Climate phase drives canopy condition in a large semi-arid floodplain forest

Journal of Environmental Management 159 (2015) 279e287

Contents lists available at ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Research paper

Climate phase drives canopy condition in a large semi-arid floodplain forest Li Wen a, *, Neil Saintilan b a b

Water, Wetlands and Coastal Science Branch, NSW Office of Environment and Heritage, 59 e 61 Goulburn Street, Sydney 2000, Australia Department of Environmental Sciences, Macquarie University, NSW 2109, Australia

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 November 2014 Received in revised form 18 May 2015 Accepted 20 May 2015 Available online 28 May 2015

To maintain and restore the ecological integrity of floodplains, allocating water for environmental benefits (i.e. environmental water) is widely practised globally. To efficiently manage the always limited environmental water, there is pressing need to advance our understanding of the ecological response to long-term climate cycles as evidence grows of intensification of extreme climatic events such as severe drought and heat waves. In this study, we assessed the alleviating effects of artificial flooding on drought impact using the canopy condition of the iconic river red gum forests in Australia's Murray Darling Basin (MDB). To achieve this, we jointly analysed spatial-temporal patterns of NDVI response and drought conditions for the period of 2000e2013, during which the MDB experienced an extreme dryewet cycle. Our results indicated that while NDVI-derived canopy condition was better at the sites receiving environmental water during the dry phases, both watered and unwatered sites displayed great similarity in seasonality and trends. Furthermore, we did not find any significant difference in NDVI response of the canopy between the sites to suggest significant differences in ecosystem stability and resilience, with watered and unwatered sites showing similar responses to the extreme wet conditions as the drought broke. The highly significant relationship between long-term drought index and NDVI anomaly suggest that climate phase is the main forcing driving canopy condition in semi-arid floodplain forests. Crown Copyright © 2015 Published by Elsevier Ltd. All rights reserved.

Keywords: River red gum Environmental water ~ o-Southern Oscillation (ENSO) El Nin BFAST (Breaks in Additive Season and Trend) Vector autoregression

1. Introduction Changes in the flow regimes of major rivers worldwide have contributed to the decline of flood-dependent biota, and to the dieback of floodplain forests in arid and semi-arid regions lez et al., 2010; Mac Nally et al., 2011). The situation is ex(Gonza pected to deteriorate with the intensification of extreme climatic events such as severe droughts and heat waves (Thornton et al., 2014). The interception of flows in large storages, and water diversion for irrigation and town water supply combine to reduce the frequency, magnitude and duration of flows required to sustain the flood-dependent biota in many river systems (Kingsford, 2000). Water delivery strategies in support of irrigated agriculture can also change the seasonality of flow, with adverse ecological impacts (Bunn and Arthington, 2002). While the hydrological alterations caused by water resource development are profound (Wen et al., 2013), they occur in the context of a naturally variable flow

* Corresponding author. E-mail address: [email protected] (L. Wen). http://dx.doi.org/10.1016/j.jenvman.2015.05.027 0301-4797/Crown Copyright © 2015 Published by Elsevier Ltd. All rights reserved.

regime, and adaptation to cycles of wet and dry are a feature of the growth and reproductive strategies of biota occupying these landscapes (Rogers and Ralph, 2010). It can therefore be challenging to differentiate between the ecological effects of water development and the over-riding signal of natural variability, particularly at the inter-decadal and inter-annual scales favoured by most monitoring programs (Colloff et al., 2015). High inter-annual and inter-decadal variability in flow is a characteristic of the rivers of the Murray Darling Basin (MDB), Australia's largest river system. In the northern catchments, river flows are amongst the most variable of all the world's rivers (Puckridge et al., 1998), and the timing and volume of discharge are strongly controlled by decaying tropical lows during the monsoon. The southern catchments supply a large and more reliable flow than the northern tributaries of the MDB, though here phases of the ~ o-Southern Oscillation (ENSO) (Allen, 1988) and the Indian El Nin Ocean Dipole (Ummenhofer et al., 2009) exert a strong influence on the timing and duration of droughts. The Pacific Decadal Oscillation appears to modulate the frequency and intensity of the ENSO (Power et al., 1999) and associated flooding patterns in the northern

280

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

and central Murray-Darling Basin (Ralph and Hesse, 2010). Complicating the assessment of the relative significance of natural and anthropogenic impacts on flow regime is the potentially over-riding signal of climate change. The best case scenario of the impact of climate change on the MDB is a projected decrease of 12% of available surface water to 2030, an estimate which combines the impact of water resource development, climate variability and climate change (CSIRO, 2008). However, variation in climate change assessments of projected surface water flow is high, as models are driven by circulation patterns sensitive to small changes in temperature. Further, the links between global radiative forcing the operation of large oceanic-atmospheric circulation cycles like ENSO, are poorly represented in the current models, and debate continues as to the influence of climate change on this source of variability (van Oldenborgh et al., 2005). Recent conditions in the MDB, illustrate well the confounding interactions between the multiple stressors of water resource development, climate change and climate variability. The “Millennium Drought” (1997e2009) was the most severe drought on record for southeast Australia with the longest uninterrupted series of years with below median rainfall since 1900 (BOM, July 2012). Floodplains in the MDB were among the most severely impacted ecosystems, with large areas of floodplain forest dieback throughout the basin widely observed and reported (Gawne et al., 2011), and attributed to a lack of flood events (Leblanc et al., 2012). The below average rainfall conditions reached greatest severity in 2007, but continued to 2009. Studies at this time reported substantial declines in tree condition in the southern MDB over the preceding decades (MDBC, 2005; Cunningham et al., 2007, 2009). Cunningham et al. (2010) reported that 79% of red gum forests on the Murray River floodplains experienced some degree of dieback with tree canopy less than 80% of potential crown. This figure, consistent between 2006 and 2009, was significantly higher than the period between 1990 and 2003 for which there was little change (Mac Nally et al., 2011). To alleviate the impacts of the extreme dry condition, targeted environmental watering was practised across the MDB. The investment of substantial resources in environment water by Federal and State government jurisdictions has focussed management attention on the development of monitoring methods capable of testing the efficacy of environmental water as an intervention strategy. The broad spatial scale of watering actions (with investment across 20 catchments within a 1 million km2 river system) favours the application of remotely-sensed metrics linked to key environmental targets. The river red gum (Eucalyptus camaldulensis) dominates the rivers, watercourses and wetlands of the inland regions of Australia across all climatic zones (Colloff, 2014), providing important ecosystem services including the provision of habitat for a range of biota (Leslie, 2001), and the cycling of carbon (Robertson et al., 1999). River red gum stand condition is commonly used as a metric of environmental monitoring because of the perceived link with resilience and dieback. Stand condition has been defined as ‘the amount of canopy present relative to the maximal potential canopy, considering stand age, and natural abiotic and biotic limitations’ (Cunningham et al., 2007). Forest dieback, has been defined as a progressive reduction in the crowns of individual trees leading to widespread mortality (Cunningham et al., 2009). Cunningham et al. (2009) tested several stand-level structural and morphological variables to identify consistent indicators of stand condition, finding that percentage live basal area (an index of mortality), plant area index, crown depth and crown vigour (the percentage of the potential crown that contained foliage) were all highly correlated. Importantly, they found that all variables closely correlated with remotely-sensed NDVI (Normalized Difference Vegetation Index), allowing for broad-scale

