Directing Succession through Adaptive Management in National Wildlife Refuges Reed Canary Grass Control and Transition to Wetland Forests and Meadows

Four experiments will be conducted to improve the model and reduce biological uncertainty associated with management of RCG-invaded sites, and each experiment addresses a different commo

National Wildlife Refuges: Reed Canary Grass Control and

Transition to Wetland Forests and Meadows

USGS Biological Resources Division USFWS Biological Monitoring Team

Study Manual Draft Version 2007

USGS Contacts:

Franklin Percival, Florida Cooperative Fish & Wildlife Research Unit, Dept of Wildlife Ecology & Conservation, University of Florida, PO Box 110485 Gainesville, FL 32611-0485, (352) 846-0543 (voice); percivaf@ufl.edu

USFWS Regional Contacts:

Hal Laskowski, USFWS Region 5 Regional Biologist, 11978 Turkle Pond Road, Milton, DE 19968, (302) 684-4028 (voice), (302) 684-8504 (fax): Harold_Laskowski@fws.gov

Frank Durbian, Squaw Creek National Wildlife Refuge, PO Box 158, Mound City, MO 64470,

660-442-5754 XT 17 (voice); 660-442-5248 (fax); Frank_Durbian@fws.gov

Rachel Laubhan, USFWS Region 6, Northern Prairie Wildlife Research Center, 8711 37 th Street, SE, Jamestown, ND 58401, (701) 253-5543 (voice); (701) 253-5553 (fax); Rachel_Laubhan@fws.gov

Science Team Contacts:

Carrie Reinhardt Adams, Dept of Environmental Horticulture, University of Florida, 107 Building 68,

PO Box 110675, Gainesville, FL 32611-0675, (352) 392-1831 XT 223 (voice); (352) 392-1413 (fax); creinhardt@ifas.ufl.edu

Susan Galatowitsch, Dept of Horticultural Science, University of Minnesota, 305 Alderman Hall, St Paul, Minnesota 55108, 612-624-3242 (voice); 612-624-4941 (fax); galat001@umn.edu

Eric Lonsdorf, Davee Center for Epidemiology and Endocrinology Lincoln Park Zoo

Chicago, IL 60614, 312-742-7216 (voice); 312-742-7220 (fax); ericlonsdorf@lpzoo.org

Clint Moore, USGS Patuxent Wildlife Research Center, Warnell School of Forestry and Natural

Resources, University of Georgia, Athens, GA 30602-2152, (706) 542-1609 (voice); cmoore@usgs.gov Julie Sorenson, Dept of Environmental Horticulture, University of Florida, 101 Building 68, PO Box

110675, Gainesville, FL 32611-0675, (352) 392-1831 XT 367 (voice); (352) 392-1413 (fax);

jsorenson@ufl.edu

Executive Summary

Invasive species present a challenge to the efforts of National Wildlife Refuges (NWRs)

to preserve appropriate plant community habitat Decision making for control of invasive speciesand subsequent revegetation efforts is clouded by a number of uncertainties Coordinating

vegetation management efforts across National Wildlife Refuges will reduce this uncertainty and promote effective decisions to achieve management goals

According to information provided by NWRs located in U.S Fish and Wildlife Service Region 3 (Midwest Region) and Region 6 (Mountain-Prairie Region), reed canary grass

(Phalaris arundinacea, RCG) has partially or heavily infested approximately 37,400 acres of

wetlands within these refuges This invasive plant species adversely affects native plant

communities, and decreases suitable habitat for target faunal species To improve the

management of RCG and assist in the recovery of these degraded ecosystems within these NWRs across Region 3 and Region 6, an adaptive management (AM) framework will be

utilized Through AM, the goal of this project is to generate the information needed for refuge

managers to make good and defensible decisions about when, where, and how to treat RCG for

purposes of maintaining or restoring target communities and the wildlife they support (from

RCG Workshop Problem Statement, July 2006, Williams et al 2007)

Refuges will implement management actions in a coordinated effort and observe

outcomes in a standardized way This information will improve a frame-based model to increase our predictive understanding of RCG and target native plant communities Additionally, this information will assist in management decision-making for RCG-dominated sites and facilitate establishment of the target ecosystem

