Photo by: K.M. Kettenring Final report to the Utah Division of Water Quality Mapping wetland vegetation in the Great Salt Lake Ecosystem Karin M. Kettenring, Chad Cranney, and Eric L.G. Hazelton Utah State University Cite as: Kettenring, K.M., C. Cranney, and E.L.G. Hazelton. 2016. Mapping wetland vegetation in the Great Salt Lake Ecosystem. Final report to the Utah Division of Water Quality. 24 pp.
Introduction Understanding the condition of Great Salt Lake (GSL) wetlands requires knowledge of the cover of wetland vegetation of interest, and how vegetation changes over time, including changes in invasive species. Furthermore, wetland managers need access to basic information about the condition of their wetlands, such as wetland vegetation coverage, particularly in response to management or restoration efforts. Assessing changes in wetland vegetation at large spatial scales, including invasive species, in response to management and restoration is methodologically challenging. The use of high resolution remote sensing imagery is one methodological option that has not been widely applied in GSL wetlands. We sought to improve the assessment and management of GSL wetlands through the application of this rapidly evolving remote sensing technology. This project is part of a larger multi-year effort conducted by a team of Utah State University researchers and partners to map the extent of invasive Phragmites australis around GSL, to assess changes in cover of Phragmites australis, and to evaluate the success of Phragmites control efforts being conducted across the GSL. The benefits of this research are increased understanding of the current extent of Phragmites, how it spreads over time, and the best ways to control it. The information generated from this project will be used to develop a comprehensive management strategy that includes water quality. In addition, maps generated by this project will be used for beneficial use assessments of GSL wetlands. Documenting changes in Phragmites australis cover are particularly important to assess in GSL wetlands because of its negative effects on wetland condition and potentially positive effects on wetland water quality. Phragmites australis (common reed) is an invasive grass that has rapidly invaded wetlands across North America (Marks et al. 1994) and is widespread and dominant in wetlands, ditches, and roadsides in northern Utah (Kulmatiski et al. 2011; Kettenring et al. 2012a). This plant is undesirable because it crowds out native vegetation and degrades habitat for wildlife including waterfowl and other migratory birds by creating large monotypic stands (Marks et al. 1994). GSL wetlands are the most important wetland habitat for migratory birds in the region and are continentally significant (Evans and Martinson 2008). Phragmites has invaded more than 23,000 acres of wetlands and open water on the GSL, reducing the availability of quality habitat (Kettenring et al. 2012c). Phragmites is also undesirable because it impedes recreation, affects property values, is a fire hazard and therefore threatens air quality, and consumes valuable water resources. Although many natural resource managers in Utah and elsewhere are trying to manage Phragmites, many questions remain as to how best to reduce its cover and foster the recovery of native plants. A variety of strategies have been widely employed including summer or fall herbicide application, mowing, burning, and flooding (Marks et al. 1994; Hazelton et al. 2014). But, as is often the case with natural resource management, due to limited time and money, there has been little monitoring of success nor any systematic evaluation of management strategies across the varied environmental conditions where Phragmites is found, particularly in Utah. Given the interest in effective management strategies for Phragmites in Utah and across North America, there is a need to evaluate different potential strategies and then monitor the success of those strategies. By developing effective methods for controlling Phragmites, managers can be more efficient in their management efforts (thereby saving time and money) and water quality impacts will be reduced by indicating when and how to apply herbicide for the greatest reduction in Phragmites cover. Here we report on changes in Phragmites cover (documented with the use of remote sensing imagery) in response to various Phragmites control treatments. Page 2 of 24