assessment of stand condition. While previous studies have documented the relationships between declining water availability and declining tree stand condition (Cunningham et al., 2009; Mac Nally et al., 2011), they have been limited by a largely uni-directional trend (i.e. decreasing) in both metrics in the MDB until the end of the last decade (i.e. within a drought event). The resilience of rive red gum to drought is best tested by post-drought recovery, and for the MDB the return of wet conditions following 2009 provides a unique natural experiment into factors limiting recovery. In particular, the efficacy of environmental watering actions during dry periods in promoting survival and recovery of tree stand condition can be tested by incorporating the history of watering actions into the experimental design. The absence of suitable hydrological models for the Murray River floodplain forests prevented this in the study of Cunningham et al. (2009). In this study, we focus on the assessment of the alleviating effects of artificial flooding on drought impact e or footprint e on the canopy conditions of the iconic river red gum forests by jointly analysing spatial-temporal patterns of canopy NDVI response and drought conditions between 2000 and 2013. Furthermore, we evaluate the appropriateness of the freely available MODIS NDVI as an indicator of forest stand condition for environmental water monitoring. The study period encompasses two climatic extremes; the longest and most intense drought on record (2000e2009), followed by the wettest year on record (2010) and several successive above-average rainfall years. We used the standardised precipitation and evaporation index (SPEI) computed from monthly climatic records (1900e2013) to evaluate drought condition. We used NDVI time series (2000e2013, during which MDB experienced an extreme drought e flooding cycle) to assess the vegetation cover response by trend and change-point analysis for sites with high canopy coverage. The underlying hypothesises include: (i) artificial flooding through environmental water application could enhance NDVI as the improved soil moisture conditions will trigger vegetation responses (Breshears et al., 2005) and promote the vigour of tree stand (Wen et al., 2009; Doody et al., 2014) reflected in the dynamics of NDVI time series (Asner and Alencar, 2010; Hl asny et al., 2015); (ii) the repetitive artificial flooding could prolong the functioning wet phase (Colloff and Baldwin, 2010) and maintaining a more vigorous vegetation cover, hence the NDVI time series would exhibit a higher time persistence, which can be expressed as the degree of temporal dependency (Dakos et al., 2012; de Keersmaecker et al., 2014), and the effects of SPEI on NDVI would be smaller; and (iii) ecosystems with longer functional wet phase could be more resilient to major stress (such as drought) and have higher rate of recovery after disruption (Pimm, 1984), which might be quantified as the magnitude of NDVI increase (decrease) during the transition of returning to wet (dry) phase. 2. Method 2.1. Study area e Yanga National Park The study site is Yanga National Park, part of the Lowbidgee Wetland Complex at the western end of the Murrumbidgee River (Fig. 1). With an area of over 200,000 ha, the ecological value of Lowbidgee is recognised as critical fish and waterbird habitats and refuge in arid and semi-arid Australia (Maher, 1990; Wen et al., 2009), and is listed in the Directory of Important Wetlands in Australia (Australian Nature Conservation Agency, 1996). The area has a semi-arid climate with low rainfall, hot summers and mild winters. The average annual rainfall (1900e2011) is about 320 mm, and there is little seasonal variation (Wen et al., 2009). The mean

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

281

located further away from the Murrumbidgee, and unlikely to be flooded during targeted environmental watering events. However, they were subject to flooding during large floods. During the study period, they were flooded only once in 2010e2011, which corre~ a peak on record (BOM, July 2012). sponded to the strongest La Nin To minimize the impacts of understorey vegetation, all sites chosen had a crown canopy cover greater than 75% (McCosker, 2008).

2.2. Drought index We computed the long-term (1900e2013) Standardised Precipitation-Evapotranspiration Index (SPEI) using monthly rainfall and temperature at Balranald (34.65oS, 143.56oE) downloaded from the Australian Bureau of Meteorology (http://www.bom.gov. au). The SPEI is based on precipitation and temperature data, and has the advantage of combining a multi-scalar character with the capacity to include the effects of temperature variability on drought assessment (Vicente-Serrano et al., 2010). We used the R package “SPEI” (Beguería and Vicente-Serrano, 2012) to compute SPEI at three time scales with 3, 6 and 12 month lagging windows, corresponding to SPEI1, SPEI2, and SPEI3. The procedure to calculate SPEI involves a climatic water balance, the accumulation of deficit/surplus at different time scales, and adjustment to a Log-logistic probability distribution. Much of the variability in Australia's climate is connected with the Southern Oscillation, which is measured by a simple index, the Southern Oscillation Index (SOI) (BOM, July 2012). To illustrate the close linkage between SPEI and SOI, we downloaded the long-term monthly SOI (1900e2013) from BOM and calculated the 6-month moving-average SOI.

