Introduction

Groundwater is an important resource in Australia representing up to 17 % of currently accessible water resources (National Water Commission 2012). The sustainable use and management of this finite resource remains a challenge for governments, compounded by growing demand for water and the combined pressures of development (e.g. coal seam gas extraction), a variable climate (Office of Climate Change 2010) and increasing surface water scarcity (Foster and Chilton 2003; Zektser and Everett 2004; Water Services Association of Australia 2010; Harrington and Cook 2014). The 2004 Intergovernmental Agreement on a National Water Initiative commits all Australian jurisdictions to “secure ecological outcomes by describing the environmental and other public benefit outcomes for water systems and defining the appropriate water management arrangements to achieve those outcomes” (Council of Australian Governments 2004). Currently all jurisdictions in Australia require consideration of the ecological values of groundwater in water planning. Further, the 2012 National Partnership Agreement on Coal Seam Gas and Large Coal Mining Development commits all signatories to seek advice from an independent committee for those “coal seam gas or large coal mining development proposals that are likely to have a significant impact on water resources” (Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development 2015). In the advice to be provided by the independent committee, specific reference is given to the consideration of impacts on “water resources, water related assets and other users, including ecological impacts that are directly related to water resources”(Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development 2015). This emphasis on ecological values in water planning and environmental impact assessment has generated a need for a consistent, transparent process to identify the ecological values associated with both surface and groundwater.

Ecosystems depend, to varying degrees, on a variety of water sources including groundwater, surface water and/or soil water. We adopt an ecohydrological definition of groundwater, which includes water that is present in the pores and cracks of the saturated and capillary zones of soils, regolith and rocks and water that is present in caves. This definition includes perched groundwater (i.e. areas where pores and cracks may be locally saturated due to a lower permeability geological layer) in the unsaturated zone.

Groundwater-dependent ecosystems (GDEs) are a sub-set of all ecosystems, specifically those that “require access to groundwater on a permanent or intermittent basis to meet all or some of their water requirements so as to maintain their communities of plants and animals, ecological processes and ecosystem services” (Richardson et al. 2011a). GDEs may include a broad range of ecosystems including aquifers, caves, lacustrine wetlands (e.g. lakes), palustrine wetlands (e.g. swamps), riverine wetlands (e.g. rivers) and terrestrial vegetation. A key point is that the dependence of GDEs on groundwater does not necessarily engender GDEs with any other particular characteristic that separates them from all other aquifers, caves, lakes, swamps or terrestrial vegetation (Eamus et al. 2006b; Nevill et al. 2010). Table 1 outlines a GDE typology based on Eamus et al. (2006b).

Table 1 Groundwater-dependent ecosystem typology (adapted from Environment Protection Agency 2005; Eamus et al. 2006b; Neldner et al. 2012)
Full size table

Approaches to GDE Mapping

As groundwater is not readily identifiable at the ground surface, the delineation of GDEs is complex. A wide array of methods is currently available to map GDEs at a range of different scales and for a variety of purposes. One key ecological principle that underpins many current GDE mapping methods is that where groundwater is accessible to an ecosystem that ecosystem will probably be using groundwater and is therefore groundwater dependent to some degree (Hatton and Evans 1998). Building on this principle, there are three main approaches typically employed to map GDEs at a catchment scale: field assessments, remote sensing approaches where GDEs are identified by characteristic signatures, and geographic information system (GIS) approaches where spatial data such as vegetation, geology and hydrology data are combined to map GDEs.

The use of numerical modelling tools and approaches have not explicitly been addressed in this paper due to the focus on catchment scale mapping of GDEs across a diverse range of catchments including data sparse regions. At this scale, GDE mapping potentially includes numerous connected, semi-connected and discrete aquifers at a range of depths, in different hydrogeological basins, and with varying scales of groundwater flow systems. This includes aquifers with groundwater flow systems that extend across multiple catchments. Some of these aquifers may not store quantities of groundwater suitable for extraction; however they remain ecologically relevant in terms of their ability to support ecosystems. A numerical modelling approach would be time and resource intensive with multiple models required to adequately map GDEs. The mapping approaches detailed below produce catchment scale GDE mapping which can be used to inform the development of numerical groundwater flow models to assess the impact of changes in the quantity and quality of a specific groundwater resource from management decisions.

Specific field techniques are useful to identify GDEs by assessing groundwater use of individual representative ecosystems. Such techniques include isotopic analysis of water in soil, groundwater and plant xylem (Chapman et al. 2003; Eamus 2009; Cartwright et al. 2010), water balance calculations (Brodie et al. 2007a, b; Eamus 2009) and assessment of leaf area index (Hatton and Evans 1998; Colvin et al. 2007; Carter and White 2009; Eamus 2009), vegetation rooting depth (Eamus 2009); and depth to groundwater (Eamus 2009; Hoogland et al. 2010). Several toolboxes (Colvin et al. 2003; Richardson et al. 2011a, b) have been produced which compile these field techniques and other methods. Field techniques are often resource intensive (Dresel et al. 2010; Gow et al. 2010) and consequently they are useful to develop or test conceptual understanding of GDEs, or to ground-truth GDE maps, but are otherwise ill-suited to catchment scale GDE mapping applications.

