Aquatic Invertebrate Abundance and Biomass in Arkansas, Mississippi, and Missouri Bottomland Hardwood Forests During Winter

Bottomland hardwood forests and associated floodplain wetlands historically covered more than10 million hectares in the Lower Mississippi Alluvial Valley (MAV; Reinecke et al. 1989; Fredrickson 2005; King et al. 2005). Wetland drainage, deforestation, and conversion of forests to agricultural and urban lands dramatically changed the MAV landscape and ecosystem during the 20th century (Sternitzke 1976; Schoenholtz et al. 2005). Today, less than 25% of the estimated bottomland hardwood forested area remains (Twedt and Loesch 1999; Ervin et al. 2006), but conservation initiatives are increasing the area of palustrine and riverine forested wetland systems (Cowardin et al. 1979 [Supplemental Material Reference S2]; Fredrickson et al. 2005).

Bottomland hardwood forests in the MAV have significant ecological, environmental, and economic values, because they are among the most productive forest communities on Earth and provide habitat for a great diversity of wildlife, including waterfowl (Fredrickson 2005; Heitmeyer et al. 2005; Mitsch and Gosselink 2007). At least eight species of ducks use bottomland hardwood forests seasonally, and these forests are particularly important to mallards Anas platyrhynchos and wood ducks Aix sponsa (Reinecke et al. 1989; Heitmeyer et al. 2005; Davis et al. 2007). Additionally, bottomland hardwood forests provide seasonal or year-round habitat for a great diversity of vertebrates and aquatic invertebrates (Wehrle et al. 1995; Batema et al. 2005; Heitmeyer et al. 2005).

Wintering mallards and wood ducks in the MAV forage on natural and agricultural seeds that provide carbohydrates and other nutrients (Delnicki and Reinecke 1986), but also ingest aquatic invertebrates primarily for protein (Dabbert and Martin 2000; Batema et al. 2005; Heitmeyer et al. 2005). Invertebrates nutritionally diversify carbohydrate-dominated diets and help build body mass, enable egg development, supply calcium for subsequent egg-shell deposition, and are important for development of basic plumage by prebreeding females (Heitmeyer and Fredrickson 1990; Richardson and Kaminski 1992; Barras et al. 1996). Additionally, invertebrate biomass estimates in flooded bottomland hardwood and other forested wetlands are necessary to quantify completely the energetic carrying capacity (i.e., based on energy from plant and animal foods) of these habitats for migrating and wintering waterfowl by the Lower Mississippi Valley Joint Venture (LMVJV; a collaborative, regional partnership of government agencies, nonprofit organizations, corporations, tribes, and individuals that conserve habitat for priority bird species, other wildlife, and people) partners of the North American Waterfowl Management Plan (Reinecke et al. 1989; Loesch et al. 2000).

Concomitant studies of invertebrate abundance and community composition in greentree reservoirs (GTRs; a forested tract surrounded partially or fully by a levee) and naturally flooded forests (NFF; an unimpounded bottomland hardwood forest) are available at a local but not a regional scale, such as the MAV (Reinecke et al. 1989; Wehrle et al. 1995; Batema et al. 2005; Fredrickson 2005). Therefore, our objectives were to 1) model spatio-temporal variation in invertebrate abundance during winter relative to selected explanatory variables (see Methods), (2) generate contemporary, precise (coefficient of variation [CV] ≤ 15%; Stafford et al. 2006, Kross et al. 2008, Straub 2012) estimates of invertebrate biomass in NFFs and GTRs at a landscape scale across parts of Arkansas, Mississippi, and Missouri in the MAV that would be useful for conservation planning and implementation by the LMVJV, and 3) provide management implications consistent with our results and others previously published.