Fig. 1. The distribution of river red gum in Yanga National Park. The sampled forest, located downstream of Redbank Weir, has crown cover greater than 75%. Inset A and B show the Murray Darling Basin in Australia and the location of Yanga National Park within the Murrumbidgee Catchment.

maximum summer and mean minimum winter temperatures in the same period are 32.2
C and 16.4
C, respectively. Within the park, river red gum (E. camaldulensis) forests and Black Box (Eucalyptus largiflorens) woodlands intermingle with freshwater swamps and open-water lakes. There are more than 20,000 ha of river red gum forest and woodland in Yanga National Park, making it Australia's fourth largest contiguous river red gum stand (Wen et al., 2009). Like many other floodplains in the MDB, Yanga National Park was affected by prolonged drought following the turn of the century, and infrequent flooding due to increasing water resource development for urban and agricultural use. Consequently, ecological degradation has been prevalent in the park (Wen et al., 2011). In recent years, the provision of environmental water is the primary government instrument to restore and maintain the ecological value of the wetlands in the park (Wen and Saintilan, 2014). We selected four river red gum sites (Fig. 1) to test the three null hypothesises. The dominate ground cover species at these sites include Sclerolaena muricata, Paspalidium jubiflorum and Eleocharis sphacelata. The coverage is generally small (<10%), and over 90% of the ground is covered by river red gum litter. Two of the sites, Two Bridges (115 ha), and three wetland patches adjacent to Piggery Lake (166 ha), were subjected to repetitive environmental watering during the period of study. The sites were frequently inundated with irrigation water prior to Yanga being designated as a national park in 2005. For comparison, two sites, the McCabes Gap (314 ha) and Top Norackwell (273 ha), were chosen representing sites

2.3. Satellite-derived vegetation indices We chose to use the MODIS NDVI because of the high temporal frequency (16-day composites) and moderate spatial resolution (250 m). The index is often referred as a proxy for photosynthetic activity, designed to provide consistent, spatial, and temporal comparisons of global vegetation conditions. They have been applied successfully to monitor vegetation dynamics in many biomes (Zhang et al., 2003; Liang and Schwartz, 2009), including wetlands (Ward et al., 2013; Powell et al., 2014a,b), and river red gum stand condition in the southern Murray-Darling Basin (Cunningham et al., 2009). The MODIS NDVI products (MOD13Q1) are originally supplied by the U.S. Geological Survey (USGS) Earth Resources Observation and Science Center (EROS, http://LPDAAC.usgs.gov). The NDVI products were further processed and distributed for Australia by CSIRO Marine and Atmospheric Research (Paget and King, 2008), and available online (https://remote-sensing.nci.org.au/u39/public/ html/modis/lpdaac-mosaics-cmar). We downloaded the 16-day NDVI products and calculated the monthly NDVI time series from Feb 2000 to July 2013 using the mean value of two adjacent 16-day NDVI composites in each month. Abnormal and null values were adjusted/filled using adjacent images or neighbourhood values using GIS focal function. For each of the four sampling sites, we created 12e23 random points depending on the size of the forest patch. The shortest distance between any two randomly placed points was restricted to be greater than 250 m in line with the grid size of NDVI. In addition, we excluded sampling points which were within 100 m of the patch boundaries to avoid the edge effect (Alignier and Deconchat, 2013). The final dataset included 57 time series (12 at Pigery Lake, 10 at Two Bridges, 20 at Top Narockwell, 15 at McCabes Gap).

282

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

2.4. NDVI time series analysis We used two approaches to assess the impacts of environmental watering on river red gum canopy conditions using NDVI time series as proxy. Firstly, we applied the BFAST (Breaks in Additive Season and Trend) algorithm developed in R (http://bfast.R-Forge. R-project.org/) to the site-averaged NDVI time series. BFAST decomposed the time series into trend, seasonal, and remainder components. The algorithm is based on an additive decomposition model that iteratively fits a piecewise linear trend and seasonal model to the data (Verbesselt et al., 2010a; 2010b):

Yt ¼ St þ Tt þ εt

for t ¼ 1; 2…; n

(1)

where Yt is the observation data at time t and Tt ; St and εt are, respectively, the trend, seasonal, and remainder variation components of the data. The algorithm assumes that the trend component (Tt) is piecewise linear with potential breakpoints that can be determined by fitting the linear models iteratively to different segments of the data in a moving window. The abrupt changes are detected by minimizing the residual sum of squares, and their optimal positions in time series can be determined based on the Bayesian information criterion (Zeileis et al., 2003). Thus, for a time series that has m breakpoints T 1 ; T 2 ; …T m , the algorithm fits m linear models for the mþ1 segments as

Tt ¼ ai þ bi t for T

i1


i

and i ¼ 1; 2; …; m:

(2)

where ai and b1 are the mean and trend for the segment i. The seasonal component is represented by a harmonic model:

St ¼

K X k¼1

 aj;k sin

2pkt þ dj;k f

 (3)

for j ¼ 1, …, l change points, f is the frequency of observation, and K is the number of harmonic terms. The technical details of BFAST algorithm can be found in Verbesselt et al. (2010a and 2010b). We applied BFAST to NDVI time series from watering and non-watering sites, and compared the modelled abrupt changes in locations, trends and seasonality to detect any effort of environmental watering. In addition, we extracted the magnitude of change at each change points from the fitted BFAST models, which has a unit of NDVI unit per month, to compare the capacity of recovery from disturbance of the watered and non-watered forests. Secondly, we used multivariate time series models, known as VARs (Vector Autoregression) to investigate the impacts of drought on river red gum forest canopy conditions. A VAR is a structure whose aim is to model the time persistence of a vector of time series observation, Yt, via a multivariate autoregression:

