Combining in-situ water quality and remotely sensed data across spatial and temporal scales to measure variability in wet season chlorophyll-a: Great Barrier Reef lagoon (Queensland, Australia)

The primary biological response to nutrient enrichment in aquatic environments, given suitable conditions such as light availability and water temperatures, is the growth of phytoplankton and higher plants. Known consequences of nutrient enrichment on the phytoplankton community can be measured by elevated phytoplankton biomass (Boynton et al. 1996;Bricker et al. 2003;Brodie et al. 2007) and alterations of the natural phytoplankton community composition, which may in turn change ecosystem food web and nutrient cycling dynamics (Cloern 2001;Brodie et al. 2005;Devlin et al. 2009).

The understanding of nutrient enrichment and potential eutrophication in the Great Barrier Reef (GBR) has progressed in recent years with key publications identifying responses to nutrient loading for corals (Fabricius 20052011;Brodie et al. 20112012) and seagrasses (Schaffelke et al. 2005;Collier et al. 2012;McKenzie et al. 2010), as well as phytoplankton communities (Furnas et al. 2005;Sorokin and Sorokin 2010;see references in Devlin et al. 2013). The nutrients introduced or released during high flow river discharge into the GBR are rapidly taken up by pelagic and benthic algae, and microbial communities (Alongi and McKinnon 2005), sometimes nourishing short-lived phytoplankton blooms and high levels of organic production (Furnas 1989;Furnas et al. 20052011). This organic matter is cycled through the marine food web and transformed, for example, into marine snow particles that may be deposited onto benthic communities, such as coral reefs, thus influencing their structure, productivity, and health (Anthony and Fabricius 2000;Fabricius and Wolanski 2000). Further, Brodie et al. (2005) and Fabricius et al. (2010) identify enhanced nutrient supply in river runoff as a critical requirement for enhanced Acanthaster planci (commonly known as the crown-of-thorns starfish or COTS) larval survival in the GBR, a finding that supports Lucas’ (1982) hypothesis that COTS suffer high levels of larval starvation in the absence of phytoplankton blooms. COTS are of great importance in the GBR as a predator on coral and are recognised as one of the leading causes of mid-shelf coral mortality in the GBR (De’ath et al. 2012).

Measurement of chlorophyll-a (hereafter chl-a) concentrations derived from phytoplankton biomass can be indicative of enhanced nutrient inputs (Spencer 1985;Furnas et al. 2005;Brodie et al. 2007) and thus provides a simple, reliable indicator of water column nutrient status in the absence of high frequency water column nutrient concentration measurements (Harding and Perry 1997;Yunev et al. 2002). Chl-a concentrations in the GBR waters have been successfully utilised as a proxy measurement of phytoplankton biomass and nutrient concentrations for a number of monitoring programs in the past (Brodie and Furnas 1994;Brodie et al. 19972007;Furnas and Brodie 1996;Steven et al. 1998).

The frequency of in-situ sampling methods is associated with cost, time, weather, and logistic constraints and only a limited number of sampling points can be monitored during any one flood event. This restriction might limit the assessment of the range and variability of coastal processes and water constituents (Novoa et al. 2012). Conversely, satellite sensors provide an effective means for frequent and synoptic water quality observations over large areas. Remotely sensed data (e.g., derived ocean colour products) have increasingly become a tool by which scientists and managers can supplement long-term monitoring datasets of in-situ water quality measurements (Brando et al. 2010;Kennedy et al. 2012;Schroeder et al. 2012;Devlin et al. 2012b). Remotely sensed data have thus provided an essential source of information related to the movement and composition of flood plumes in GBR waters, extending the utility of the in-situ sampled data at both regional and temporal scales.