Four experiments will be conducted to improve the model and reduce biological

uncertainty associated with management of RCG-invaded sites, and each experiment addresses a different commonly encountered site condition Two experiments focus on floodplain forest ecosystems The first addresses a nearly monotypic stand of RCG surrounded by desirable floodplain forest vegetation, and assesses the effect of seed rain from the surrounding forest, as well as restoration seeding, on revegetation following RCG eradication The second experiment addresses a nearly monotypic RCG understory overtopped by a desirable tree canopy overstory, also explores the effect of seed rain and restoration seeding on woody species establishment afterRCG eradication Two experiments target wet meadow ecosystems One focuses on determining the extent to which management can shift vegetation composition of a meadow partly invaded byRCG towards wet meadow vegetation Sites will be treated with herbicide or mowing, depending

on the ratio of meadow vegetation: RCG The second experiment concerns RCG-dominated sites, and compares shifting a heavily infested mixed meadow to meadow by eradicating all

vegetation and reseeding, or by managing the existing vegetation Refuge biologists, with the

assistance of the science team, will determine which experiments will be conducted at their NWR

Response data collected by refuge personnel will be uploaded to the project website (http://hort.ifas.ufl.edu/rcgam), and will be accessible to other participants in the study The goal

of this study is to use the data collected by the refuge personnel to improve the frame-based model and reduce biological uncertainty associated with RCG management and the

establishment of target ecosystems The employment of the AM process “provides a framework for making good decisions in the face of critical uncertainties, and a formal process for reducing

uncertainties so that management performance can be improved over time” (Williams et al 2007) The information learned from this project will be used to decrease uncertainties regardingRCG, and increase the knowledge base regarding the management of RCG and the establishment

of target ecosystems, both at the individual NWRs, and across Region 3 and Region 6

Table of Contents

Executive Summary 2

Model Summary 5

Experimental Design and Field Methods 7

Managing RCG-invaded Floodplain Forests 7

Experiment A: Floodplain Forest: Seed rain and active seeding 7

Experiment B: Floodplain Forest: Understory removal 11

Managing RCG-invaded Wet Meadows 14

Schedule of Refuge Participant Involvement 18

Literature Cited 22

understanding of RCG-invaded forests and meadows based on manager knowledge Frames

(WM, Mixed, RCG or Bare) represent vegetation states with distinctly different management

needs For example, the logical management for a monotypic stand of RCG is very different

than for a recently invaded wet meadow The model will integrate deterministic biological

processes (growth, competition, seed dispersal) and management (burning, mowing, chemical application) processes with stochastic environmental events (floods, drought) to predict the

change in vegetative response to management actions Furthermore, the model will incorporate the consequences of external drivers, which we consider attributes that influence the interaction between management and biological process, e.g site specific soil nutrients, hydrology, etc

By organizing our best knowledge, the model also serves to guide the entire adaptive

management process In other words, the model is the tool that integrates the multi-refuge

experiments that address biological uncertainty reduction and projects how uncertainty reductioninfluences efficient management choices For example, the biological uncertainty related to the effects of native plant cover and diversity on RCG spread currently limits the predictability of

particular management actions Prior to monitoring, we can use the model to predict outcomes

of management, where the confidence in our predictions reflects our current uncertainties After coordinated multi-refuge monitoring we can: a) quantify the relationship between diversity and RCG growth, b) integrate the updated information into the model to determine how much better our predictions have become and c) determine whether more monitoring is needed or if the

decision-making process can be revised with the increase in knowledge from our coordinated

efforts The adaptive management process accommodates an iterative return of information and assessment in response to decision making By accounting for the biology and organizing the

uncertainties, the model exposes and places the effects of uncertainty in biological process,

guides the choice of experiments or monitoring that best reduce uncertainty and then provides a tool for choosing management actions

Ultimately, the interplay of monitoring and experiments with the frame-based model should

provide a transparent and adaptively evolving plan to guide efficient management of RCG Withsmall sample sizes, one cannot differentiate the effects of stochastic events (floods) from

deterministic actions (management) and without the model, one cannot project the consequences

of mistaking one for the other Thus, the model and coordinated management enhance the

information that is already gathered and galvanize, rather than paralyze, the ability to make

confident, rational and transparent choices

Figure 1 The frame-based model

Figure 1 A prototype of the frame-based model developed during an initial coordinating meeting with FWS Regions 3 & 6