Yt ¼ A1 Yt1 þ / þ Ap Ytp þ BXt þ 2t

(4)

where p is the number of lags (order of the VAR); the vector Xt contains a set of exogenous variables, including a constant, trend and drought indices; and the vector 2t is assumed to be white noise. For each fitted VARs, we calculated the sum of AR (AutoreP gression) coefficients (i.e. A ¼ pi¼1 Ai ). As the sum of AR coefficients quantifies the time persistence expressed as the degree of temporal relation between observations (Dakos et al., 2012), it offers a measurement the stability of a system (de Keersmaecker et al., 2014). Hence, the differences in the sum of AR coefficients might reveal the impacts of environmental watering on the long-

tern stability of river red gum forests. To enable the direct comparison between sampling sites, we standardized the NDVI time series before fitting the multivariate time series models, and specified the intercept to be zero. 3. Results 3.1. Long-term behaviours of NDVI time series The fitted seasonal, trend and remainder components of the site-averaged NDVI time series are presented in Fig. 2. The seasonal cycles were similar for all sites with the annual magnitude of around 0.12e0.15 corresponding to annual wetedry cycle, and no change in seasonal patterns was detected for all sites. For the trend component, however, three abrupt change points were discovered (Fig. 2). Although the exact time when the changes occurred varied among sampling sites, the entire period (i.e. 2000e2013) was divided into four well defined phases, and the direction of trends was the same for all sites (Fig. 2) being: 1) a significantly decrease (p < 0.001 for all sites) followed by a short-term jump; 2) the NDVI continuously decreased significantly to the all-time low in 2008; 3) a rapid recovery phase when increase in NDVI value was significant (p < 0.001) and the NDVI value reached all-time high in 2010 or 2011; and finally 4) the NDVI again dropped significantly to the end of study period. The comparison between NDVI trends at different sampling sites is illustrated in Fig. 3. NDVI was generally higher at the watered sites than at unwatered sites. On average, the NDVI value was 0.08 higher at the watered sites, but the difference could be as high as 0.13 during drought and as low as 0.002 in wet years (Fig. 3). Despite the above differences, the fitted trends corresponded well to the 12-month SPEI in terms of both direction and slope (Fig. 3). The relatively short and low magnitude wet period of February to December 2005 (i.e. positive SPEI3) led to the first jump (abrupt change) of 0.17, 0.18, 0.20, and 0.21 in the average NDVI values within a month at Piggery Lake, Two Bridges, McCabes Gap and Top Narockwell respectively (Fig. 3). The second longer and higher magnitude wet period of May 2010 to November 2012 led to the recovery in NDVI trends for all sites (Fig. 3). Within a time frame of two and half years, the spatially averaged NDVI value increased 0.41, 0.36, 0.36, and 0.38 for Piggery Lake, Two Bridges, Top Narockwell, and McCabes Gap, respectively (Fig. 3); and the increasing trends were all significant (p < 0.001, Fig. 2). As the mean NDVI value over one site might flatten the spatial heterogeneity within a forest patch, and conceal the real system behaviour, we fitted 57 BFAST models for each of the NDVI time series. From the fitted the models, we extracted and compared the increase in NDVI values between the watered and non-watered sites at the abrupt changed point at the 2009/2010 flood. The mean NDVI increase was slightly higher at the water sites but the difference was insignificant (DNDVI/month was 0.29 and 0.27 for watered and non-watered sites, respectively, ANOVA, F(1, 55) ¼ 1.58, p ¼ 0.21) (Fig. 4). In general, the long-term drought conditions in the study area responded well to the Southern Oscillation Index (SOI), which is an ~ o/La Nin ~a indication of the development and intensity of El Nin events in the Pacific Ocean (Fig. 5). In most cases, the timing of ~o onset and intensity of drought corresponded strongly the El Nin episodes indicated by sustained negative SOI. Similarly, floods were ~ a phases indicated by persistent positive correlated with the La Nin SOI (Fig. 5). 3.2. The relationship between NDVI and drought index The NDVI time series at the river red gum forest sites were

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

283

Fig. 2. Fitted seasonal (St), trend (Tt) and remainder (et) components of the site-averaged NDVI time series (Yt) at watered (Two Bridges and Piggery Lake) and non-watered (McCabes Gap and Top Narockwell) river red gum forest sites. Time (dotted vertical lines) with confidence intervals (red), magnitude and direction of abrupt changes, slope of trends (b) and the corresponding significant levels (p) are shown for the modelled long-term trends. Note that the NDVI values were multiplied by 100.

sufficiently modelled with AR(1)/AR(2) and SPEI3 (i.e. 12-month SPEI) (adjusted R2 > 0.85, Table 1). Although we included trend in the initial multivariate autoregression time series model, the stepwise model selection procedure dropped the trend term as it was not significant (p ¼ 0.41, 0.86, 0.72, and 0.50 for Top Narockwell, McCabes Gap, Piggery Lake, and Two Bridges, respectively), suggesting that the revealed trends using BFAST were largely explained by the variation in drought index.

The coefficients for both the SPEI and autoregressive terms were P comparable (i.e. b and A were very close for all models, Table 1). The models for the four forest sites had similar structure; therefore, the similarity in the sum of autoregressive coefficients suggested equivalent time persistence. Furthermore, the estimated coefficient of the SPEI3 was also close (Table 1) indicating the effects of drought on forest canopy conditions were more or less the same for both watered and unwatered sites.