Catchment scale GDE mapping approaches are commonly based on less resource intensive remote sensing approaches which ideally require a basic conceptual understanding of the interaction between groundwater and ecosystems within a landscape, and the effect of this interaction on ecosystem spectral signatures in remote sensing imagery (Barron et al. 2012). Remote sensing approaches are often based on searches for patterns that suggest relatively high evapotranspiration rates during dry periods (O’Grady et al. 2011). For example, Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat satellite imagery are commonly used to map GDEs based on prolonged vegetation ‘greenness’ during drier months or periods of drought, or through the identification of thermal anomalies associated with groundwater discharge and evaporation (Barron et al. 2012). MODIS and Landsat remote sensing imagery are at coarse (250–1000 m) and moderate (25 m) spatial resolutions respectively. Common indicators used for these types of remote sensing approaches include the Normalised Difference Vegetation Index (NDVI) (i.e. vegetation greenness), MODIS Enhanced Vegetation Index and the Normalised Difference Wetness Index (NDWI) (Gu et al. 2007; Tweed et al. 2007; Guerschman et al. 2009; Gow et al. 2010; Contreras et al. 2011; Barron et al. 2012).

There are two major technical limitations in using such remote sensing approaches for GDE mapping. First, there is a significant spatial mismatch between the resolution of remote sensing data and the scale of the ecological feature being mapped (Aplin 2004). The relatively coarse spatial resolution of widely available remote sensing data (e.g. MODIS and Landsat) reduces the effectiveness of remote sensing approaches in mapping smaller GDEs such as palustrine wetlands associated with permanent, point source springs. However, as finer scale remote sensing data becomes more widely available the potential for GDE mapping across catchments will improve and the impact of this limitation should diminish. Second, commonly used remote sensing indices such as NDVI (greenness) or NDWI (wetness) can vary for reasons unrelated to groundwater use. For example, across northern Australian landscapes fire drives significant variation in vegetation structure, with dense green rainforests associated with fire protected landscape positions within vast expanses of fire prone savanna (Bowman 2000). Other technical limitations with remote sensing approaches can arise because of variability within pixels, lags between changes in water availability and vegetation condition (Gow et al. 2010), variability in groundwater dependency within a GDE (Eamus and Froend 2006) and variability in groundwater dependence between similar GDEs in different sites and/or accessing groundwater at different depths (Zencich et al. 2002). Remote sensing approaches are thus often better suited to applications in areas with distinct seasonal wet and dry periods (Barron et al. 2012). For example, the use of NDVI has been shown to be an appropriate surrogate for wetland vegetation cover in arid regions of Queensland (White and Lewis 2011). Application of remote sensing approaches across extensive areas at a scale compatible with groundwater management and planning activities would likely require cost prohibitive, very high resolution remote sensing.

As well as technical limits to their reliability, remote sensing approaches can be criticised because of conceptual issues. Remote sensing approaches look for GDEs as a distinct ecosystem type, ecosystems that are greener or wetter than surrounding areas. However, as discussed above, groundwater dependence is just one characteristic of an ecosystem. GDEs range from lakes to woodlands, so the use of remote sensing approaches to develop a comprehensive map of GDEs depends on recognising a suite of ‘GDE signatures’ that will vary substantially across the full range of GDEs in any given landscape. A threshold value for a single index cannot capture the complexity of groundwater dependence in ecosystems in most real landscapes.

A final significant problem with GDE maps derived from the analysis of remote sensing imagery is that they typically identify potential GDEs with little or no information about how each GDE is connected to groundwater or to broader hydrological processes in the landscape. Understanding how a GDE is linked to groundwater and the nature of this connection is very useful for landscape planning and assessing potential impacts from proposed development.

An alternative catchment scale approach for mapping GDEs is the prediction of the presence or absence of GDEs based on associations with other landscape features through a GIS analysis (Gou et al. 2014). GDE mapping methods using a GIS approach tend to focus on the spatial analysis of multiple datasets (some possibly derived from interpretations of remote sensing data). For example, data on ecosystems have been overlaid with climatic (Howard and Merrifield 2010; Ozdemir 2011), geological and geomorphological (Colvin et al. 2003; Brodie et al. 2007b; Walsh 2008; Howard and Merrifield 2010; Ozdemir 2011), hydrological (Smith et al. 2006; Brodie et al. 2007b; Howard and Merrifield 2010) and topographical data (Colvin et al. 2003; Ozdemir 2011) to map GDEs. This approach allows a certain level of flexibility in the inclusion and omission of different factors based on data availability and the scope of the target study.