Frames (WM, Mixed, RCG or Bare) represent vegetation states with distinctly different management needs Numbered arrows

designate transitions between these states

(inc.

annuals)

RCG RCG/WM

Soil Characteristics Hydrology

Site History Landscape

Biological Processes

External Drivers

Wet Meadow

Floodplain Forest (FF)

Mixed RCG/ FF

RCG

Hydrology

Site History Landscape

Biological Processes External Drivers

Soil Characteristics

Floodplain Forest

Bare Soil (inc.

Experimental Design and Field Methods

This experimental design addresses commonly encountered RCG management scenarios and related biological uncertainties as discussed by USFWS Regions 3 and 6 personnel Habitat typestargeted by this study are typical of RCG-invaded ecosystems identified by potential project participants

Management treatments and treatment response monitoring are described briefly below

Methodology will be detailed in a photo-based methods guide that will be available via the project website

Managing RCG-invaded Floodplain Forests

Competition from invasive perennial grasses restricts forest regeneration in different kinds of ecosystems worldwide, limiting the success of forest restoration Not surprisingly, RCG is often implicated in the lack of forest regeneration in floodplain systems in the Midcontinental US Because the risk of RCG reinvasion is typically very high along waterways, knowing how to restore a system so the forest canopy closes before RCG re-establishes is a key concern

Restoration decision-making is currently problematic because of uncertainties related to the regeneration dynamics of the system Controlling RCG in floodplain forests is also complicated

by the unpredictability of major flood events

The following experiments address the uncertainties related to establishing forest vegetation following RCG eradication Experiment A addresses a nearly monotypic stand of RCG (no/few desirable woody species) surrounded by desirable floodplain forest vegetation The first step of the experiment involves eradication of existing RCG, resulting in bare soil (model transition

#12) The second step involves establishment of desirable tree canopy on the bare soil (model transition #4) Experiment B addresses a nearly monotypic RCG understory which is overtopped

by a desirable tree canopy overstory The goal is to eradicate the RCG while maintaining existingtree canopy, and to subsequently encourage establishment of desirable trees (model transition

#10), either by relying on recruitment from the surrounding forest vegetation, or by

seeding/planting appropriate species

Throughout experiments A and B, Rodeo may be substituted for Roundup Ultra where use formulation is appropriate, i.e there is standing water 150ft from the treatment area

aquatic-Experiment A: Floodplain Forest: Seed rain and active seeding

Site Selection Criteria

Sites should be at least 5 acres in size and range from 5-10 acres where the desirable community

is floodplain forest, but RCG is the most dominant species (a nearly monotypic stand), with no/few tree species present within the RCG stand Floodplain forest vegetation may be

dominated by cottonwood, silver maple, and willow The RCG stand should be surrounded by a fairly intact floodplain forest, from which seed recruitment of desirable woody species can be expected (but interior RCG areas should be large enough such that a portion of the site would notreceive seed rain)

Sites can be part of a larger floodplain forest system, but sites should be single units in which theentire area is treated the same way Participating refuges should select sites with regard to

elevation: sites may be located on the channel edge of floodplains (frequent flooding), or at higher floodplain positions (rare/never flooded) A single refuge may choose to select sites at varying elevations, if units are available

Baseline Site Characterization

During site selection field visits in summer 2007, a soil profile description (from a probe) should

be made to determine site characteristics with respect to landform and long-term hydrology Notethe flooding frequency of the site from any information available (for example, flooded 3 times

in the past 10 years), and determine the elevation of the site The edge of the management unit (site) should be recorded with a GPS (the edge here is defined by the border of the RCG stand)

During the 2007 visit, the location all water inflows and outflows from the site should be noted with arrows on a topographic map Also, a preliminary assessment should be made of off-site RCG propagule sources (note if RCG seed sources are hydrologically connected to the site) Soil samples should be collected for nutrient/texture analysis (see details below) An annotated floristic list should also be compiled during this site visit, noting all species above 5% cover This floristic list will be useful for refining a vegetation sampling strategy and as a reference for the seedbank assay