Assessing chl-a concentrations with remotely sensed data in optically complex coastal waters, such as the GBR coastal waters, where suspended sediment and coloured dissolved organic matter (hereafter CDOM) co-occur with phytoplankton, is notoriously difficult (e.g., Gitelson et al. 2009;Odermatt et al. 2012). The standard and global bio-optical algorithms used in oceanic waters (Case 1 waters; Morel and Prieur 1977;Gordon and Morel 1983;Morel 1988) are inaccurate when applied to optically complex coastal waters (Case 2 waters; Morel and Prieur 1977;Gordon and Morel 1983;Morel 1988), although regional parameterisation of these algorithms can help increase their accuracies (Naik et al. 2013;Brando et al. 2012). Some recent studies have investigated the use of remote-sensing products in the mapping of water quality parameters, particularly the use of chlorophyll and CDOM products, both for global and regional assessments (Devlin et al. 2012b;Brando et al. 20102012;Schroeder et al. 2012). However, the retrieval of chlorophyll data based on the application of bio-optical algorithms (e.g., Level 2, chl-a product) is still problematic in Case 2, turbid nearshore waters (Wang and Shi 2007;Gitelson et al. 2009;Odermatt et al. 2012).

Our study utilises the variation in ocean colour to map the gradient in water quality concentrations across flood plumes to avoid these inherent problems of retrieving chl-a measurements in Case 2 waters. The same approach has been applied to map flood plumes using moderate resolution imaging spectroradiometer (MODIS) imagery, allowing assessment of their extent and influence on GBR ecosystems (Bainbridge et al. 2012;Brodie et al. 2010;Devlin et al. 2012a, b;Schroeder et al. 2012;Álvarez-Romero et al. 2013).

Our study extends earlier work on spatial and temporal patterns of chl-a distribution in the GRB (Steven et al. 1998;Brodie et al. 2007) with an analysis of a long-term dataset of chl-a concentrations collected in high flow conditions between December to April (wet season) from 2006 to 2013. This paper presents in-situ chl-a measurements sampled over variable conditions associated with the North Queensland wet season, coupled with spatial mapping outputs derived from MODIS true-colour images. The method applied in this work allowed us to assess the extent and characteristics of the dominant water of the flood plumes, without depending on the retrieval of Level 2 MODIS data in Case 2 waters. The information on the spatial extent of different flood plume water types, integrated with high frequency wet season sampling, provides a valuable tool from the environmental management perspective to assess the extent of the spatial and temporal distribution of chl-a concentrations and, hence, phytoplankton biomass in the Great Barrier Reef.

Simple indices of phytoplankton biomass, as measured by chl-a, provide an accurate means of quantifying broad-scale water quality within the Great Barrier Reef region. These data can be used as a baseline for ongoing investigation of impacts of increased nutrient discharges into Great Barrier Reef waters, such as assessing the role of altered water column nutrient status on COTS outbreaks and the influence of agriculture and urban coastal settlement on regional water quality.

The onset of the wet season and high flow conditions provides nutrient-rich waters that can nourish and enhance phytoplankton growth. The magnitude and timing of flow are important co-factors in the distribution of flood waters and the range of water quality concentrations measured over the sampling period. Inner-shelf reefs of the wet tropics region—an area of high rainfall (1,200–4,000 mm average annual)—are exposed regularly (one to several times per year) to land-based pollutants via flood plumes (Devlin et al. 2001;Devlin and Schaffelke 2009;Lewis et al. 2012), while inner-shelf reefs of the regions to the south of the wet tropics (i.e., Burdekin, Mackay, Fitzroy, and Burnett regions) are exposed less frequently but to larger flood plumes that often travel great distances (Devlin et al. 20112012a, b). Flow conditions have always been variable, however there is a strong consensus that we are now seeing the impacts from more extreme weather conditions through these higher flow conditions (Elmhirst et al. 2009;Johnson et al. 2013).

Comparative analysis of phytoplankton biomass measured in plume conditions with co-factors such as flow and salinity (Figure 6) is key information for understanding the response of phytoplankton standing stock to the input of nutrient-enhanced flood waters. Mean concentrations of chl-a measured in the flood plumes (Figure 8) (0.77–2.55 μg L^-1) are higher than the annual water quality guideline (0.45 μg L^-1,GBRMPA 2010). The highest concentrations were recorded during high peak flows (Q₉₅; Figure 6) for the Tully, Burdekin, and Fitzroy rivers. The Herbert River also showed high concentrations at lower flow quartiles (Q₀₅, Q₁₀), potentially reflecting the complex hydrodynamics around Hinchinbrook Island and the southern influence of the Burdekin River. First flush events, even at low flow conditions, are responsible for the increased delivery of dissolved inorganic nitrogen into the coastal zone (Devlin and Schaffelke 2009) from rivers with catchments containing large areas of fertilised cropland (Mitchell et al. 2009). These higher values at lower flow quartiles also suggest that the productivity in the coastal system is enhanced outside of the high flow periods. High chl-a in low river discharge conditions can be driven by sediment resuspension events in shallow waters (depth less than 10 m) where dissolved inorganic nitrogen and dissolved inorganic phosphorus mineralised from particulate nitrogen and phosphorus in benthic sediment are injected back into the water column (Chongprasith 1992;Walker 1981;Muslim and Jones 2003;Ullman and Sandstrom 1987). Resuspension of benthic sediment during cyclonic conditions occurs down to depths of 30 m and also causes phytoplankton blooms on the GBR shelf (Furnas 1989).

