Upper Mississippi River Restoration Program Long Term Resource Monitoring

Results and Discussion

Average Sedimentation Rates During the Study Period

Poolwide mean rates (in cmּyr^­1) of backwater sedimentation over the entire length of the sampling transects during the 5-yr period were 0.04 (SE = 0.13) in Pool 4, 0.27 (SE = 0.11) in Pool 8, and 0.52 (SE = 0.20) for Pool 13. These rates often differed between aquatic and terrestrial portions, both within and among pools (Table 2). However, because of relatively large variances, we found no significant differences in mean rates among pools within aquatic or terrestrial portions (t < 1.8, df = 54, p > 0.08). We also found no significant differences between aquatic and terrestrial areas within pools (t < 1.5, df = 54, p > 0.15). 

The maximum poolwide estimate of 0.47 cmּyr^­1 (Pool 13) for aquatic portions of transects was lower than most rates presently used to evaluate effects of sedimentation in UMR backwaters (e.g., backwater life-expectancy). Considering that one of the objectives of this study was to determine if rates obtained from previous studies reasonably depict present day rates, this comparison is noteworthy. To better assess the differences between our rates and previous estimates, we compared mean rates of the past studies to the upper 95% confidence limits for our mean rates. We are limited to this method of comparison because past rates are often presented without estimates of variance, or past estimates were obtained from a biased sampling of backwater areas (e.g., sampled only in deep areas). The treatment of past estimates as true means in our comparison seems appropriate, given that the past rates have often been used without regard to the confidence in the estimate. Upper confidence limits for our poolwide estimate of rates (cmּyr^­1) in aquatic portions of transects for Pools 4, 8, and 13 ranged from 0.27 to 0.84, which are much lower than rates of between 1 and 4 cmּyr^­1 reported by McHenry et al. (1984).  Whether or not these differences are a result of the biased sampling designs of past studies, changes in sedimentation rates over time, or atypical rates during our study period is unknown. We will address the potential for atypical rates in our study later in this report.

The high variability we observed in rates within pools and within the aquatic and terrestrial portions of transects is undesirable for estimating a mean, but is interesting from a modeling standpoint. Explaining the variability among transects or within transects may be of greater importance to resource managers than obtaining mean rates at large scales (e.g., pool). One potential explanatory variable that may be useful is bed elevation.  In some lentic systems, relations between sedimentation and bed elevation can be strong, as sediments tend to accumulate at greater rates in deeper areas (i.e., sediment focusing; Hakanson 1977; Hilton et al. 1986). To examine this potential relation, we developed simple correlations between bed elevation and sedimentation rates at both transect and sample location levels.

Although we expected relations between sedimentation rates and elevation to be stronger at the location level, we began by looking for relations at the transect level because it would generate a simpler predictive model. Mean sedimentation rates were highly variable among transects (Appendix B). However, little of the variability in mean sedimentation rates among transects within pools could be explained by correlations with measures of mean, maximum, or selected quantiles of bed elevation at the transect level. The lack of a correlation may be because of differences in dominant sediment transport processes among backwaters, as water velocities can vary greatly among backwater areas. In addition, great differences in the relations between bed elevation and sedimentation rate may be present at the location level, thus making relations at the transect level difficult to detect.

To investigate the relation between sedimentation rate and bed elevation within transects, we determined mean poolwide sedimentation rates for eight bed elevation ranges for the period 1997 to 2001. The relation between sedimentation rate and bed elevation within transects differed between aquatic (negative) and terrestrial (positive) portions of transects (Figure 2). Bed elevation, as an interaction term with the aquatic or terrestrial portion of transect, had a significant effect on sedimentation rates when using mean rates from all three pools (F = 8.8, df = 1,20, p = 0.008).

The loss of water surface area (i.e., terrestrial encroachment) and the filling of the deepest areas of backwaters are two concerns of managers. Encroachment of terrestrial areas into aquatic areas was not apparent in Pools 4 and 8, as erosion occurred in the near-shore bed elevation range of -0.2 to 0 m. (Figure 2). High rates of sedimentation in the deepest areas (bed elevation <­0.6 m) were not found in these data. The deepest areas in all pools tended to have low sedimentation rates (<0.5 cmּyr^­1), with rates lower than 0.15 cmּyr^­1 in Pools 4 and 8. However, an exception to this occurred in the recent dredge cuts (bed elevation <-1.0  m) in Brown's Lake, Pool 13 where sediments accumulated at over 7.5 cmּyr^­1.

As indicated, we observed slight associations between bed elevation and average sedimentation rates within transects. This association may be confounded by effects of variables not included in our simple model, such as velocity and vegetation. Any effects of bed elevation may become clearer after adjusting for such effects, but we were unable to examine these effects because suitable data for these other variables do not exist.

Annual Sedimentation Rates During the Study Period

We considered annual discharge the most likely factor driving annual variability in sedimentation. Annual discharge varied greatly during our study (Table 3), allowing us to investigate this relation by comparing poolwide mean sedimentation rates for terrestrial to aquatic portions of transects during the highest (2001) and lowest (2000) discharge years.  Accumulation occurred in terrestrial areas and erosion in aquatic areas during the high discharge year of 2001, with the inverse occurring in 2000 (Figure 3). Differences between 2000 and 2001 are significant for terrestrial areas in all three pools (t > 3.1, df = 55, p < 0.005) and for aquatic areas in Pool 8 (t = 4.72, df = 52, p < 0.0001), assuming year effects are independent of pool effects. Differences between aquatic and terrestrial portions of transects were significant in 2000 and 2001 (t > 2.5, df = 55, p < 0.02), except for Pool 13 in 2000 (t = 1.5, df = 55, p = 0.13).