284

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

Fig. 3. The alignment of the 12-month SPEI (brown bars) and modelled trend components of site-averaged NDVI (line) at watered (Two Bridges and Piggery Lake) and non-watered (McCabes Gap and Top Narockwell) river red gum forest sites. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. The abrupt NDVI increase at the change point revealed by BFAST models for watered and non-watered river red gum forests during the transition of returning to wet phase in 2009/2010.

To account for the spatial heterogeneity within a vegetation patch, we fitted 57 autoregressive models for all the NDVI time series. The results showed that there was no significant difference in tree stand condition response in terms of sum of autoregressive coefficients between sites with and without environment watering (ANOVA, F(1, 55) ¼ 2.25, p ¼ 0.14, Fig. 6). The difference in effects of SPEI on NDVI was also insignificant (ANOVA, F(1, 55) ¼ 1.00, p ¼ 0.32, Fig. 7). 4. Discussion The drought of 1997e2009 was the lowest average rainfall period in the southern Murray Darling Basin on record, eclipsing the 1935e1945 drought (Leblanc et al., 2012; also see Fig. 5). The drought was characterised by a large negative anomaly in autumn rainfall, with lesser negative anomalies across winter, spring and

summer (Leblanc et al., 2012; also see Fig. 3). The hydrological drought was further exacerbated by impact of farm dams, ground and surface water extraction intercepting river flow (Potter and Chiew, 2011), and the lowland wetland systems were arguably the most impacted environments over the period (Leblanc et al., 2012). The combined effect of the drying climate and water extraction on hydrological drought has been implicated in the declining condition of floodplain forests across the basin, described at the time as one of the most dramatic examples of floodplain forest dieback in the world (Horner et al., 2009; Cunningham et al., 2009). Cunningham et al. (2009) used NDVI to estimate that 70% of the river red gum forests of the Murray River floodplain were in poor or declining condition. The millennium drought broke abruptly in 2010, as the basin came under the influence of a strong La Nina event (Fig. 5). The calendar year 2010 was the wettest yet recorded in the Murray-Darling Basin (Beard et al., 2011), and was followed by two above-average rainfall years. The two sites receiving managed environmental water, Piggery Swamp and Two Bridges, maintained higher NDVI values during the drought period than the non-watered sites, Top Narockwell and McCabes Gap. These observations confirmed our first hypothesis that artificial flooding enhanced tree-stand condition during the dry phase, and concord with previous studies showing a relationship between drought-phase red gum condition and gradients in water availability (Mac Nally et al., 2011). If an intention of environmental water application over the drought period was to maintain crown condition, the strategy achieved its desired effect. However, the direction (Figs. 2 and 3) and degree of response to wet phase (Fig. 4) is similar in watered and unwatered sites in spite of the history or inundation, suggesting that rainfall rather than inundation is sufficient to explain the recovery response. An important caveat on this conclusion is the possibility that groundwater recharge may extend beyond the inundation zone, with high lateral connectivity of lenses commonly observed across floodplains in the region (Doody et al., 2014). Eucalyptus camaldulensis at the Top Narockwell and McCabes sites may be aided by groundwater recharge from events inundating Piggery Swamp and Two Bridges.

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

285

~ o-Southern Oscillation cycles on Fig. 5. The relationship of SPEI at Lowbidgee (red) and 6-month smoothed SOI (blue) from 1900 to 2013 showing the strong effluence of the El Nin local drought conditions (SOI downloaded from Bureau of Meteorology Australia at http://www.bom.gov.au/).

Table 1 Summary of fitted VAR models for site-averaged NDVI time series. The exogenous variable included in the model is the 12-month SPEI. P Site b А Max lags Adjusted R2 p value Top Narockwell McCabes Gap Piggery Lake Two Bridges

0.32 0.27 0.25 0.34

0.89 0.89 0.90 0.88

1 2 2 2

0.85 0.88 0.89 0.88

<0.001 <0.001 <0.001 <0.001

Fig. 6. The sum of AR coefficients for individual autoregressive models at five sampling sites.

The change from dry to wet in 2009e2010 was unprecedented in magnitude in the climatic record (Fig. 5), and the NDVI response is remarkable in that all red gum communities exhibited similarly high NDVI responses (Fig. 4). Though these declined in ensuing years, they did so in unison, with watered and unwatered sites maintaining similar values. We interpret this as a resetting of condition across the watering gradient in response to the high rainfall and large natural floods over this period. In a

Fig. 7. The estimates of effects of SPEI on NDVI autoregressive models at five sampling sites.