Mingo National Wildlife Refuge (NWR) and Duck Creek Conservation Area are contiguous, and the U.S. Fish and Wildlife Service and Missouri Department of Conservation manages them, respectively (hereafter Mingo/Duck Creek). Mingo/Duck Creek covers 10,400 ha and is located in the northern MAV near Puxico, Missouri (36°N, 90°W). The site contains 7,000 ha of bottomland hardwood forests, the only remaining large tract of bottomland hardwood forests in the Missouri MAV (Heitmeyer et al. 1989; Supplemental Material S3). The GTRs at Mingo/Duck Creek underwent development during the 1940s (Fredrickson 2005) and currently contain approximately 60% red oaks Quercus palustris, Q. phellos, and Q. pagoda of similar age (70–80 y; Straub 2012).

White River NWR is located in west-central MAV near St. Charles, Arkansas (34°N, 91°W). White River NWR encompasses a 145-km stretch of the lower 160 km of the White River near its confluence with the Mississippi River. White River NWR contains about 62,300 ha of bottomland hardwood forests and other wetlands (Oli et al. 1997). Average stand age for trees within our study plots was 70–80 y old, and stands contained 30%–40% red oaks (e.g., mostly Q. texana; Straub 2012). No GTRs existed in White River NWR; hence, we confined our study to NFFs.

Delta National Forest is managed by the U.S. Department of Agriculture (USDA) Forest Service in east-central MAV and located 23 km southeast of Rolling Fork, Mississippi (32°N, 90°W). Delta National Forest contains over 24,000 ha of bottomland hardwood forests interspersed with palustrine wetlands and is the only national forest comprised exclusively of bottomland hardwoods in the United States (Lowney and Hill 1989). There are approximately 2,000 ha of bottomland hardwood forests managed as GTRs, which generally are flooded annually from mid-November to early February (Wehrle et al. 1995). The bottomland hardwood forest in the Sunflower GTR is estimated to be 75–80 y old and contains approximately 42% red oaks (Q. texana, Q. phellos; Straub 2012).

We also sampled GTRs and NFFs at the Sam D. Hamilton Noxubee NWR (Noxubee NWR), located outside the MAV in the Interior Flatwoods (IF) Region in east-central Mississippi, 25 km south of Starkville, Mississippi (33°N, 88°W; Pettry 1977). We sampled this site because of previous aquatic invertebrate research at this study site (Duffy and LaBar 1994; Wehrle et al. 1995). Major hydrological features of Noxubee NWR include the Noxubee River, its tributaries, and Bluff and Loakfoma Lakes. Noxubee NWR is approximately 19,400 ha and comprises approximately 18,000 ha of bottomland hardwood forest and upland forest with four GTRs, one from which we collected samples to compare results with those of Wehrle et al. (1995; GTR1). The GTRs date from the 1960s and generally flood annually from late November to mid-February. The forest in GTR1 is an estimated 65–75 y old and contains approximately 35% red oaks Q. pagoda, Q. phellos, Q. texana, and Q. nigra (Straub 2012).

We used the grts design option of the SPSURVEY package (Kincaid and Olsen 2011) in Program R 2.11.1 (R Development Core Team 2008) to select random plot centers within each GTR and NFF. We established ten 0.2-ha circular sampling plots within each GTR (when one existed) and within an associated NFF at each study area (Foth 2011; Straub 2012). Our a priori goal was to obtain four random samples within all plots per GTR and NFF per month (n  =  40/GTR and NFF/month; Table S1, Supplemental Material). Occasionally, some plots in NFFs were not inundated or incompletely inundated; consequently, we took one to three samples from flooded plots within NFFs to obtain greater than 20 NFF samples per month. We attempted to collect samples monthly at all study areas during November–February 2008–2010. However, lack of inundation, ice, or deep flooding precluded sampling some areas and months (n  =  19; Foth 2011).