Project objective To use remote sensing imagery to look at wetland vegetation cover changes (particularly Phragmites) over time and in response to Phragmites control efforts. Methods Phragmites control experiment This experiment was designed to evaluate strategies that may be more effective and logistically feasible for dealing with large, well-established stands of Phragmites. We have four sites with extensive stands of Phragmites where we are conducting the management treatments: Ogden Bay Waterfowl Management Area (WMA), Farmington Bay WMA, sovereign lands west of Ogden Bay WMA, and sovereign lands northwest of Farmington Bay WMA. At each site, we applied 5 treatments to each 3 acre Phragmites stand (15 acres total per site). The Phragmites treatments were chosen based on our initial survey of GSL wetland managers (Kettenring et al. 2012b); extensive conversations with Randy Berger and other state, federal, and private managers; and our reading of the Phragmites management literature (reviewed in Hazelton et al. 2014). We chose treatments that were logistically feasible for managers to apply, and chose a balance of treatments that represented commonly applied strategies as well as less common ones that hold great promise for GSL wetlands (Hazelton et al. 2014). The five treatments we applied were: (1) summer glyphosate spray followed by winter mow, (2) summer imazapyr spray followed by winter mow, (3) fall glyphosate spray followed by winter mow, (4) fall imazapyr spray followed by winter mow, and (5) untreated area. These treatments were first applied in 2012, and were repeated in 2013 and 2014. In summer 2015 and 2016, we are monitoring treatment effectiveness now that the herbicide treatments are complete. It is critical to monitor this experiment for multiple years after treatments have ceased to conclusively determine if we are actually restoring wetlands (rather than just temporarily knocking back Phragmites). Chad Cranney has been the lead graduate student on this study but he is graduating with his MS this spring. Christine Rohal will take over from Chad to complete the final year of data collection (2016). Treatment effectiveness (changes in vegetation cover) is being assessed with on-theground vegetation surveys (data not included in this report) but given the logistical challenges of working with 3 acre treatment plots, remote sensing of entire plots provides a more detailed assessment of whole-plot changes. Aerial imagery acquisition Aerial imagery were acquired of the research sites 2012-2015; multiple vendors for the imagery were used year-to-year due to budget constraints. Dates of flights, sensors used, and flight platforms varied across the study (Table 1). Highest quality imagery were collected during the two most critical years: 2013, a full year after the first Phragmites treatments, and 2015, a full year after the final Phragmites treatments. Since multiple sources were used for imagery, across-year analyses are not possible, however all data are suitable to compare treatment effects within a given year. In 2012, only 3 of the four sites were flown due to federal restrictions on the use of unmanned aerial vehicles (UAVs) in controlled airspace. We conducted pilot studies in attempts to automate the identification of Phragmites and other plant communities. In all years imagery, ERDAS Imagine could not effectively differentiate between Phragmites and other cover types. The variation in spectral signatures and Page 3 of 24
textures within the Phragmites vegetation class were greater than those between Phragmites and other cover types. Other researchers have witnessed this limitation as well and determined that Phragmites is best identified using a combination of multispectral imagery and active remote sensing methods such as LiDAR (Gilmore et al. 2008) or Side Aperture RADAR (Bourgeau- Chavez et al. 2013). In the absence of additional data sources, we determined that manually digitizing the Phragmites cover would be the most efficient analysis. Images were analyzed visually by a single expert observer. Due to the high resolution of the imagery, Phragmites could be identified in the imagery at the sub-meter scale. A combination of the RGB and NIR bands allowed for differentiation between Phragmites and native vegetation. This Phragmites classification was then confirmed using texture (patterning, shading, stature). Subsequent comparison of these methods to known control points (field data collected by graduate student Chad Cranney) within each image confirmed that the method was accurate at determining Phragmites near-monocultures as small as 1m 2. Digitized Phragmites area within each treatment plot was then used to determine the percent cover of Phragmites for each 3 acre treatment plot. Results and discussion The remote sensing imagery acquired each year for this project provided fine-scale details in vegetation changes, particularly Phragmites, across the four sites in each