Gou et al. (2014) noted that assumptions made in GIS approaches potentially result in products that overestimate ecosystem groundwater dependence in comparison to remote sensing approaches. However, as discussed above, it is likely that remote sensing approaches fail to identify some types of GDEs due to biases inherent in the use of vegetation indicators such as NDVI or NDWI. GIS approaches that integrate expert knowledge with spatial data can offer more conceptually comprehensive assessment of GDEs than assessments based solely on indices derived from remote sensing.

Of-course GDE mapping methods can combine remote sensing and GIS approaches. The benefits of a mixed-assessment approach are widely recognised and have been trialled in various studies at a range of scales. For example, Münch and Conrad (2007) used NDVI to identify likely GDEs in the Western Cape, South Africa. These data were then integrated with other geospatial data on groundwater level and soil moisture availability to further refine and improve the accuracy of final products. Similarly, Barron et al. (2012) integrated NDVI with a set of terrain characteristics to develop a groundwater recharge and discharge map for western Victoria (Australia). This concept was taken further by Gou et al. (2014) who generated products at various scales, suited to different management needs. A regional scale (i.e. multi-catchment) analysis considered climatic, topographic, ecological and hydrogeological factors to identify those aquifers in Texas (USA) most likely to support GDEs. More detailed GDE mapping was then undertaken at the aquifer scale with NDVI derived from MODIS and Landsat ETM+ imagery indicating vegetation more likely to be using groundwater.

Australia’s National Atlas of GDEs

In 2012, the Australian Government funded the development of the National Atlas of Groundwater-Dependent Ecosystems (National GDE Atlas). Intended primarily as a management tool, the National GDE Atlas was developed by integrating analysis of MODIS and Landsat imagery identifying areas which may be accessing groundwater with supporting spatial data including ecosystem mapping to map likely GDEs (Dowsley et al. 2012). The National GDE Atlas provides information on a sub-set of GDEs, excluding information on subterranean aquatic ecosystems (e.g. aquifer and cave ecosystems), estuarine GDEs and near-shore marine GDEs.

Given the expansive area covered in the National GDE Atlas, it used broader scale spatial data in its assessment because finer scale data were not consistently available across all Australian states and territories. The use of these broader scale data compromised spatial accuracy and detail in the interest of consistency, which is inherently problematic when mapping GDEs because features such as shallow alluvial deposits that are associated with GDEs are not represented in broader scale data. Inadequate data for GIS analysis of GDEs in the National GDE Atlas meant that the final product was largely driven by the analysis of remote sensing imagery. As a result the National GDE Atlas, like many GDE maps, provides limited information about why a particular location is mapped as a GDE or how those GDEs are likely to be connected to groundwater. The National GDE Atlas is best considered as a guide to the location of potential GDEs rather than a decision tool for groundwater managers in specific catchments and therefore is insufficient to address the requirements of water planners and managers who often manage groundwater at catchment and regional scales.

In response to identified limitations in existing GDE mapping methods including the approach used to develop the National GDE Atlas, the National Water Commission and Queensland Government co-funded the Queensland Groundwater-Dependent Ecosystem Mapping Project. This pilot project developed a mapping method that facilitates multi-disciplinary synthesis of catchment-scale knowledge and capitalises on the best available spatial data in its approach to mapping GDEs. This method addresses the gap identified by Mackay (2006) between developed national scale management tools, policies and regulation and the typically catchment and regional scale management issues associated with groundwater exploitation.

Delineating GDEs using this mapping method requires the development of a shared, multi-disciplinary and comprehensive conceptual understanding of the relationships between groundwater and ecosystems in a landscape. The method has been applied to varied catchments in Queensland, Australia. It has proven to provide readily usable products for a range of applications at the sub-catchment and catchment scale including water planning and environmental impact assessment.

A New Method for Conceptually Comprehensive GDE Mapping

The Queensland Groundwater-Dependent Ecosystem Mapping Method (the method) was developed to produce a map of GDEs, at the finest practical scale as determined by source datasets (often 1:100,000 in Queensland but in some coastal areas the scale is 1:50,000 Queensland Department of Science, Information Technology and Innovation 2015).

This method uses a rigorous, iterative, heuristic approach that synthesises multidisciplinary local, expert knowledge with relevant spatial data in a GIS, to identify where and, equally importantly, why ecosystems are or are likely to be groundwater dependent. The method is based primarily on a GIS approach, but is sufficiently flexible so as to incorporate any relevant remote sensing derived datasets. This approach was developed because it was recognised that remote sensing approaches are unlikely to produce comprehensive assessments of GDEs across a range of catchments as climatically and biologically diverse as Queensland. However, remotely sensed data, (like other spatial datasets), can be incorporated into GIS approaches where local expert knowledge identifies that the information is appropriate to delineate GDEs.