We used a rectangular sweep net (23 cm × 45 cm, 500-µm mesh) to collect invertebrates from the substrate and water column (Cheal et al. 1993; Murkin et al. 1994; Wehrle et al. 1995; Gray et al. 1999). At each sample location, we also measured water depth in centimeters with a meter stick to relate depth to invertebrate abundance. We placed samples on ice at each site soon after collection and prior to transporting them to Mississippi State University. We stored all samples in a freezer at −10°C (Murkin et al. 1994; Stenroth and Nyström 2003). We used tap water for processing all samples, because other flotation media did not increase recovery of invertebrates from samples (Foth et al. 2012). We removed invertebrates by hand and identified them to Family (Pennak 1989; Merritt and Cummins 2008). We placed processed samples in an oven at 60°C for 18–24 h until they were dried to a constant mass and then weighed each Family (in micrograms) of invertebrates to extrapolate and estimate kilograms per hectare (Murkin et al. 1994; Foth 2011; Hagy and Kaminski 2012).

We used the glmmADMB package (Fournier et al. 2012) in R version 2.15 (R Development Core Team 2009) to model relationships between the mean number of invertebrates from two to four sweep net samples per plot or the number from a single sweep net sample when only one sample was obtained (hereafter, invertebrate abundance) and measured explanatory variables. We analyzed data at plot level, because we detected minor, yet significant, correlation among samples within plots (i.e., intraclass correlation coefficient within plots  =  0.314, P < 0.001) likely as a consequence of large sample size of sweep net samples across areas and years (n  =  791). We modeled invertebrate abundance rather than biomass (or transformation of mass data), because discrete counts of all invertebrates recovered from samples allowed more appropriate modeling techniques given strongly right skewed distribution of our data (Zuur et al. 2009). As explanatory variables, we selected sampling period (early winter [November–December] and late winter [January–February]), average water depth per plot (centimeters), percentage of red oak leaves among all intact identifiable leaves in sweep net samples (% RO), and species richness of trees in plots (Tree Richness).

Because our dataset had right skewed distributions with overdispersion, we fit abundance data with a generalized linear mixed model using the negative binomial distribution and log link function. The link function uses the natural logarithm of all raw data to linearize the relationship with measured covariates. However, we back-transformed (i.e., antilog) all parameter estimates because this allowed us to express our invertebrate data on the original scale of aggregate total invertebrates (Zuur et al. 2009). A priori, we formulated 12 ecologically important candidate models (Table 1) for possible explanation of variation in invertebrate abundance and ranked models according to Akaike's Second Order Information Criteria (AIC_C) to identify the best explanatory models or models (Akaike 1974; Burnham and Anderson 2002). We considered models competitive if AIC_C was within two delta AIC units of our top model (Burnham and Anderson 2002). We included YEAR as a categorical random effect to account for among year variation. We present parameter estimates (β), unconditional standard errors (SE), and 95% confidence intervals from the best model because only one model met our a priori AIC criterion.

For arithmetic mean estimates of invertebrate biomass at sites with GTRs and NFFs, we pooled data across GTRs and NFFs. However, we only used data from NFFs to generate the MAV-wide estimate, because GTRs comprise less than 2% of total area of bottomland hardwood forests in the MAV (Fredrickson 2005).

The best model explaining variation in invertebrate abundance included the interaction of time period within winter and water depth (Table 1). We did not consider other models competitive because they were greater than 25 AIC_C units from our best model. Our top model indicated that invertebrate abundance responded differently to water depth based on the time period; however, there consistently was a negative relationship (Figure 1). In early winter invertebrate abundance slightly decreased with increasing water depth (β  =  −0.0019, SE  =  0.005, 95% CI: −0.0126, 0.0087); in late winter this decrease was greater (β  =  −0.0522, SE  =  0.008, 95% CI: −0.0682, −0.0362). During both early and late winter periods, invertebrate abundance was greatest in depths ranging from approximately 10–20 cm of surface water (Figure 1).