286

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287

contemporaneous study at the same location, Doody et al. (2015) showed similar transpiration rates post-flooding in E. camaldulensis stands in frequently flooded sites (1:2 year), and those subject to prolonged drought (1:5 years). Our study demonstrates that changes to NDVI-based metrics of leaf area and photosynthetic response change over inter-annual and interdecadal cycles are dominated by response to climatic phase (i.e. ENSO) rather than antecedent watering history. NDVI is therefore a problematic indicator of the long-term changes to floodplain vegetation condition induced by environmental water application in a climatically variable system. We could not detect a difference in the pattern of NDVI response to wet phase initiation between river red gum stands receiving planned environmental water flows and those not receiving these flows during the drought phase, or at least these responses were insignificant compared to the over-riding climate signal of the historically unprecedented shifts in rainfall conditions. Casting ecological watering objectives in terms of the proportion of forest in good condition (e.g. MDBA, 2010) may subject success to the vagaries of a variable climate. In relation to our second and third hypotheses, we could not detect any significant difference in time persistence and the effects of drought index, and resilience (as indicated by recovery capacity) between the watered and unwatered forests. Over the entire study period of Feb 2000 to Jul 2013, during which the MDB experienced ~ o and La Nin ~ a episodes, the NDVI anomalies a cycle of strong El Nin at all forest sites displayed similar significant response to the drought index indicated by the comparable estimates of coefficients. The difference in the sum of autoregressive coefficients was also insignificant among forest stands. Furthermore, the magnitude of NDVI increase during the transition of returning to wet phase was comparable. This historic dry was not sufficient to induce wide-spread mortality and periods of drying may be as important to the persistence of these forests as flooding (Sims and Colloff, 2012). Aerobic conditions prevalent during the dry phase promote the release of plant nutrients facilitating flood-induced productivity responses (van Oorschot et al., 1998; Powell et al., 2014a,b) River red gums respond to lower water availability by reducing leaf area, at the individual and stand scales (Cunningham et al., 2007). Horner et al. (2009) followed stand condition and tree mortality across experimental plots with a range of planting densities to show that the millennium drought led to mortality of trees in stands of artificially high stand density, but not a lower stand densities, suggesting competition for water triggered the thinning of the high density stands. Loss of canopy condition, described as the thinning of leaf area, was a more commonly described drought response than mortality or even changes in physiological indicators of performance during the millennium drought (Cunningham et al., 2007). Water potential was not noticeably different between trees in poor and good condition sites along the Murray River (Cunningham et al., 2007), an indication that leaf drop is a healthy drought response, further illustrated in the restoration of canopy condition in the subsequent La Nina phase across several monitoring sites (Colloff et al. in press). The immediate shedding of leaves with the onset of soil moisture after deficit indicates a strong reliance on water in the upper layers, and Doody et al. (2015) attribute this to the production of adventitious roots during the inundation phase. This is not to say that environmental flows are not of primary importance to the maintenance of floodplain wetland vegetation. In a previous study (Wen et al., 2012), we showed that while rainfall dominated NDVI responses of floodplain vegetation in the Macquarie Marshes, the spatial heterogeneity of response corresponded closely to flooding regime. That is, flooding was the key to maintaining the mosaic of different vegetation communities in the

marshes, each with distinct inundation regime requirements. It is this mix of vegetation communities along inundation gradients that is likely to drive biological diversity in the system (Rogers et al., 2012). The monitoring of the effects of flooding regime on wetland resilience might be better focussed on the spatial distribution of wetland plant communities, and their relation to flooding periodicity (Rogers et al., 2013; Saintilan and Imagraben, 2012), than on short-term NDVI responses of canopies. To conclude, we assessed the effects of artificial watering on the long-term dynamics of canopy condition of river red gum forest ~ o and La Nin ~ a cycle using MODIS NDVI stands during a strong El Nin as a surrogate. We found the canopy condition was better during the drought period at the watering site. Despite this, both watered and unwatered sites displayed great similarity in seasonality and trends. Furthermore, we did not find any significant difference in ecosystem stability and resilience between the watered and unwater sites. The highly significant relationship between longterm drought index and NDVI anomaly suggest that climate phase is the main forcing driving canopy condition in semi-arid floodplain forest. Acknowledgement The authors thank three anonymous reviewers for their insightful comments and criticisms. The views expressed in this article are those of the authors, and do not necessarily reflect the official opinion of the NSW OEH. References Alignier, A., Deconchat, M., 2013. Patterns of forest vegetation responses to edge effect as revealed by a continuous approach. Ann. For. Sci. 70 (6), 601e609. Allen, R.J., 1988. El Nino southern oscillation influences in the Australasian region. Prog. Phys. Geogr. 12, 313e348. Asner, G.P., Alencar, A., 2010. Drought impacts on the Amazon forest: the remote sensing perspective. New. Phytol. 187 (3), 569e578. Australian Nature Conservation Agency, 1996. A Directory of Important Wetlands in Australia, second ed. Australian Nature Conservation Agency, Canberra, Australia. Beard, G., Chandler, E., Watkins, A.B., Jones, D.A., 2011. How does the 2010e11 La ~ a compare with past La Nin ~ a events. Bull. Aust. Meteorological Oceanogr. Nin Soc. 24, 17e20. Beguería, S., Vicente-Serrano, S.M., 2012. SPEI: Calculation of the Standardised Precipitation-evapotranspiration Index. R Package Version 1.2. R Foundation for Statistical Computing, Vienna, Austria. ~ a Events BOM, Australian Bureau of Meteorology, July 2012. Record-breaking La Nin ~ a Life Cycle and the Impacts and Significance of the - an Analysis of the La Nin ~ a Events in Australia. The Bureau of Meteorology, 2010e11 and 2011e12 La Nin Melbourne Australia. : www.bom.gov.au/cgi-bin/climate/change/timeseries.cgi. Available at. Breshears, D.D., Cobb, N.S., Rich, P.M., Price, K.P., Allen, C.D., Balice, R.G., Romme, W.H., Floyd, J.H., Belnap, J., Anderson, J.J., Myers, O.B., Meyer, C.W., 2005. Regional vegetation die-off in response to global-change-type drought. Proc. Natl. Acad. Sci. U S A 102 (42), 15144e15148. Bunn, S.E., Arthington, A.H., 2002. Basic principles and ecological consequences of altered flow regimes for aquatic biodiversity. Environ. Manag. 30 (4), 492e507. Colloff, M., 2014. Flooded Forest and Desert Creek: Ecology and History of the River Red Gum. Csiro Publishing, Melbourne, Australia. Colloff, M.J., Baldwin, D.S., 2010. Resilience of floodplain ecosystems in a semi-arid environment. Rangeland J. 32 (3), 305e314. Colloff, M.J., Caley, P., Saintilan, N., Pollino, C.A., Crossman, N.D., 2015. Long-term ecological trends of flow-dependent ecosystems in a major regulated river basin. Mar. Freshw. Res. Published online April 2015 http://dx.doi.org/10.1071/ MF14067. CSIRO, 2008. Water Availability in the Murray Darling Basin. A Report to the Australian Government from the CSIRO Murray-Darling Basin Sustainable Yields Project. CSIRO, Australia, p. 67. Cunningham, S.C., Read, J., Baker, P.J., Mac Nally, R., 2007. Quantitative assessment of stand condition and its relationship to physiological stress in stands of Eucalyptus camaldulensis (Myrtaceae). Aust. J. Bot. 55 (7), 692e699. Cunningham, S.C., Mac Nally, R., Read, J., Baker, P.J., White, M., Thomson, J.R., Griffioen, P., 2009. A robust technique for mapping vegetation condition across a major river system. Ecosystems 12 (2), 207e219. Cunningham, S.C., Griffioen, P., White, M., Mac Nally, R., 2010. Mapping the Condition of River Red Gum (Eucalyptus camaldulensis Dehnh.) and Black Box