The method is divided into 5 stages containing a total 15 steps (Fig. 1): development of mapping rule-sets and pictorial conceptual models (steps 1–3); development of draft GDE products (steps 4–6); refinement (steps 7–11); product testing (step 12); and finalisation and product release (step 13–15). This method has been used to delineate terrestrial and surface expression GDEs over approximately 12 % (or 209,038 km2) of Queensland and mapping is underway to extend this to a further 42 % (or 736,377 km2) of Queensland (Fig. 2). As an example, this paper will outline the results from the application of the method to map GDEs in the Mackay–Whitsunday region.

Fig. 1
figure 1

Overview of the Queensland groundwater-dependent ecosystem mapping method

Full size image
Fig. 2
figure 2

Current extent of GDE mapping using the Queensland GDE mapping method in Queensland, Australia

Full size image

The Mackay–Whitsunday region covers 9335 km2 (Fig. 3) and contains five major drainage basins which discharge into the Great Barrier Reef: O’Connell; Pioneer; Plane; Proserpine; and Whitsunday Island (Queensland Department of Environment and Heritage Protection 2014a). This region features at least 3529 native species of plants and animals including 56 rare or threatened species,Footnote 1 68 protected areas,Footnote 2 26 national parks,Footnote 3 nine nationally important wetlands,Footnote 4 and the Great Barrier Reef World Heritage Area (Queensland Department of Environment and Heritage Protection 2014a). GDE mapping of terrestrial and surface expression GDEs in the Mackay–Whitsunday region demonstrates the broader value of GDE mapping, particularly in high priority areas such as Great Barrier Reef catchments.

Fig. 3
figure 3

Mackay–Whitsunday study area showing all areas identified as potentially containing GDEs during the technical workshop (held in March 2012) based on local expert knowledge

Full size image

The present method has not been used to map aquifer, estuarine or near-shore marine GDEs. While this is a critical gap, the method is flexible and could incorporate the mapping of these GDEs in the future.

Stage 1: Development of Mapping Rule-Sets and Pictorial Conceptual Models

Stage 1 utilises a process we call “walking the landscape” to systematically capture and synthesise existing knowledge on GDEs (step 1), available spatial data (step 2) and local expert knowledge of landscapes including groundwater and ecosystem interactions (step 3). In step 1 a thorough review is conducted to collate relevant information on the relationship between groundwater and ecosystems, such as scientific publications, reports and survey information. This review of existing knowledge begins to build the essential conceptual understanding of groundwater and ecosystem interaction and the conditions that control this interaction. Knowledge derived from this review can, where appropriate, be used to inform expert discussions in step three. Similarly, in step 2 relevant available spatial data (including remote sensing derived data) is compiled to inform expert discussions in step three and to support the generation of GDE mapping. A review of available spatial data for the Mackay–Whitsunday region identified at least 100 datasets held by the Australian Government, Queensland Government, local governments and regional natural resource management bodies that could be used to support the delineation of GDEs in Queensland. During compilation of spatial datasets priority is given to the use of the best available datasets considering data relevance, currency, accuracy, reliability, attribution, consistency and extent. An inventory of collated datasets, including information on the scale and extent of each dataset, provides an indication of the potential detail of GDE mapping and key data gaps. All spatial datasets for Mackay–Whitsunday region were compiled and stored in a consolidated ESRI ArcGIS® (Redlands, CA, USA) database and sorted by thematic category.

Based on the conceptual understanding and detailed spatial datasets identified in steps 1 and 2, structured technical workshops are held in step 3 to capture multidisciplinary local expert knowledge relevant to the mapping of GDEs and to establish a comprehensive conceptual understanding of groundwater and GDEs in the study area. This step involves walking the landscape, systematically moving through each part of each catchment in the study area, on a catchment by catchment basis. During the technical workshops experts are asked to identify the approximate areas of potential GDEs on large hard copy maps and from here:

  • Develop a pictorial conceptual model;

  • Develop mapping rule-sets; and

  • Identify data needs for implementation of mapping rule-sets (Queensland Department of Environment and Heritage Protection 2012a).