In NFFs, Delta National Forest contained the greatest overall invertebrate biomass across both winters, whereas biomass was least at Noxubee NWR in both winters. White River NWR had the greatest invertebrate biomass in winter 2009–2010, but only 28 samples were obtained there because deep flooding limited our accessibility for sampling. The MAV-wide estimate of invertebrate biomass in NFFs for both years combined was 18.39 kg/ha and equaled our a priori goal for precision (CV  =  15%; Table 2). When we combined data across regions, sites, GTRs and NFFs, and years, the estimate of mean invertebrate biomass decreased to 6.6 kg/ha and precision increased (CV  =  9%; Table 2).

Our study was the first we are aware of to quantify invertebrate resources in bottomland hardwood forests at a regional scale during winter despite previous localized studies within the MAV and IF (papers cited in Batema et al. 2005). We found that invertebrate abundance decreased with increasing water depth during winter, with greatest abundances in early and late winter occurring in depths ranging from 10 to 20 cm and generally less than 30 cm. The trend lines associated with Figure 1, during late winter, also suggest that depths less than 10 cm might provide even greater invertebrate abundances, but this trend may be more associated with concentrated invertebrates as floodwaters recede and GTRs are drawn down post waterfowl season. The steeper decline in abundance during late winter also was reflected in invertebrate biomass (Foth 2011). More deeply flooded forests, such as GTRs, may promote anoxic conditions as winter progresses due to decaying plant matter, leaching of metals (e.g., iron) and tannic acid, and stagnation of impounded water, unlike NFFs with temporally dynamic flooding from hydrologic flows and allochthonous inputs of leaf litter and nutrients from overbank flooding (Batema et al. 2005). Additionally, increased invertebrate abundance and biomass in shallower waters during winter may be related to warming ambient water temperatures, increasing day length, nutrient release from decomposition of organic matter, or a combination of these and other factors (White 1985; Duffy and LaBar 1994; Manley et al. 2004; Hagy and Kaminski 2012).

An additional explanation regarding increased invertebrate abundance and biomass in more shallowly flooded forests may be related to our unit of measurement. We scaled invertebrate abundance data two dimensionally (i.e., invertebrates per square meter) to be consistent with previous literature and so estimates could be scaled up for conservation uses (i.e., abundance or biomass per hectare). Because we sampled and collected invertebrates only from the water column and not benthos, abundances may have been conservative but also greater in shallow water because organisms were concentrated more so than in deeper wetlands. However, a post hoc analysis revealed a similar negative relationship between volumetric scaling of invertebrate abundance (invertebrates per cubic meter). Therefore, we are confident our results are robust regardless of measurement units.

In the northern MAV, Mingo/Duck Creek's NFFs had greater invertebrate biomass than the GTR at that site (Foth 2011). This pattern may be related to an earlier seasonal transition into autumn at Mingo/Duck Creek, where managers flood impoundments earlier for fall waterfowl hunting seasons, thereby possibly creating anoxic conditions in impounded waters earlier in winter (Batema et al. 2005). The NFF plots at Mingo/Duck Creek had dried by midwinter and were not accessible by wintering waterfowl that could exploit and reduce standing crops of invertebrates. However, water remained in the Mingo/Duck Creek GTR during fall–winter and provided wintering ducks with access to invertebrates, perhaps also contributing to reduced invertebrate biomass in the GTR.

Similarly, in the southern MAV, the NFF at Delta National Forest contained greater biomass than its associated GTR. The GTR there was flooded in early November and drained in mid-February. Invertebrates likely colonized newly available wetlands in the GTR allowing their numbers to increase and plateau by late January. Conversely, water levels and frequency of flooding in the NFFs at Delta National Forest were temporally and spatially dynamic. Flood pulses were brief (e.g., 1–10 d), occurred three to four times per winter, and created localized ponding after recession of floodwaters. These dynamic conditions may have enhanced aerobic conditions and nutrient release from decomposing leaf litter and allochthonous inputs of organic matter and nutrients conducive to increasing invertebrate standing crops (Batema et al. 2005).

