correlation within transects. This correlation assumption is an approximation for these potentially spatially correlated data, but we assumed this approach to have only minor effects on standard error estimates. Weighting was used to account for the disparities in sampling location intervals along and among transects and surface areas among sampling strata. Differences in means were tested using contrast statements provided by the SAS^® macro, smsub (available at http://ftp.sas.com/techsup/download/stat/smsub.html), an extension of SURVEYMEANS.  We used a Bonferroni correction (α = 0.05 ÷ number of comparisons) to adjust for multiple comparisons. Data were not summarized for strata used in the allocation of sampling, as sample size precluded such analyses.

We used a model-based approach to investigate correlations between sedimentation rates and predictor variables, such as bed elevation and discharge. Using linear mixed modeling software (SAS^® PROC MIXED; SAS 2000), we modeled relations between predictor variables and mean sedimentation rates of levels (e.g., pool, year, elevation class) produced by SURVEYMEANS. Bed elevation class means for sedimentation were graphically depicted by the center of the category range, but the mean bed elevations within the depth category were used in models to assess bed elevation effects. We assumed independence across depth category means within pools. (The range of spatial correlation for these sedimentation data appeared to be approximately 25 m [J. Rogala, unpublished data]. Because aquatic section means typically derive from groups of observations with combined length of >25 m along a transect, this assumption seems reasonable for means from that section, but may be liberal for means from the terrestrial section groups that typically have a combined length of <5 m.)  Discharge effects on annual sedimentation were investigated using a high discharge condition that was represented by the discharge value exceeded 5% of the time within each study year; daily discharge data were obtained using U.S. Army Corps of Engineers discharge records (U.S. Army Corps of Engineers 2002a,b).