Over the wet season, however, the dominant driver of high chl-a concentration is the delivery of nutrient-enriched river waters. A comparison of daily flow data from the Tully River with daily in-situ chl-a data for the period 2006–2013 showed peaks in river flow are highly correlated temporally with elevated chl-a concentrations (Figure 7). Using data collected with high frequency for one key catchment shows that chl-a concentration is high in comparison to the GBRMPA annual guideline (GBRMPA 2010) and elevated above other reported concentrations (Brodie et al. 2007;Schaffelke et al. 2012; Figure 8). The recommended GBRMPA water quality guideline for chl-a is an annual mean measurement and, as such, is not directly comparable against chl-a data collected through the wet season only. However it does provide a reference at which to compare the chl-a data collected in the high flow periods over multiple wet seasons and illustrates the higher values measured within the wet season in the Tully marine region. Brodie et al. (2007) reported on a long-term Great Barrier Reef chl-a monitoring program which demonstrated a significant cross-shelf difference in chl-a concentrations. Data from this program also highlighted a strong regional pattern, where the inshore wet tropics had high wet season concentrations in comparison to the northern and southern transects. Brodie et al. (2007) suggested this variation in baseline and seasonal chl-a concentration was related to the differences in catchment loads where fertilised agriculture in the wet tropics contributed to high concentrations of dissolved inorganic nitrogen in riverine flow (Kroon et al. 2012;Brodie and Waterhouse 2012;Wooldridge et al. 2006). In the Tully River nitrogen concentrations rose significantly in the period 1987–2000 (Mitchell et al. 2001). Particulate nitrogen concentrations doubled and nitrate concentrations increased by 16% over this period (Mitchell et al. 20012009). The cropping activities, primarily sugar cane cultivation, are concomitant with high fertiliser application rates and higher nutrient delivery in the rivers (Brodie and Mitchell 2005Brodie et al. 2013;Kroon et al. 2012;Waterhouse et al. 2012).

The frequency of water types across the flood plume is comparable with our current understanding of water quality gradients and supports the validity of the classification method applied in this study. The higher frequency associated with the primary flood plume water type (f_p) reflects the intensity, duration, and constituent concentrations of the river discharge but is limited to a small near-shore zone (< 10 km across the shelf). These characteristics are strongly linked to the catchment hydrology and land use practices. For example, the two larger catchments over the GBR, which are under extensive agricultural development (i.e., Burdekin and Fitzroy rivers, which have greater than 80% of area utilised for agriculture), are associated with a larger area of turbid primary waters.

Secondary and tertiary plume frequency (i.e., f_s and f_t) cannot always be attributed to an individual river, particularly in the case of wet tropics rivers, which can merge into one heterogeneous plume. The area of secondary flood plume types is influenced by the onset of the primary plume through the river discharge and the local hydrodynamic conditions, mainly controlled by tides (e.g., Valente and da Silva 2009), wind regime (Dzwonkowski and Yan 2005;Lihan et al. 2008), bathymetry (e.g., Lee and Valle-Levinson 2013), and Coriolis force (Geyer et al. 2004). Mapping the annual frequency of the secondary water type (Figure 12) gave a qualitative estimate of the area where high concentrations of chl-a have occurred during wet season conditions between 2006 and 2012.