The walking the landscape process has a wide range of applications from assessing aquatic biodiversity to identifying landscape connectivity. Mackay–Whitsunday participants commended the use of the walking the landscape approach in GDE mapping, noting that the approach is simple, systematic, thorough and conducted at an appropriate scale. The range of experts attending these technical workshops should be carefully selected as each expert will bring different knowledge to the process. The best outcomes are derived from a workshop where a diverse range of experts can contribute their information to deliver a consensus driven and comprehensive outcome. This supports the quantitative analysis undertaken by Hampton and Parker (2011) that determined face-to-face interactions was vital to the success of both synthesis activities and multi-institutional collaboration exercises likely due to increased time allotted to the activity and the beneficial sociological aspects of these interactions. This ensures the GDE mapping is guided by a pragmatic holistic understanding of the relevant issues and regional context with local expert input spanning disciplines including aquatic ecology, botany, ecology, geology, hydrogeology, hydrology and soil science. Participants in the technical workshops also need to have detailed local knowledge of how water movements through the study area. For Mackay–Whitsunday, a facilitated technical workshop was held on 12th and 13th March 2012 with 16 experts representing over ten disciplines including aquatic ecology, botany, hydrology and remote sensing.

During the Mackay–Whitsunday technical workshop, experts identified 43 locations that potentially contain GDEs (Fig. 3). A pictorial conceptual model is then developed (step 3a) for each identified location as a simplified representation of the components, processes and interrelationships of a specific system. Pictorial conceptual models are “representations of observed objects, phenomena and processes in a logical and objective way with the aim of constructing a formal system whose theoretical consequences are not contrary to what is observed in the real world” (Queensland Department of Environment and Heritage Protection 2012b). In terms of GDEs, conceptual models illustrate at a minimum:

  • How groundwater moves through a catchment;

  • Any information on the depth to groundwater;

  • The likely location of groundwater recharge and discharge; and

  • The likely location and type of ecosystems potentially groundwater dependent.

Supporting these pictorial conceptual models are detailed descriptions of local ecological, geological and hydrological contexts. These pictorial conceptual models provide natural resource managers and other decision-makers with valuable information on how and why GDEs exist in their catchments in a format that is easy to understand.

The development of these pictorial conceptual models is useful in synthesising multidisciplinary knowledge because they graphically represent the current collective knowledge of expert participants and ensure that any assumptions are explicit. The information contained in these pictorial conceptual models is a key strength to the method with Mackay–Whitsunday participants identifying the pictorial conceptual models as being a valuable tool for a range of future activities including decision-making and science communication. In addition, pictorial conceptual models are often used to underpin the development of numerical modelling designed to assess the impact of proposed or current management decisions on specific groundwater or surface water resources. For example within the Australian Government’s Bioregional Assessment Programme, pictorial conceptual models are used in the Clarence-Moreton bioregion to characterise the interaction between groundwater and surface water in different landscapes (e.g. basalt in areas with different levels of rainfall, areas with different levels of alluvial development, etc.) (Raiber 2015, pers. comm., May 8).

Based on the pictorial conceptual models and expert deliberations, mapping rule-sets are developed (step 3b) to identify the process to delineate GDEs using GIS. A mapping rule-set is simply a combination of attributes that describe the drivers and processes in a landscape that, when applied to spatial datasets in a GIS analysis delineate where ecosystems are or are likely to be groundwater dependent at a catchment scale. Alongside each mapping rule-set a confidence rating is recorded reflecting the degree of confidence experts had in the prediction that identified ecosystems were groundwater dependent and a suite of attributes are captured describing the nature of the connection between groundwater and ecosystems. In step 3c, experts assess available spatial datasets (as identified in step 2) and select the most appropriate dataset to delineate GDEs at the greatest practical level of detail. The absence of a suitable spatial dataset that aligns to a mapping rule-set or component thereof may prevent the implementation of part or all of that mapping rule-set. One major limitation encountered during the Mackay–Whitsunday work was the availability, scale and/or coverage of several specific spatial datasets including existing aquifer and depth to groundwater mapping.

Where the study are being assessed for GDEs is adjacent to an area which has already been mapped, then knowledge acquired from the previous study should be collated in Step 1 and used to guide decision-making in the technical workshop. This existing knowledge, including conceptual models and mapping rule-sets, may be wholly accepted by the local experts for application to the current catchment or modified to suit local variation or extra knowledge gained since the previous mapping was compiled. This process should ensure that the edges of mapped catchments are contiguous with one another and that the mapping continues to be refined through improved understanding of GDEs.

Stage 2: Development of Draft GDE Products

In stage 2, the outcomes from stage 1 are used to develop a draft set of GDE products including pictorial conceptual models (step 4), mapping rule-sets (step 5) and mapping data (step 6).

Hand-drawn pictorial conceptual models developed at the technical workshop (step 3a) are digitised and refined (step 4) based on similarities in the drivers, processes and/or key conditions controlling groundwater ecosystem interaction. During the implementation of the method, the authors found a tapering in the rate of the development of refined pictorial conceptual models indicating that the suite of available pictorial conceptual models is increasingly comprehensive. However, the authors also found that the suite of pictorial conceptual models continued to be refined at a steady rate with revisions to incorporate increased level of detail or develop specific examples of these pictorial conceptual models that are applicable to a local area or region. Currently the suite of pictorial conceptual models is primarily driven by rock type and porosity. For example, alluvial pictorial conceptual models capture GDEs associated with groundwater found extensively in the relatively high porosity unconsolidated sand, gravel and clay of alluvial systems, coastal sand mass systems or inland sand dune fields. Similarly, pictorial conceptual models of rock aquifers capture GDEs associated with groundwater found in metamorphic or igneous rocks with either predominantly primary porosity (i.e. permeable rocks) or secondary porosity (i.e. fractured rocks).