L. Wen, N. Saintilan / Journal of Environmental Management 159 (2015) 279e287 (Eucalyptus largiflorens F. Muell.) Stands in the Living Murray Icon Sites. Stand Condition Report 2010. MurrayeDarling Basin Authority, Canberra. fi, S., Dakos, V., Carpenter, S.R., Brock, W.A., Ellison, A.M., Guttal, V., Ives, A.R., Ke Livina, V., Seekell, D.A., van Nes, E.H., Scheffer, M., 2012. Methods for detecting early warnings of critical transitions in time series illustrated using simulated ecological data. PloS One 7 (7), e41010. Doody, T., Benger, S., Pritchard, J., Overton, I., 2014. Ecological response of Eucalyptus camaldulensis (River Red Gum) to extended drought and flooding along the River Murray, South Australia, 1997e2011 and implications for environmental flow management. Mar. Freshw. Res. 65, 1082e1093. Doody, T.M., Colloff, M.J., Davies, M., Koul, V., Benyon, R.G., Nagler, P.L., 2015. Quantifying water requirements of riparian river red gum (Eucalyptus camaldulensis) in the MurrayeDarling Basin, Australia e implications for the management of environmental flows. Ecohydrology. http://dx.doi.org/10.1002/ eco.1598. Gawne, B., Butcher, R., Hale, J., Kingsford, R., Brandis, K., Leo, V., Wheaton, L., Hladyz, S., 2011. A Review of River Ecosystem Condition in the Murray-Darling Basin. Final Report prepared for the Murray-Darling Basin Authority by The Murray-Darling Freshwater Research Centre. MDFRC Publication, p. 289, 01/ 2011, June.  Comín, F.A., Muller, E., 2010. Recent lez, E., Gonza lez-Sanchis, M., Cabezas, A., Gonza changes in the riparian forest of a large regulated Mediterranean river: implications for management. Environ. Manag. 45 (4), 669e681. , Z., Bucha, T., Kono ^pka, M., Luka  Hl asny, T., Barka, I., Sitkova c, M., 2015. MODIS-based vegetation index has sufficient sensitivity to indicate stand-level intra-seasonal climatic stress in oak and beech forests. Ann. For. Sci. 72 (1), 109e125. Horner, G.J., Baker, P.J., Mac Nally, R., Cunningham, S.C., Thomson, J.R., Hamilton, F., 2009. Mortality of developing floodplain forests subjected to a drying climate and water extraction. Glob. Change Biol. 15 (9), 2176e2186. de Keersmaecker, W., Lhermitte, S., Honnay, O., Farifteh, J., Somers, B., Coppin, P., 2014. How to measure ecosystem stability? an evaluation of the reliability of stability metrics based on remote sensing time series across the major global ecosystems. Glob. Change Biol. 20, 2149e2161. Kingsford, R.T., 2000. Ecological impacts of dams, water diversions and river management on floodplain wetlands in Australia. Austral Ecol. 25 (2), 109e127. Leblanc, M., Tweed, S., Van Dijk, A., Timbal, B., 2012. A review of historic and future hydrological changes in the Murray-Darling Basin. Glob. Planet. Change 80, 226e246. Leslie, D.J., 2001. Effect of river management on coloniallyenesting waterbirds in the BarmaheMillewa forest, southeEastern Australia. Regul. Rivers: Res. Manag. 17 (1), 21e36. Liang, L., Schwartz, M.D., 2009. Landscape phenology: an integrative approach to seasonal vegetation dynamics. Landsc. Ecol. 24 (4), 465e472. Mac Nally, R., Cunningham, S.C., Baker, P.J., Horner, G.J., Thomson, J.R., 2011. Dynamics of murrayedarling floodplain forests under multiple stressors: the past, present, and future of an Australian icon. Water Resour. Res. 47 (12). Maher, P., 1990. Bird Survey of the Lachlan/Murrumbidgee Confluence Wetlands. NSW National Parks and Wildlife Service, Hurstville, NSW. McCosker, R., 2008. Yanga Vegetation Mapping: Historical Community Extent and Condition. Technical report to NSW. Department of Environment and Climate Change, Sydney. MDBA, 2010. Guide to the Proposed Basin Plan. MDBA Publication No. 61/10. In: Technical Background Part II, vol. 2. Murray eDarling Basin Authority, Canberra. MDBC, 2005. Survey of River Red Gum and Black Box Health along the River Murray in New South Wales, Victoria and South Australia- 2004. Murray Darling Basin Commission, Canberra, ACT. ~ o in a changing climate: a van Oldenborgh, G.J., Philip, S., Collins, M., 2005. El Nin multi-model study. Ocean Sci. Discuss. 2 (3), 267e298. van Oorschot, M., Hayes, C., van Strien, I., 1998. The influence of soil desiccation on plant production, nutrient uptake and plant nutrient availability in two French floodplain grasslands. Regul. Rivers: Res. Manag. 14 (4), 313e327. Paget, M.J., King, E.A., 2008. MODIS Land Data Sets for the Australian Region. Internal report 004. CSIRO Marine and Atmospheric Research. Pimm, S.L., 1984. The complexity and stability of ecosystems. Nature 307 (5949), 321e326. Potter, N.J., Chiew, F.H.S., 2011. An investigation into changes in climate characteristics causing the recent very low runoff in the southern MurrayeDarling Basin using rainfallerunoff models. Water Resour. Res. 47 (12). Powell, S.J., Jakeman, A., Croke, B., 2014a. Can NDVI response indicate the effective