The chl-a concentration distribution is variable through the three water types, and in both primary and secondary flood plume plots, there is a peak of concentration where f_p and f_s equal 0.5, suggesting that there is a transition zone between these two water types that presents optimal environmental conditions leading to high chl-a concentrations. Figure 11 suggests that these chl-a peaks are driven by a reduction in both TSS and K_dPAR values. Waters dominated by primary (e.g., f_p > 0.8) or secondary (e.g., f_s > 0.8) plumes usually exhibit lower chl-a concentrations compared to waters where plume type is more variable (e.g., f_p or f_s < 0.6).

These transitional conditions potentially represent an ocean front where the denser water under-rides lighter water, giving rise to an inclined interface and a strong convergence at the surface, which can concentrate phytoplankton and pollutants (Klemas 2012).

In complex coastal areas, field data and ecological evidence are very difficult to acquire because of the temporal variability in spatial patterns of all ecological conditions and biophysical drivers causing change. Flood plume water types over multi-annual (Figure 9), annual (Figure 12), and weekly (Figure 13) time scales provide synoptic information in monitoring the water quality conditions of GBR coastal waters.

Some limitations are inherent in the true-colour classification approach used to categorise plume type in this study. One limitation described in Álvarez-Romero et al. (2013) is the presence of other surface water masses that can be confounded with plume waters. For example, Broad Sound Channel and Shoalwater Bay, north of the Fitzroy River mouth, are areas where high turbidity occurs in the absence of plumes, mostly associated with strong tidal currents (Kleypas 1996), resulting in low confidence for accurate classification. Another limitation is the use of ocean colour data alone. Stretching of the red-green-blue images produced in SeaDAS is image-dependent and, thus, true-colour classes might be inconsistent over time. Furthermore, no atmospheric correction is applied to the true-colour images and might lead to misclassification. The impact of such misclassifications on the estimation of the primary, secondary, and tertiary water type maps should be further investigated. Despite these acknowledged limitations, the utility of the classification method is corroborated by the measurements of in-situ water quality parameters among the three water types observed over multiple years (Figure 9). The field observations reveal meaningful quantitative differences when organised along the gradients of flood plumes classified with the true-colour approach.

Connections between environmental variations and ecological systems occur across a large range of interacting spatial, temporal, and organisational scales. Collecting information at temporal or spatial scales that capture the environmental frequency and variability of patterns is essential for the decision-making process (Ostendorf 2011). Remote-sensing technologies can provide the synoptic window necessary for inventory and mapping of ocean ecosystems with information at different spatial and temporal resolutions. Using remotely sensed information in conjunction with in-situ data provides complementary approaches which allow extrapolation of the site data to a larger area and, conversely, provides validation of the remote-sensing data.

Mechanisms by which phytoplankton biomass is controlled within GBR waters are through a number of co-determinants including existing nutrient pools, export and import of GBR nutrients, disturbance, grazers, and physical conditions. The higher nutrient concentrations and shallower water present in the inshore areas would typically result in high episodic pulses of phytoplankton growth. However, the difference in the productivity of the inshore GBR driven by natural processes and productivity which has increased through nutrient enrichment, resulting in the accelerated growth of phytoplankton beyond a natural threshold, can be difficult to resolve.

The current marine monitoring water quality program provides valuable information on a regional basis which can be and has been used to map acute pressure from polluted river waters (Devlin et al. 20102012a). The value of intensive sampling around the formation and development of flood plumes in the coastal area is essential in our understanding of the short-term influence of river flow enhanced by sediments, nutrients, and pesticides.

Knowledge of the areas and types of ecosystems that are the most likely to be impacted by changing water quality can help focus our understanding on what types of ecological impacts are occurring to those systems to help inform marine, coastal, and catchment management (Brodie et al. 2012). Details of the movement of pollutants and frequency of inundation can be key measurements in attributing water quality decline to ecosystem change. Detection and monitoring of nutrient enrichment and associated impacts in the Great Barrier Reef require a focus on the analysis of long-term in-situ data (Furnas et al. 2011;Schaffelke et al. 2012) and the inclusion of data collected over high flow events. Resolving the spatial extent of nutrient-enriched waters through the measures of high chl-a requires the analysis of remotely sensed data to fully capture the wet season variability. Combining these data sources enhances the ability to describe and distinguish anthropogenic impacts from natural wet season processes in the GBR coastal system.