Mapping rule-sets developed at the technical workshop (step 3b) are refined (step 5), to increase the efficiency of the GIS analysis and ensure logical consistency between mapping rule-sets. Mapping rule-set parts are also identified in step 5 which explicitly provide the rationale for an ecosystem’s inclusion in the GDE mapping. One of the major benefits of this mapping method is that different mapping rule-sets parts are developed for different sub-types of GDEs. This allows for confidence in groundwater dependence to be assigned independently for each mapping rule-set part and for improvements in the delineation of different GDE sub-types to be completed independently based on improvements in knowledge or spatial data pertaining to that sub-type of GDE.

In Mackay–Whitsunday consolidation, 10 mapping rule-sets linked to pictorial conceptual models (Table 2; Fig. 4) were developed (Queensland Department of Environment and Heritage Protection 2014b) to describe how groundwater moves through the study area and where and how ecosystems may interact with this groundwater. Six of these mapping rule-sets delineate which ecosystems are potentially dependent on shallow alluvial aquifers. One of the key differences between these six alluvial mapping rule-sets is the nature of ecosystems potential connection to groundwater resources including variability in temporal connectivity (e.g. whether ecosystems have seasonal groundwater connectivity, near-permanent groundwater connectivity, etc.).

Table 2 Example mapping rule-set (MW_RS_01) developed for the Mackay–Whitsunday study area (Queensland Department of Environment and Heritage Protection 2014b)
Full size table
Fig. 4
figure 4

Example pictorial conceptual model, linked to MW_RS_01, developed for the Mackay–Whitsunday study area (Queensland Department of Environment and Heritage Protection 2014c). Refer to WetlandInfo ® for a high resolution, interactive, colour version of this pictorial conceptual model

Full size image

The GIS analysis process used to develop draft GDE mapping data (step 6) utilises specialised software (e.g. ESRI ArcGIS® (Redlands, CA, USA)) to apply each mapping rule-sets to spatial datasets to identify those ecosystems that are or potentially are groundwater dependent.

The 10 mapping rule-sets for Mackay–Whitsunday region were implemented using data models constructed in ESRI ArcGIS Model Builder® (Redlands, CA, USA). Each mapping rule-set was represented by one or more mapping rule-set parts which divided the mapping rule-set into individual mappable components (Table 3). The data model constructed for each mapping rule set contained the implementation of all mapping rule-set parts and results in the output of one dataset per mapping rule-set part (Fig. 5).

Table 3 Example mapping rule set and rule-set parts, for MW_RS_01, developed for the Mackay–Whitsunday region
Full size table
Fig. 5
figure 5

Example model developed using ESRI ArcGIS Model Builder® (Redlands, CA, USA) to implement MW_RS_01 for the Mackay–Whitsunday region

Full size image

Stage 3: Refinement

Stage 3 iteratively refines the developed draft GDE products, repeating component steps until expert consensus is achieved so the products reflect the current knowledge of groundwater and GDEs in the catchment. This stage consists of field validation (optional) and technical review workshop(s). This stage is critical to ensure that the generated products are fit for their intended purpose and are accepted by local experts. Step 7 involves the optional verification of draft GDE mapping products including mapping data, pictorial conceptual models and/or mapping rule-sets using through systematic field validation or comparison of draft products with known GDEs.

In step 8, a technical review workshop is held with local experts to review the draft GDE mapping products developed in stage 2. This second workshop is structured so as to systematically review each mapping rule-set and rule-set part, the spatial data used to implement the mapping rule-set, the related pictorial conceptual model and the resultant GDE mapping data.

For the Mackay–Whitsunday region, a technical review workshop was held on 9th of May 2013 with 14 experts representing 8 disciplines. Feedback captured during the technical review workshop was used to inform the development of refined pictorial conceptual models (step 9), mapping rule-sets (step 10) and GDE mapping data (step 11). The authors have found that the review process undertaken in step 8 has been critical in ensuring the robustness and overall quality of the final GDE mapping products. These technical review workshops produce substantial revisions to the draft GDE mapping products, often addressing the tendency for the over-estimation of GDEs in draft products and rectify any omissions (e.g. areas not assessed for GDEs at earlier technical workshops).

