Mapping the maximum inundation extent of lowland intermittent riverine wetland depressions using LiDAR

Highlights

• Wetland depression volumes were modelled using LiDAR data for series of water depths.

• Fill levels corresponded with change points in derivatives of rates of change in volume.

• Maximum inundation extent of wetland depressions were delineated using the mapped fill levels.

• Boundaries were validated by temporal analysis of inundation using Landsat imagery.

• The boundary products inform management of water allocated to the environment.

Abstract

Accurate high-resolution maps of maximum wetland inundation extents are valuable inventorial resources, but mapping such boundaries can be difficult, time consuming and involve a level of subjectivity from the surveyor. A novel, objective and efficient method is presented for delineating the maximum inundation extents of lowland intermittent riverine wetland depressions using a high resolution LiDAR-derived digital terrain model (DTM). The method was specifically developed for wetland depressions associated with the lower reaches of the Murrumbidgee River in the southern Murray Darling Basin, Australia, but is suitable for any similar intermittent wetland depressions where the natural inundation extent is otherwise unclear. The method is based on the premise that immediately after rising water reaches a threshold level, much larger volumes of water are required to inundate not just the discrete wetland depression but also its surrounding local area; that threshold level can be considered the wetland fill level. Using a series of water levels and a corresponding series of the volume between each water level and a DTM of a wetland area, the maximum rate of increasing acceleration in volume with respect to water level, i.e. the wetland fill level, can be identified. The fill levels of eight wetland areas, ranging in size from 44 to 384 ha, were produced by this method and subsequently used to map their maximum inundation extents. Corresponding maps of the inundation frequency gradient were independently determined from a time series of Landsat derived inundation maps. The level of separation in inundation frequency of map cells inside the maximum inundation extents compared to those outside, determined by the Kolmogorov Smirnov statistic, validated the technique, while identifying that a minimum level of variation in the elevation of the wetland area may be necessary for it to be successfully applied. The technique has potential to be applied more widely as an objective and relatively inexpensive procedure to identify the potential maximum inundation extent of intermittent floodplain wetlands when a good quality high resolution DTM, acquired during a dry phase, is available.

Introduction

Floodplain wetlands are productive ecosystems with high biodiversity, providing important ecosystem services (Brander et al., 2006; Dudgeon et al., 2006; Maltby and Acreman, 2011; Ward et al., 2002). However, the natural flow regimes that deliver water to wetland areas are increasingly being altered globally by water resource development and climate change (Mittal et al., 2016; Poff et al., 1997), resulting in the degradation of flow-dependent freshwater ecosystems (Kingsford et al., 2015; Kuiper et al., 2014). To mitigate the impacts of altered flow regimes on floodplain wetlands reliant on regulated rivers, environmental water is increasingly being managed to restore degraded ecosystems (Arthington, 2012; Gordon et al., 2010; Poff and Zimmerman, 2010). For best decision making around the application of limited environmental water, detailed knowledge of dependent wetland areas is required (Finlayson et al., 1999). In the case of intermittently inundated wetland depressions, morphometry data are a requirement for the derivation of metrics that describe the response of water balance, hydroperiod and zonation to particular flow events. Inundation metrics can then be linked to ecological function and processes (Brooks and Hayashi, 2002; Johnston et al., 2001; Pollock et al., 1998) in order to predict ecological outcomes. For example, spatiotemporal depth patterns influence the distribution of aquatic plants, driving community structure (Pollock et al., 1998), and carbon and nutrient dynamics (Johnston et al., 2001). Inundation metrics can be used to estimate required water volumes, which is useful when aiming to restore inundation regimes through managed environmental water. However, effective application of inundation metrics is predicated on knowledge of the natural maximum inundation extent, which may not be obvious for long-term altered wetland systems. A key component for intermittent wetland morphometric inventories are therefore spatial data describing wetlands' maximum inundation extents.

On-ground survey methods used to define wetland areas generally involve the analysis of vegetation, soil and hydrology, and, for areas regularly inundated, take advantage of identifiable differences in these properties to surrounding areas (e.g. Department of Environment and Resource Management, 2010; Environmental Laboratory, 1987; Haag et al., 2005). But the application of on-ground methods can be time consuming and difficult to implement over large areas (Halabisky et al., 2016; Scott and Jones, 1995). For this reason, at most scales of wetland inventory development, the principle source of data is remote sensing (Rebelo et al., 2009). A summary of relevant research involving the development of methodologies for open water wetland mapping utilising remote sensing is presented in Table 1.

Optical remote sensing is regularly employed to map inundation extents in ephemeral systems (e.g. Chen et al., 2014; Thomas et al., 2015), to monitor wetland vegetation (Adam et al., 2010), and characterise temporal variability in wetland extent and quality (e.g. Yuan et al., 2005). Many of the papers in Table 1 describe methodologies suited to systems at large scales (over 1000 km²) and/or dynamic monitoring of inundation. For inventories of wetland systems over relatively large areas, methods that produce a low level of spatial resolution in the resulting map or product is often acceptable, but such methods do not translate to smaller scale evaluations. In addition, several of the studies were principally concerned with classifying wetland areas into subtypes, using methods that involve multitemporal analysis of remotely sensed imagery (Betbeder et al., 2014; Evans et al., 2014; Frohn et al., 2009; Reif et al., 2009; Yuan et al., 2005), which are not suitable for the precise mapping of the maximum extent of inundation in wetlands.

Studies that aim to map relatively small ephemeral wetlands at a level of morphometric precision required to enable accurate estimates of water volumes and depths utilise LiDAR (e.g. Lane and D'Amico, 2010), which is often used in conjunction with optical remote sensing (e.g. Huang et al., 2014; Maxa and Bolstad, 2009; Rapinel et al., 2015a). Rapinel et al. (2015a) used LiDAR to specifically estimate isolated wetland water storage capacities, by producing three-dimensional models of wetlands from digital terrain models (DTM) derived from LiDAR acquired during a wetland dry phase. Polygon boundaries were then used by Rapinel et al. (2015a) to delineate the study wetlands based on prior classification using Landsat-7 data (Frohn et al., 2009; Reif et al., 2009). The volume bound by a plane representing a water level at the same elevation as the points of the DTM that coincided with the polygon boundaries could then be calculated within a GIS (Rapinel et al., 2015a). While providing a useful methodology to model discrete wetland waterbody volumes, the Rapinel et al. (2015a) method does not extend to mapping inundation extent; the method required pre-mapped polygons to define waterbody boundaries.

Our study describes a novel method for delineating the maximum inundation extent of specific discrete wetlands using LiDAR-derived DTMs. Boundaries were derived from only a high-resolution DTM of a user-selected wider local area containing the target wetland water body; the method does not require optical remote sensing, except for validation purposes. Our motivation was the need for accurate high-resolution boundaries of maximum inundation extent of specific wetlands in order to estimate morphometrics for the analysis of ecological responses to inundation. The method needed to be rapid and applicable to a range of geomorphologically variable sites. The objectives of this study were to: (i) produce a rapid desktop based method to identify fill level elevations of discrete ephemeral water bodies; and (ii) map the maximum inundation extent of the water bodies based on the identified fill levels and validate them using historical inundation frequencies derived from optical remote sensing data.