of the five treatment plots. Changes in other vegetation types beyond Phragmites were minimal (based on field surveys conducted by Chad Cranney), therefore we focus here on changes in Phragmites cover. Phragmites cover was relatively stable in all the untreated control plots across sites and years (Table 2; Figures 1-16). The remote sensing imagery showed that Phragmites response to the treatments was dramatic, particularly after the 2012 treatments when the 2013 imagery shows large reductions in Phragmites cover (Figures 1-15). However, in 2015, Phragmites cover rebounded in the summer herbicide treatment plots (Figure 16) and these changes were captured clearly in the imagery (Figures 1-15). These results have important management implications. For the short term, for effective Phragmites control, summer or fall herbicide applications are equally as effective. However, when looking beyond this short term control, the summer treatments were much less effective (see 2015 imagery and Figure 16). Therefore, we recommend to managers that they consider herbicide application in the fall, except when trying to specifically prevent Phragmites seed production. In a companion study with field surveys, we found that summer treatments were substantially more effective at reducing Phragmites inflorescence production (Christine Rohal, MS thesis). Therefore, a summer spray in year 1 may be preferable to deal with seed production (the main form of Phragmites spread; Kettenring and Mock 2012) followed by fall treatments in years 2 and 3 of a three year herbicide application sequence. There were a number of reasons why we focused on using high resolution remote sensing approach for this study. First, because the experimental plots were so large, we expected that remote sensing imagery would be more effective at describing whole-patch changes in vegetation. We still believe that this was the appropriate approach to achieve that goal. We also expected that collecting high resolution imagery (~5cm resolution) would allow us to monitor fine-scale changes in other vegetation types as Phragmites was removed. However, because native plant recovery has been so sparse in this study (confirmed with Chad Cranney s on-the- Page 4 of 24
ground field surveys), we were not able to fully address this application use for that purpose. (In other words, there was such a limited cover of native vegetation that we did not have sufficient native vegetation cover to analyze. We also expected that this fine-scale resolution imagery would allow us to set up automated vegetation classification of vegetation types using ERDAS Imagine software, following the methods previously developed by MS student Lexine Long. However, the extremely fine resolution imagery we acquired for this project had much higher noise in the spectral signature within the Phragmites vegetation class (compared with the 1m resolution data that Lexine worked with), that we could not reasonably automate the Phragmites cover classification (vs. open water or other plant species). Therefore, we ended up having to manually digitize Phragmites cover changes over time. This approach was significantly more time-intensive than if we had been able to automate the vegetation classification. Conclusions Here we document using high resolution remote sensing imagery large changes in Phragmites cover in response to various Phragmites control treatments. Our study yields important management recommendations, specifically that a summer herbicide application is not effective for long term Phragmites control. We also document that these high resolution imagery are of such high quality, that they are not effective for distinguishing Phragmites from other vegetation types or open water, due to the high variation in spectral signature found within the Phragmites vegetation class. Future projects should consider pairing such high resolution imagery with other data sources (e.g., LiDAR) or continuing to hand digitize Phragmites stands. Literature cited Bourgeau-Chavez LL, Kowalski KP, Mazur MLC, Scarbrough KA, Powell RB, Brooks CN, Huberty B, Jenkins LK, Banda EC, Galbraith DM (2013) Mapping invasive Phragmites australis in the coastal Great Lakes with ALOS PALSAR satellite imagery for decision support. Journal of Great Lakes Research 39:65-77 Evans K, Martinson W (2008) Utah's featured birds and viewing sites: a conservation platform for Important Bird Areas and Bird Habitat Conservation Areas. Sun Lith, Salt Lake City, Utah Gilmore MS, Wilson EH, Barrett N, Civco DL, Prisloe S, Hurd JD, Chadwick C (2008) Integrating multi-temporal spectral and structural information to map wetland vegetation in a lower Connecticut River tidal