Article Preview
TopIntroduction
Land use and land cover (LULC) change substantially alters the structure and functioning of wetland ecosystems which affects plant communities and microclimate thereby changing fluxes in water, nutrients, and energy across the land-atmosphere interface. By definition, wetland ecosystems are transitional regions between terrestrial and aquatic ecosystems with unique soil conditions, plants, and animals, and are essential components of terrestrial carbon and nutrient cycles (Mitsch & Gosselink, 2007; Cui et al., 2009; Li et al., 2012). Initially, conversion of wetland areas to agricultural lands was a main reason for wetland loss. Large area wetlands were converted to cropland to sustain food production for the ever-increasing population (Rijsberman & De Silva, 2003). In recent time population growth and urbanization have further undermined wetland areas and gradually converted them to built-up areas and agricultural lands (Bolca et al., 2007; Cui et al., 2010; Parihar et al., 2013). Drainage ditches for cropland have significantly lowered the water table depth of wetlands and reduced water-storage capacity and changed local hydrological cycling such as precipitation and runoff (Bartzen et al., 2010). Wetland conversion to built-up area results in an increase in impervious surfaces which may cause increases in evapotranspiration and runoff (Carlson & Arthur, 2000). These changes may also increase temperature due to rising greenhouse gas emissions from human-related sources and reduced vegetation covers (Chen et al., 2006; Balçik, 2014). Therefore, a thorough investigation of LULC dynamics on wetland ecosystems is the key to understandings of LULC change-induced hydrological, ecological and climatic processes.
Various methods have been developed in LULC dynamic studies using satellite data and can be generally referred to classification and post-classification comparison (Zhou et al., 2008). The classification method refers to techniques that use various classification algorithms to directly obtain land cover maps (Yuan et al., 2005). It mainly includes two techniques - Pixel based and objected based classifications (Dronova et al., 2011; Whiteside et al., 2011). Recently, Geographic Object Based Image Analysis (GeOBIA) has been widely used for image classification. Compared with the pixel-based methods, the GeOBIA not only considers spectral and textural information that are main factors influencing pixel-based classifiers, but also includes shape characteristics and context in adjacent pixels (Myint et al., 2011). A number of geographical/geometric feature attributes (e.g. shape, adjacency, and topological entities) are included in objects, which provide useful information that cannot be obtained from single pixels and can be easily processed by setting a series of classification rules (Whiteside et al., 2011). Therefore, the performance of GeOBIA is usually more efficient than the pixel-based classifier in processing data with relatively high spatial resolutions (Gao & Mas, 2008; Kindu et al., 2013). Besides the high efficiency, the object-based analysis can eliminate the salt and pepper effect that is caused by closely located pixels classified into different land cover types, which usually occurs in pixel-based classification. In addition, the GeOBIA has fewer errors in identifying distinct edges of different land cover types and performs better in temporal analysis of LULC change than the pixel-based methods (Dingle Robertson & King, 2011). For wetland classification, since high spectral and spatial heterogeneities exist due to differences in water depths and vegetation, the process is highly context-dependent (Wright & Gallant, 2007; Cui et al., 2010). The features of object-based analysis make GeOBIA a very useful method in investigating wetland changes over long periods (Dingle Robertson & King, 2011; Dronova et al., 2011; Moffett & Gorelick, 2013).