In step 11, development of revised GDE mapping data, the individual datasets developed for each mapping rule-set and rule-set part are aggregated into five GDE datasets. The five GDE datasets produced are based on GDE type and dataset geometry (e.g. point, line, or polygon). Dataset geometry is driven by GDE size and detail in input datasets:

  • Surface expression GDEs (polygons)

  • Surface expression GDEs (lines)

  • Surface expression GDEs (points)

  • Terrestrial GDEs (polygons)

  • Subterranean GDEs (polygons)

Each GDE dataset includes, at a minimum: relevant attributes of ecosystems (e.g. type of vegetation community or wetland); attributes describing groundwater dependency (e.g. temporal connectivity, etc.); and several method specific attributes (e.g. source data used to delineate GDE extent, mapping rule-set name, pictorial conceptual model and expert confidence in groundwater dependence, etc.). It is important that the mapping contains these attributes as they are necessary for users to understand why an ecosystem has been identified as potentially groundwater dependent in an informative, accessible and visually stimulating manner. These attributes are also useful in informing the development of management actions (e.g. managing groundwater extraction to maintain natural base flow patterns) and complement the development of numerical models.

Stage 4: Product Testing

After the refinement process (stage 3), targeted product testing is conducted to identify any errors in the GDE mapping products and assess the usability of the intended delivery mechanism. The suite of GDE mapping products is made available via a closed version of the online interface to a wide range of stakeholders including natural resource managers at national, state and regional levels (incorporating experts from across different jurisdictions), those experts who contributed to the technical workshops, and other experts that have explicit knowledge of the mapped landscapes. A closed release of the suite of Mackay–Whitsunday GDE mapping products was made available to 23 natural resource managers. After a 2 week period, web delivered surveys requesting feedback on the Mackay–Whitsunday GDE mapping products were distributed and 10 responses were received (43 % response rate). Feedback revealed that GDE mapping products were easily accessible, reflected expert knowledge of the landscape and contained appropriate levels of supporting information. Several suggestions were received on ways to expand and improve upon the GDE mapping products. These will be discussed further in “Discussion and Conclusion” section.

Stage 5: Finalisation and Product Release

Stage 5 balances feedback received during the wider quality assurance process (stage 4) with the resource requirements to implement those suggestions. All GDE mapping products including GDE mapping data are an approximation and will be updated over time as data and knowledge improve. Any identified errors from stage 4 should be corrected (step 13) but feedback of a conceptual nature may need to be addressed in later iterations of GDE mapping products. Once all GDE mapping products are finalised they are integrated with any existing GDE mapping and released through available web delivery mechanisms (step 14). GDE mapping products for Queensland, updated with information from Mackay–Whitsunday, data were publicly released in May 2013. GDE mapping data were released as a minor version update, version 1.1, through both the Queensland Government WetlandInfo ® delivery mechanism and the Queensland Globe®, and were also made available for download through the Queensland Spatial Catalogue® and Queensland Government data®. Future updates to GDE products are recommended (step 15) reflecting the dynamism of ecosystems, our rapidly improving understanding of groundwater and ecosystem interaction and improvements to available spatial datasets.

Discussion and Conclusion

Immediate and foreseeable threats to GDEs are widely acknowledged in literature. They include land use change (Kløve et al. 2011), climate change (Eamus and Froend 2006; Kløve et al. 2011) and activities involving groundwater extraction (e.g. irrigation, mining and water supply) (Boulton 2005; Eamus and Froend 2006; Brown et al. 2011). These threats have the potential to alter existing groundwater regimes (Murray et al. 2006; Boulton 2009) including changing groundwater availability (e.g. depth to groundwater, temporal variability, groundwater pressure and flow rate) (Boulton 2005; Kløve et al. 2011) and groundwater quality (e.g. contamination by salts, nitrates, pesticides, petrochemicals, industrial chemicals and heavy metals) (Boulton 2005; Brown et al. 2011; Kløve et al. 2011). Natural resource managers often lack sufficient information at an appropriate scale on the location and attributes of GDEs for adequate consideration of these threats (Kløve et al. 2011). The GDE mapping method described in this paper addresses this fundamental information gap by recognising the value of local expert knowledge and integrating this knowledge with available spatial data and information. The GDE mapping products developed using this method are at a scale compatible with management and planning activities and can be scaled up (e.g. to whole regions) or down (e.g. to a local area) to suit other projects through the integration of experts with knowledge at that spatial scale. Each mapped GDE has attributes describing its potential connection to groundwater (e.g. temporal connectivity, etc.), linking the GDE to the mapping rule-set developed by local experts, linking the GDE to the relevant pictorial conceptual model that illustrates current understanding of hydrogeological function, and describing the ecosystem itself. This allows the user to clearly understand the rationale and data used to delineate an ecosystem as groundwater dependent. The level of information provided in the GDE mapping products is critical in informing assessments of the impact of specific activities on groundwater and hence GDEs in a particular landscape. For example, this conceptual understanding informs the development of numerical models to assess the impact of a proposed level of drawdown from an activity in a specific hydrogeological basin on GDEs.