marsh. Remote Sensing of Environment 112:4048-4060 Hazelton ELG, Mozdzer TJ, Burdick D, Kettenring KM, Whigham DF (2014) Phragmites australis management in the Unites States: 40 years of methods and outcomes. AoB Plants 6:plu001 Kettenring KM, de Blois S, Hauber DP (2012a) Moving from a regional to a continental perspective of Phragmites australis invasion in North America. AoB Plants:1-18 Kettenring KM, Garvie K, Hazelton ELG, Hough-Snee N, Ma Z (2012b) Phragmites invasion and control in the Great Salt Lake watershed: 2012 land manager survey. Final report to the Utah Department of Natural Resources, Division of Forestry, Fire & State Lands. pp. 26 Kettenring KM, Long AL, Neale CM (2012c) Determining the current extent of Phragmites australis in Great Salt Lake wetlands using multi-spectral remote sensing techniques. Page 5 of 24
Kettenring KM, Mock KE (2012) Genetic diversity, reproductive mode, and dispersal differ between the cryptic invader, Phragmites australis, and its native conspecific. Biological Invasions 14:2489-2504 Kulmatiski A, Beard KH, Meyerson LA, Gibson JR, Mock KE (2011) Nonnative Phragmites australis invasion into Utah wetlands. Western North American Naturalist 70:541-552 Marks M, Lapin B, Randall J (1994) Phragmites australis (Phragmites communis): threats, management, and monitoring. Natural Areas Journal 14:285-294 Page 6 of 24
Figures and Tables Table 1. The vendor and imagery details for each flight conducted 2012-2015. Flight year 2012 2013 2014 2015 Vendor used for flight USU AggieAir Aero-graphics, Salt Lake City Sensor Canon S-95 Microsoft UltraCam Eagle Platform Unmanned Aerial Vehicle (UAV) USU Remote Sensing Services Laboratory Imperx B4820 Aero-graphics, Salt Lake City Microsoft UltraCam Eagle Fixed wing Fixed wing Fixed wing Flight date June 25, 2012 October 5, 2013 October 3, 2014 August 31, 2015 Imagery type RGB + NIR 4-band 4-band 4-band Pixel resolution RGB 7cm NIR 6cm 5cm 6cm 5cm Page 7 of 24
Table 2. Changes in Phragmites cover 2012-2015 by treatment type and site. Treatments were FG = fall glyphosate, FI = fall imazapyr, SG = summer glyphosate, SI = summer imazapyr, and UC = untreated control. Sites were FB1 = Farmington Bay WMA Site #1, FB2 = Farmington Bay WMA Site #2, OB = Ogden Bay WMA, and HS = Howard Slough WMA. NA = not applicable for 2012 FB1 site when the UAV flight was not allowed near the Salt Lake City airport and therefore no imagery could be collected. 2012 Phragmites cover (%) 2013 Phragmites cover (%) 2014 Phragmites cover (%) 2015 Phragmites cover (%) Treatment Site FG FB1 NA 25% 4% 7% FG FB2 84% 7% 49% 13% FG HS 86% 30% 40% 45% FG OB 100% 15% 28% 9% FI FB1 NA 6% 10% 4% FI FB2 91% 1% 22% 3% FI HS 92% 10% 31% 25% FI OB 99% 5% 4% 3% SG FB1 NA 32% 31% 64% SG FB2 91% 6% 23% 65% SG HS 94% 11% 35% 98% SG OB 100% 65% 64% 100% SI FB1 NA 41% 61% 91% SI FB2 90% 9% 18% 61% SI HS 86% 41% 57% 98% SI OB 100% 21% 3% 14% UC FB1 NA 98% 100% 100% UC FB2 89% 92% 93% 100% UC HS 81% 100% 100% 100% UC OB 100% 100% 100% 100% Page 8 of 24
Figure 1. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 9 of 24
Figure 2. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 10 of 24
Figure 3. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 11 of 24
Figure 4. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 12 of 24
Figure 5. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 13 of 24
Figure 6. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 14 of 24
Figure 7. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 15 of 24
Figure 8. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 16 of 24
Figure 9. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 17 of 24
Figure 10. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 18 of 24
Figure 11. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 19 of 24
Figure 12. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 20 of 24
Figure 13. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 21 of 24
Figure 14. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 22 of 24
Figure 15. Phragmites cover (in blue) in five experimental treatment plots in a Great Salt Lake Page 23 of 24
1.0 2012 2013 2014 2015 0.8 Phragmites cover 0.6 0.4 0.2 0.0 Control Summer glyphosate Summer imazapyr Fall glyphosate Fall imazapyr Phragmites treatments Figure 16. Changes in Phragmites cover (mean ± 1 SE) in four large-scale restoration experiment sites on the Great Salt Lake subject to different Phragmites control treatments. Note that the imagery acquisition methods differed by year such that year-to-year variability may occur beyond the vegetation changes (but 2013 and 2015 imagery were acquired with the same methods and therefore are directly comparable). Page 24 of 24