287

flood extent in macrophyte dominated floodplain wetlands? Ecol. Indic. 45, 486e493. Powell, S.J., Jakeman, A., Croke, B., 2014b. Can NDVI response indicate the effective flood extent in macrophyte dominated floodplain wetlands? Ecol. Indic. 45, 486e493. Power, S., Casey, T., Folland, C., Colman, A., Mehta, V., 1999. Inter-decadal modulation of the impact of ENSO on Australia. Clim. Dyn. 15, 319e324. Puckridge, J.T., Sheldon, F., Walker, K.F., Boulton, A.J., 1998. Flow variability and the ecology of large rivers. Mar. Freshw. Res. 49, 55e72. Ralph, T.J., Hesse, P.P., 2010. Downstream hydrogeomorphic changes along the Macquarie River, southeastern Australia, leading to channel breakdown and floodplain wetlands. Geomorphology 118, 48e64. Robertson, A.I., Bunn, S.E., Boon, P.I., Walker, K.F., 1999. Sources, sinks and transformations of organic carbon in Australian floodplain rivers. Mar. Freshw. Res. 50 (8), 813e829. Rogers, K., Ralph, T. (Eds.), 2010. Floodplain Wetland Biota in the Murray-Darling Basin: Water and Habitat Requirements. CSIRO PUBLISHING. Rogers, K., Ralph, T.J., Saintilan, N., 2012. The use of representative species as surrogates for wetland inundation. Wetlands 32 (2), 249e256. Rogers, K., Saintilan, N., Colloff, M.J., Wen, L., 2013. Application of thresholds of potential concern and limits of acceptable change in the condition assessment of a significant wetland. Environ. Monit. Assess. 185 (10), 8583e8600. Saintilan, N., Imgraben, S., 2012. Principles for the monitoring and evaluation of wetland extent, condition and function in Australia. Environ. Monit. Assess. 184 (1), 595e606. Sims, N.C., Colloff, M.J., 2012. Remote sensing of vegetation responses to flooding of a semi-arid floodplain: implications for monitoring ecological effects of environmental flows. Ecol. Indic. 18, 387e391. Thornton, P.K., Ericksen, P.J., Herrero, M., Challinor, A.J., 2014. Climate variability and vulnerability to climate change: a review. Glob. Change Biol. 20, 3313e3328. Ummenhofer, C.C., England, M.H., McIntosh, P.C., Meyers, G.A., Pook, M.J., Risbey, J.S., Gupta, A.S., Taschetto, A.S., 2009. What causes southeast Australia's worst droughts? Geophys. Res. Lett. 36 (L04706) http://dx.doi.org/10.1029// 2008GL036801. Verbesselt, J., Hyndman, R., Newnham, G., Culvenor, D., 2010a. Detecting trend and seasonal changes in Satellite image time series. Remote Sens. Environ. 114 (1), 106e115. Verbesselt, J., Hyndman, R., Zeileis, A., Culvenor, D., 2010b. Phenological change Detection while accounting for abrupt and Gradual trends in Satellite image time series. Remote Sens. Environ. 114 (12), 2970e2980. pez-Moreno, J.I., 2010. A multiscalar drought Vicente-Serrano, S.M., Beguería, S., Lo index sensitive to global warming: the standardized precipitation evapotranspiration index. J. Clim. 23 (7), 1696e1718. Ward, D.P., Hamilton, S.K., Jardine, T.D., Pettit, N.E., Tews, E.K., Olley, J.M., Bunn, S.E., 2013. Assessing the seasonal dynamics of inundation, turbidity, and aquatic vegetation in the Australian wetedry tropics using optical remote sensing. Ecohydrology 6 (2), 312e323. Wen, L., Saintilan, N., 2014. Linking local ecological outcomes with basin-wide water planning: a case study of Yanga National Park, an important Australian inland forested wetland. Hydrological Sci. J. 59 (3e4), 904e915. Wen, L., Ling, J., Saintilan, N., Rogers, K., 2009. An investigation of the hydrological requirements of River Red Gum (Eucalyptus camaldulensis) Forest, using Classification and Regression Tree modelling. Ecohydrology 2 (2), 143e155. Wen, L., Rogers, K., Saintilan, N., Ling, J., 2011. The influences of climate and hydrology on population dynamics of waterbirds in the lower Murrumbidgee River floodplains in Southeast Australia: Implications for environmental water management. Ecol. Model. 222 (1), 154e163. Wen, L., Yang, X., Saintilan, N., 2012. Local climate determines the NDVI-based primary productivity and flooding creates heterogeneity in semi-arid floodplain ecosystem. Ecol. Model. 242, 116e126. Wen, L., Macdonald, R., Morrison, T., Hameed, T., Saintilan, N., Ling, J., 2013. From hydrodynamic to hydrological modelling: Investigating long-term hydrological regimes of key wetlands in the Macquarie Marshes, a semi-arid lowland floodplain in Australia. J. Hydrology 500, 45e61. €mer, W., Hornik, K., 2003. Testing and dating of structural Zeileis, A., Kleiber, C., Kra changes in practice. Comput. Statistics Data Analysis 44 (1), 109e123. Zhang, X., Friedl, M.A., Schaaf, C.B., Strahler, A.H., Hodges, J.C., Gao, F., Huete, A., 2003. Monitoring vegetation phenology using MODIS. Remote Sens. Environ. 84 (3), 471e475.