Unlike the MAV, the GTR at Noxubee NWR had greater invertebrate biomass than its adjacent NFF, which may have been a function of hydrology and water quality of the Noxubee River watershed and Bluff Lake, contiguous with this GTR. Noxubee NWR is in the upper reaches of the Noxubee River, and the surrounding hardwood bottomlands contribute significant allochthonous detritus (Vannote et al. 1980). During overflows, flood water disperses leaf litter and possibly invertebrates from the floodplain. During significant floods, riverine water enters the GTR from overbank flooding of the Noxubee River and terrestrial runoff, likely transporting leaf litter and nutrients to the GTR from the watershed. Thus, the impounded GTR may receive increased amounts of allochthonous detritus that serves as substrates and food for invertebrates, and when it combines with fresh, oxygenated water invertebrate abundance and mass may increase. Moreover, flowing water from Bluff Lake initially floods the GTR at Noxubee NWR each fall (Wehrle et al. 1995). This diversion of fresh water and associated nutrients and invertebrates may influence the increased standing crop of invertebrates in GTR1. However, none of these hypotheses have undergone testing, thus justifying future research.

Differences in flooding regimes and landscape agricultural practices between the MAV and IF probably influenced invertebrate population and community dynamics and biomass. Soil and water fertility is greater in the MAV compared to the IF (Wehrle et al. 1995). The MAV primarily contains fertile alluvial clays, whereas the Noxubee NWR flood plain has fine sandy loams (Pettry 1977). Also, MAV bottomlands are surrounded by agricultural lands, which typically contain greater nitrogen and phosphorus levels, sediments, and pollutants than IF sites (Stanturf et al. 2000). Inputs of nitrogen and phosphorus may influence algal and microbial growth and increase primary and secondary production. The IF region surrounding Noxubee NWR has had little row crop agriculture since the late 1980s and now primarily is range and forest lands, likely reducing nutrient inputs into the Noxubee River and Noxubee NWR lands (Kaminski et al. 2005; USDA 2007).

Seasonally dynamic invertebrate populations provide wildlife with abundant and renewed food during important annual life cycle events, such as prebasic molt of female mallards and wood ducks in late winter (Richardson and Kaminski 1992; Barras et al. 2001; Heitmeyer 2006). Although GTRs only contribute less than 2% to the overall area of bottomland forests in the MAV, these habitats provide food and other resources for waterfowl, especially in drought years, and waterfowl hunting opportunities (Fredrickson 2005). Moreover, sound red oak acorns persist through most winters if not depredated, and some species reach peak abundance in January (e.g., Nuttall oak Quercus texanii); thus, flooded GTRs and NFFs may provide important foraging habitats for ducks in mid to late winter when other wetlands may be depleted of foods (e.g., Leach et al. 2012; Straub 2012).

Previously, the LMVJV conservation planners had geographically limited data on invertebrate abundance and biomass and taxonomic composition from bottomland hardwood forests (Batema et al. 1985; Wehrle et al. 1995). Energetic and abundance estimates of waterfowl foods exist for the MAV (Kaminski et al. 2003; Stafford et al. 2006; Reinecke and Kaminski 2007 (S6 Supplemental Material); Kross et al. 2008; Straub 2012), but the current accepted value of 11.4 kg/ha for invertebrate biomass, by the LMVJV, in bottomland hardwood forests lacked spatial replication across the MAV. Our study provided a precise and contemporary estimate of invertebrate biomass in the MAV during winter. Thus, we recommend the LMVJV adopt 18.39 kg[dry]/ha as a revised estimate for invertebrate biomass in naturally flooded bottomland hardwood forests.