The observed difference in sedimentation rates for the lowest and highest discharge years suggests annual poolwide sedimentation rates are correlated with discharge. To see if a correlation existed across all discharge levels during our study period, we examined the association between discharge and poolwide annual sedimentation rates in aquatic and terrestrial areas for all years with sufficient data in each pool (n = 24 out of a possible n = 30). The effect of discharge was significantly associated with aquatic or terrestrial portion of transect in this model (F = 22.1, df = 1,20, p = 0.0001). Discharge (in thousands of m³ּsec^­1) effects on sedimentation rate (cmּyr^­1) were positive for the terrestrial portion of transects (slope = 0.57, SE = 0.15) and negative for the aquatic portion of transects (slope = -0.21, SE = 0.10). Based on the model, sedimentation rates are similar in both terrestrial and aquatic portions of transects at discharges near 3,000 m³ּsec^­1 (Figure 4). This model excluded pool effects and assumed no temporal correlation for annual means.

Using the associations we observed between annual sedimentation rates and annual discharge, we assessed whether our study period was a typical period of sedimentation, at least for discharge effects. Of concern was the occurrence of a high discharge year within our study period. For example in Pool 8, the year 2001 was the highest discharge year during the period 1959-2001, as measured by the discharge value exceeded 5% of the time within each year (Figure 5). Therefore, the mean sedimentation rates obtained for aquatic areas over our study period may be lower than long-term rates, considering our finding of net erosion in 2001. Furthermore, our study did not include a low discharge year; our lowest year (2000) was only in the lower 25% of occurrence. As a result, this may have also contributed to our rates potentially being lower than long-term rates, given our finding of high accumulation in lower discharge years. However, the effect of a low discharge year on sedimentation is unknown, as we did not have a drought year during our period of sampling.

The observed differences in annual mean rates of sedimentation in aquatic and terrestrial portions of transects may be a simplification of relations along a bed elevation gradient. To further assess these relations, we compared mean sedimentation rates for eight bed elevation ranges for each year by pool. Patterns for the lowest and highest discharge years were quite different, with negative correlation between bed elevation and sedimentation rate in the low discharge year of 2000, and positive correlation in the high discharge year of 2001 (Figure 6). Using annual rates for all years in study period, bed elevation was significant (F = 4.3, df = 24,60, p < 0.0001) as an interaction term (bed elevation*year). The significant effect was not present simply because of differences between 2000 and 2001, as bed elevation effects were still highly significant when 2001 data was excluded (F = 2.7, df = 17,44, p = 0.0041). These findings differ from the bed elevation effects observed in the average sedimentation rate over the study period, where bed elevation effects were only present when the portion of the transect (aquatic versus terrestrial) effect was included in the model.

Several patterns in the relations between sedimentation and bed elevation classes were similar across years. For example, we found low sedimentation rates (<0.2 cmּyr^­1) in near-shore areas over the study period during most of the study years. This suggests that discharge does not play a large role in determining sedimentation in these areas. Similarly, rates of sedimentation in the dredge cuts in Brown's Lake, Pool 13 were high (>6 cmּyr^­1) in all years, suggesting these areas are not susceptible to scour in high discharge years. Although discharge affected the relations between sedimentation and bed elevation as a whole, the effects may be limited to only some bed elevation classes along transects and can differ among transects.

The annual sedimentation patterns we observed can be attributed to several processes that drive sedimentation. Our findings of erosion in aquatic portions of transects in high discharge years (2001) suggests that, while the suspended sediment delivery may be high in the backwaters during high discharge, sediments are kept in suspension and transported out of the backwaters. Furthermore, previously accumulated sediments that formed the existing bed are probably resuspended because of high shear stress associated with high discharge periods. Accumulation of sediments in normally terrestrial areas during high discharge events can be attributed to the delivery of sediments during high water and the ability of terrestrial vegetation to reduce flows and allow deposition of suspended sediments.

In contrast, suspended sediments delivered to the backwaters in a low discharge year, albeit a lesser mass of sediments, are potentially accumulated because lower velocity conditions are present in backwaters. For the same reason, existing bed sediments are not eroded in low discharge years. In the normally terrestrial portions of transects, there is net erosion in low discharge years because accumulation during the maximum discharge period within low discharge years is not great enough to offset erosion occurring the rest of the year. Wind-generated wave action is a likely mechanism for erosion of near-shore areas.

More complex models, developed using our observed correlations as a foundation, show promise in strengthening our capability to predict sedimentation spatially and temporally.  Inclusion of other explanatory variables (e.g., water velocity) would further strengthen future modeling efforts. In particular, an approach that analyzes effects from different levels (e.g., pool, transect, and year) simultaneously seems appropriate. Such “mixed” statistical models will allow us to model sedimentation in UMR backwaters more explicitly.