For soil characterizations, soil should be collected at 3 sampling locations/acre These locations should be selected to represent the variation in soils and vegetation Site locations should be recorded with a GPS At each sampling location, 1 L of soil should be collected with a shallow corer (e.g., tulip-bulb planter) to a depth of approximately 10 cm Core samples from each location should be homogenized in a bucket, deposited in two airtight bags (each containing 0.5

L of soil) One soil sample should be stored frozen (used for texture analysis), and one sample should be dried and sieved through a 2mm mesh screen (used for nutrient analysis) The dried and sieved sample will be sent to University of Florida Analytical Research Lab for a

Floodplain forest community

(trees present)

Large area

of mostly RCG

Recruitment from forest?

Figure 2 Aerial view of potential site with overstory of desirable trees and RCG understory (schematic diagram of appropriate site for experiment A)

determination of total C, total N, pH The frozen 0.5 L sample should be sent to UF Soils

Judging

Experimental Design and Management Treatments

Management treatments for this experiment are divided into 2 consecutive steps Step 1 will be the same at every site, and step 2 may vary with site

Step 1: Control RCG, resulting in bare soil

To control RCG, glyphosate herbicide should be broadcast applied (Monsanto’s Roundup Ultra, 41% glyphosate isopropyl amine salt, 2% solution

applied 187 L/ha, see label specifications)between September 12 and October

30 2007, or following the first hard freeze of the season (0ºC or lower for 2-4 hours or more) Another broadcast glyphosate application in mid-late April 2008 (or when RCG germination and growth begins) will target recolonizing RCG seedlings An ATV may be needed for broadcast spraying to move easily over debris on the forest floor

To estimate RCG rhizome viability, a rhizome check will be performed in April 2008 with a shovel (semi-quantitative) in a subset of 3 randomly chosen sampling points If this check shows that RCG rhizome mortality is low, repeat glyphosate herbicide

application between September 12 and October 30 2008 If rhizome mortality is high, another fall herbicide application is not needed at this time, and proceed to Step 2 for Spring 2008

Step 2: Revegetation of target woody species

Once RCG control treatments are complete, there are two possible treatment options for revegetation of target woody species, which will take place in April 2009 (or April 2008

if one herbicide application produces significant rhizome mortality) The appropriate revegetation treatment will be based on feasibility and site-level factors, including

composition of surrounding plant community

A Planting Strategy (select one):

1 Seed bare soil with target plant community For sowing in April, seed should be obtained during the previous fall Note the sowing rate of seeding treatments (seeds/m2)

2 Do not plant or seed and rely on natural recolonization from surrounding

floodplain forest

B Follow-up control strategy: spot apply recolonize RCG with glyphosate

Management treatments and treatment response monitoring (see below) may be delayed by high water events, but should be implemented after water levels recede

Treatment Response Monitoring

Prior to treatments in April 2008, three observation wells and one staff gauge (at lowest elevation

on treatment site) should be installed on each site for water table and level measurements An site or nearby pre-existing staff gauge may be used for this purpose If a site does not have the potential to flood, a staff gauge may not be necessary Appendix A includes the USGS protocol for observation well construction and installation Three to five photostations should be

on-established at vantage points on the perimeter of each site If possible, observation wells/staff

gauges should be located down hill from photostations (this leaves 2 photostations without corresponding wells/staff gauges) These should be marked with 8 foot steel T-posts, painted or treated to be highly visible If the addition in vertical structure associated with the posts is

unacceptable, other existing landmarks may be used for vantage points, or posts may be located

in areas where vertical structure already exists Prior to revegetation efforts, water table

measurements should be made 3 times/yr (April, August, November) and after extreme events After initial RCG eradication (e.g., Step 1 above) water table measurements should be made every month from April-October, and after extreme events Photos from photostations should be taken mid summer annually in an internally consistent manner noting the direction of image

Vegetation composition (herbaceous vegetation and seedlings/saplings) will be surveyed prior to treatments and annually at mid-summer to determine changes in response to treatments For each acre of a site, at least ten points should be randomly selected to be vegetation sampling points To assess the influence of seed rain from the surrounding forest, vegetation sampling points will be stratified across 4 categories of increasing distance from edge of floodplain forest, e.g 5, 15, 45, and 150 m (or maximum distance) from edge (see Appendix B) Any atypical areasshould be excluded from the randomization