Water management that mimics natural hydrologic ebbs and flows in GTRs provides wet–dry pulses beneficial for invertebrate survival and reproduction (Wehrle et al. 1995; Batema et al. 2005). If fuel or other management costs preclude intentional fluctuations in water levels and durations in GTRs, a complex of GTRs that incorporates successive gravity-fed flooding between GTRs during winter may be logistically and fiscally efficient. If only a single GTR is present, removal of boards from water control structures during natural flood events and replacement of boards before flood cessation would help mimic dynamic hydroperiods. Additionally, flooding within GTRs should provide maximal coverage of depths less than 30 cm to facilitate foraging ducks' access to nektonic and benthic invertebrates, acorns, and other natural seeds (This study; Foth 2011; Hagy and Kaminski 2012).

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Data for analysis on samples collected from forested wetlands of Arkansas, Mississippi and Missouri, winters 2008–2009 and 2009–2010. Data are organized by year, round (November, NOV; December, DEC; January, JAN; February, FEB), study site (Delta National Forest, DNF; Mingo National Wildlife Refuge/Duck Creek Conservation Area, MINGO; Sam D. Hamilton Noxubee National Wildlife Refuge, NOX; White River National Wildlife Refuge, WR), flooding regime (naturally flooded forest, NFF; greentree reservoir, GTR), survey plot, compass azimuth, and invertebrate Family (count, weight, kg/ha, and g/m³).

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S1 (69 KB XLSX)

Reference S1. Classification of Wetlands and Deepwater habitats of the United States. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe. 1979. U.S. Department of the Interior, Fish and Wildlife Service, Washington DC, USA.

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S2; also available at http://www.co.monterey.ca.us/planning/major/Pebble%20Beach%20Company/Pebble_Beach_DEIR_Nov_2011/Pebble_Beach_DEIR_Admin_Records_Nov_2011/Cowardin/Cowardin_1979_%20wetland.pdf (78 KB PDF)

Reference S2. Water and habitat dynamics of the Mingo Swamp in Southeastern Missouri, Fish and Wildlife Research, 6 of series of technical reports. Heitmeyer ME, Fredrickson LH, Krause GF. 1989. U.S. Department of the Interior, Fish and Wildlife Service, Washington, DC.

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S3; also available at http://www.dtic.mil/dtic/tr/fulltext/u2/a322649.pdf (1977 KB PDF)

Reference S3. Development of objectives for waterfowl and shorebirds in the Mississippi Alluvial Valley. Loesch CR, Twedt DJ, Tripp K, Hunter WC, Woodrey MS. 2000. Pages 8–11 in USDA Forest Service Proceedings Rocky Mountain Research Station (RMRS-P) Publication 16, Ogden, Utah, USA.

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S4; also available at http://www.lmvjv.org/library/research_docs/2000%20RMRS-P-16_8-11%20Loesch%20et%20al.PDF (539 KB PDF)

Reference S4. Reinecke, K. J., and R. M. Kaminski. 2007. Lower Mississippi Valley Joint Venture, waterfowl working group memorandum. Subject: Final Revision of Table 5 (Duck Energy-days:DEDS) U.S. Fish and Wildlife Service, Vicksburg, Mississippi, USA.

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S5 (70 KB PDF)

Reference S5. 2007 Census of Agriculture, County Profile, Oktibbeha County, Mississippi. U. S. Department of Agriculture, National Agricultural Statistics Service,

Found at DOI: http://dx.doi.org/10.3996/092013-JFWM-061.S6; also available at http://www.agcensus.usda.gov/Publications/2007/Online_Highlights/County_Profiles/Mississippi/cp28105.pdf (378 KB PDF)

We thank A. Leach, K. Wigen, C. Asa, J. McInnis, K. Brock, and E. Interis for dedicated field assistance, processing samples in the laboratory, and providing continuous input. Additionally, we thank the USDA, Forest Service, Center for Bottomland Hardwoods Research, Stoneville, MS, the U.S. Fish and Wildlife Service, and Mississippi State University Forest and Wildlife Research Center (MSU-FWRC) for funding and additional support.

Additionally, we would like to thank the Journal reviewers and the Subject Editor for providing helpful input and improving our manuscript.

Our manuscript has been approved for publication as MSU-FWRC Publication No. WFA-WF389.

Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.