Introduction

Wetlands provide important habitat for plants and animals such as nesting sites for waterfowl and spawning grounds for fish. Wetland vegetation protects shorelines from erosion and traps sediment before it enters water bodies. The quality of water passing through wetlands can be improved because wetlands filter contaminants and nutrients. Wetlands also renew groundwater supplies, and help to control flooding which in turn reduces damage caused by flooding. Finally, wetlands are an important economic resource as they provide natural products and provide recreational opportunities such as hunting, fishing, and bird watching (Jaworski, 1978, as cited by Herdendorf, 1992). Unfortunately, wetlands are being lost at a fast rate due to drainage, conversion,

pollution, and over-exploitation of their resources. In southern Ontario prior to European settlement, it is estimated that there were 2.4 million ha of wetlands. Roughly 933,000 ha or 61% of this wetland area had been lost to development by 1982 (Snell, 1987). As of 2011, southern Ontario was home to 13.4 million people and the population is expected to increase to 17.7 million by 2036 (Ontario Ministry of Finance, 2012), which will increase the level of pressure on the existing wetland areas. Another problem facing wetlands is the detrimental effects of the introduction of invasive species. Invasive plant species can decrease plant diversity, threaten rare and endangered native species, and decrease habitat quality for birds and animals (Laba et al., 2008). Of particular concern in North American wetlands is the introduction of Phragmites australis (Cavenilles) Trinius ex. Steudel subsp. australis (common reed) (Saltonstall, 2002, Mal and Narine, 2004), hereafter referred to as Phragmites. This invasive subspecies has been displacing native

Phragmites (Saltonstall, 2002) as well as other wetland vegetation species (Lavoie, 2008). The effects of the invasion by Phragmites on fish, birds, and mammals is mixed and not well studied in freshwater wetlands (Lavoie, 2008).

An important tool for the management of invasive wetland vegetation is accurate and up- to-date vegetation maps (Parker Williams and Hunt, 2002). If the location and abundance of the species of interest is known, then an effective plan for controlling the spread can be implemented. Traditionally on-ground field surveys have provided information about wetland vegetation species distribution and abundance. However, this method is time consuming and labour intensive (Laba et al., 2008), especially when conducted over large areas. Also, field surveys may not be feasible due to restrictions on accessibility posed by the wetland environment. Remote sensing provides an alternative to on-ground field surveys. Imagery from aerial or satellite platforms can be captured over large areas, and can make repeat observations of the same area (Ozesmi and Bauer, 2002). This makes remote sensing imagery an ideal source for mapping wetlands for widespread

management of invasive plant species such as Phragmites (Adam et al., 2010). Mapping of Phragmites using satellite remote sensing in North American freshwater coastal wetlands and estuarine marshes has been limited. Laba et al. (2008) used a single- date QuickBird-2 image from August and the maximum likelihood classifier to map

Phragmites in an estuarine marsh. User’s and producer’s accuracy of Phragmites was 76% and 100% respectively. Ghioca-Robrecht et al. (2008) conducted an unsupervised classification using multi-date QuickBird-2 multispectral imagery (April and September) of Erie Marsh, one of the largest marshes in Lake Erie. Phragmites was mapped with a user’s and producer’s accuracy of 76% and 53% respectively. Separation of Phragmites

and other wetland plant species from multi-date imagery alone was not good. Individual plant species can be mapped from multi-date high resolution multispectral imagery with high accuracy if additional data are included. Gilmore et al. (2008) used LiDAR along with multi-temporal QuickBird-2 high spatial resolution imagery to map three wetland vegetation species Phragmites, Typha spp., and Spartina patens in a brackish tidal marsh. Based on a fuzzy accuracy assessment, Phragmites was mapped with 97% accuracy. However, LiDAR data is expensive for use in widespread wetland vegetation mapping on an annual basis. Also, multiple images increase the cost and may not be available due to weather conditions or satellite tasking restrictions.

Therefore, there is some room for improvement for mapping Phragmites using single- date satellite imagery. Individual land managers are limited in resources and expertise, therefore obtaining data gathered from multiple satellites, from multiple seasons, or non- optical data such as LiDAR may not be possible. An accurate method of mapping

Phragmites from single-date imagery is a more attractive mapping solution.

Worldview-2 (WV-2) is a recently launched, high resolution satellite with eight 2 metre spatial resolution multispectral bands and a 0.5 metre Panchromatic band. WV-2 contains a Coastal Blue, Yellow, Red-Edge, and a second Near Infrared band (NIR2) in addition to the Blue, Green, Red, and Near Infrared (NIR1) bands of other high resolution

multispectral satellites such as IKONOS-2 and QuickBird-2. Despite the increased spatial and spectral resolution, the WV-2 satellite has similar imaging capabilities as QuickBird- 2 and IKONOS-2 including image acquisition size, large area and long strip collection, as well as stereo imaging abilities (refer to eoPortal Directory, 2012). With the launch of Worldview-3 in 2014, the Worldview constellation will be able to image a location on the earth’s surface every two days (eoPortal Directory 2012).

WV-2 imagery has not been previously used for mapping individual wetland species. The goal of this research was to evaluate the use of WV-2 for mapping the invasive emergent species Phragmites in a Great Lakes coastal wetland. It was hypothesized that the eight multispectral bands possessed by Worldview-2 would increase classification results compared with other multispectral sensors (such as IKONOS-2, QuickBird-2) containing only four bands. Object-based methods were developed for both four and eight band imagery and the results compared. This comparison showed the advantage of using the full eight bands provided by the Worldview-2 sensor.