Sampling points should be monumented with large diameter PVC recessed into the ground to create a receptor for a conduit pole that would be inserted during data collection The PVC receptor will be flush with ground level and covered with a PVC cap which will be even with thesoil surface (this marker should not interfere with mowing/herbicide treatments) The receptor would be marked with GPS and will be lined with iron to allow location with a magnetometer

Every sampling point is the southwest corner of a large plot (10m x 15m) for cover of all plants above 5% cover (and RCG cover, regardless of cover class), and a small nested plot 5m x 5m for seedling and sapling (trees only) counts with heights (by species) See Appendix C for plot specifications For each plot the total cover will be recorded using cover classes: <1% 1-5%, 5-25%, 25-50%, 50-75%, and 75-100% (species may overlap in cover, such that total cover in any plot sums to more than 100%) The stem density of RCG (rhizome-based shoots and seedlings) will also be estimated and recorded using the following classes: 0, 1-10, 10-50, 50-100, 100-200,200-500, 500-1000, 1000-2000, more than 2000

Seed traps will be used to measure seed rain Seed traps will consist of fine-mesh hardware cloth suspended in an inverted pyramid below 0.5 x 0.5 m wood frames raised above the ground by

0.5 m stick (see Holl, K D 1999 Biotropica 31: 229-242) These traps will be placed in

receptors at node sampling points along a transect, with three traps each set at 5, 15, 45, and 150

m (or maximum distance) from forest edge Traps would be placed out from April to August (seed shed ranges from spring to summer), and will be checked for seed every two weeks All visible, apparently healthy woody species seeds will be separated from leaves and other debris, identified, and counted

Changes in seed bank composition will be important for characterizing the remnant RCG

population following eradication as well as for determining changes in on-site woody plant propagule availability Seed bank samples should be taken after fall 2007 RCG control efforts (inNovember 2007) Soil should be collected at 5 sampling locations/acre These sampling locationsshould be several distances out from the forest edge, i.e 5, 15, 45, and 150 m (or maximum

distance) from edge (see Appendix B) At each sampling location, from within a 5m x 2m plot (this is the smaller plot from vegetation sampling), 2 L of soil should be collected with a shallow core sampler (e.g., tulip-bulb planter) to a depth of approximately 10 cm (about 5 cores) Core samples from each location should be homogenized in a bucket and a 1L subsample should be deposited in an airtight bag The sample should be stored cold (10 C), but not frozen Seedbank species composition (density for each species) should be determined with a germination assay (Boedeltje et al 2002) in winter 2007/8 Seed bank samples will be grown out for 5 months (December 2007-April 2008) To diagnose change in RCG seed bank, samples should also be taken and assayed in November 2008 and November 2009.

Experiment B: Floodplain Forest: Understory removal

Site Selection Criteria

Sites should range in size from 1-10 acres of floodplain forest with RCG as the most dominant species (a nearly monotypic stand) in the understory, overtopped by an overstory of desirable tree species (50-100% cover) Floodplain forest vegetation may be cottonwood or silver maple dominated

Sites can be part of a larger floodplain forest system, but sites should be single units in which theentire area is treated the same way Participating refuges should select sites with regard to

elevation: sites may be located on the channel edge of floodplains (frequent flooding), or at higher floodplain positions (rare/never flooded) A single refuge may choose to select sites at varying elevations, if units are available

Baseline Site Characterization

During site selection field visits in summer 2007, a soil profile description (from a probe) should

be made to determine site characteristics with respect to landform and long-term hydrology Notethe flooding frequency of the site from any information available (for example, flooded 3 times

in the past 10 years) The edge of the management unit (site) should be recorded with a GPS (the edge here is defined by the border of the area to be treated with RCG control and revegetation)

Region 3/6 Reed Canary Grass Study

Understory of

continuous RCG

Overstory of desirable

overstory of desirable trees and RCG understory (schematic diagram of appropriate site for experiment A)

Page 11 of 23