The participatory approach utilised in this method was not only a valuable exercise in knowledge sharing and network building, but also ensured knowledge held by local experts was included in the mapping. The level of participation sustained throughout the method implementation encouraged expert ownership of the mapping, with increased expert ownership promoting and accelerating the uptake in the use of final GDE mapping products. Pictorial conceptual models synthesising expert knowledge at a range of scales (e.g. site, local and landscape) were also clearly identified by stakeholders as being valuable, particularly in communication (e.g. communication between state government representatives and local landholders), planning (e.g. prioritisation of field assessments and informing targeted assessments such as numerical modelling) and decision-making activities. One of the unintended consequences of the walking the landscape process was that this knowledge synthesis had an educational outcome with workshop participants gaining an improved awareness of groundwater in their area which could then be applied to their other work (e.g. as consideration in development assessment).

The GDE mapping products developed using this method reflect and communicate the reality that groundwater dependence is an attribute of an ecosystem rather than separate ecosystems in a landscape. For example, some wetlands may have some degree of groundwater dependence while others may not. These groundwater-dependent wetlands and non-groundwater dependent wetlands are not necessarily isolated from one another within the landscape nor isolated from one another in many ecological classification frameworks (e.g. the Interim Australian National Aquatic Ecosystem (ANAE) Classification Framework; Aquatic Ecosystems Task Group 2012). However, groundwater dependence can result in ecologically distinct communities that may contain endemic species (Fensham and Fairfax 2003) particularly in areas where groundwater provides sustained water in an otherwise dry landscape. This message is communicated primarily by placing all the GDE mapping products in a broader landscape context and by using existing ecosystem classification frameworks and mapping data. Similarly, pictorial conceptual models place GDEs in a broader landscape context by including information on relevant ecological, geological and hydrological conditions. This approach makes a valuable contribution to developing a holistic understanding of GDEs.

The final GDE mapping products have multiple applications including informing regional water resource planning, assessment and regulation of development activities, future scientific research and identifying and prioritising funding allocation for future GDE related work. The GDE mapping products are not intended as a substitute for field assessments and may be updated when such groundwater investigations are made. Instead, the GDE mapping products are intended to support future management and planning activities including providing a conceptual foundation to support more detailed groundwater resource assessments that may incorporate other mapping and modelling techniques. Recognising the usefulness of the GDE mapping products, stakeholder demand has driven additional GDE mapping using this method with mapping currently underway or completed for approximately 54 % of Queensland. Key stakeholders driving this spatial expansion vary between regions but broadly include various levels of governments and community groups (e.g. regional Natural Resource Management bodies or catchment management groups).

These GDE products, from the mapping to the pictorial conceptual models, are not static as reflected through the versioned releases of GDE products. This updatability is critical given the dynamic nature of ecosystems and our evolving understanding of GDEs and landscape processes. WetlandInfo ®, one delivery platform, enables progressive updates to GDE products through individual product versioning (i.e. versioning for each pictorial conceptual model and GDE dataset independently). Since these GDE mapping products are proposed to be one of the preferred datasets informing water management decisions, it is important that any broader scale GDE mapping (e.g. the National GDE Atlas) is consistent with the best available GDE mapping. Integration with any existing GDE mapping will minimise the risk of inconsistencies in GDE mapping products and ensure that users are presented with a consistent range of products at the highest detail available.

Two key limitations were encountered when implementing the GDE mapping method which impacted upon the quality of the final products. The first limitation is in our current knowledge of the interaction between groundwater and ecosystems. It was evident from the technical workshops that there are significant knowledge gaps, particularly regarding the presence and characteristics of local groundwater regimes, vegetation water use, rooting depths of different plant species and ecosystem response to changes in groundwater availability. Similarly, expert knowledge in specific disciplines (e.g. soil science) in some locations was difficult to obtain. In these circumstances, it was necessary to rely on several experts in related disciplines with long-term knowledge of those locations to address this limitation.

The second limitation arose from the lack of several key spatial datasets and challenges relating to the spatial and temporal scales of available spatial datasets. For example, datasets were available containing thousands of depth to groundwater records unequally distributed across study areas, collected using different methods and at varying temporal frequencies. However, without a consistent method for deriving a depth to groundwater spatial dataset from this database, this data could not be incorporated into GDE mapping as a spatial data set despite its usefulness.

On-going collaboration with stakeholders throughout the implementation of this method has prioritised the range identified knowledge gaps. The key opportunities for future work include: the enhancement of pictorial conceptual models to include further detail on spatial and temporal variations in hydrogeological and ecological conditions; the development of complementary conceptual models to identify ecological responses to different pressures and/or threats; and research to identify the range of rooting depths for key phreatophytes within different landscape types and climatic conditions.