Historical photographs revisited: A case study for dating and characterizing recent loss of coral cover on the inshore Great Barrier Reef

Long-term data with high-precision chronology are essential to elucidate past ecological changes on coral reefs beyond the period of modern-day monitoring programs. In 2012 we revisited two inshore reefs within the central Great Barrier Reef, where a series of historical photographs document a loss of hard coral cover between c.1890–1994 AD. Here we use an integrated approach that includes high-precision U-Th dating specifically tailored for determining the age of extremely young corals to provide a robust, objective characterisation of ecological transition. The timing of mortality for most of the dead in situ corals sampled from the historical photograph locations was found to coincide with major flood events in 1990–1991 at Bramston Reef and 1970 and 2008 at Stone Island. Evidence of some recovery was found at Bramston Reef with living coral genera similar to what was described in c.1890 present in 2012. In contrast, very little sign of coral re-establishment was found at Stone Island suggesting delayed recovery. These results provide a valuable reference point for managers to continue monitoring the recovery (or lack thereof) of coral communities at these reefs.


Supplementary Information
Absence of 230 Th ages dating to 1990.4 ± 1.3 event at Stone Island One possible reason for the absence of coral colonies at Stone Island dating to the 1990.4 ± 1.3 event that caused extensive mortality at Bramston Reef, is that there were very few living coral colonies present at that point in time. As discussed in the main text, there is anecdotal evidence to suggest that there was a healthy reef flat at Stone Island sometime during the 1960-1970s 1 . Following the destruction of the coral community in 1918, the time between disturbance and recovery is therefore estimated to be between 40-50 years. If there was a mortality event around 1970.4 ± 9.6 (determined by the 230 Th ages obtained from Stone Island and neighbouring Bramston Reef), then Stone Island may have still been in the very early stages of recovery with low coral cover; subsequently reducing the probability of being able to date a coral colony to this time period.
A second theory is that remnants of an existing population could have been dislodged from the unstable substrate characterised by loose coral rubble during one of the many category 3 or 4 cyclones that took place in the years leading up to the time of the observations made by Wachenfeld 1 (Supplementary Table  S5). On the other hand, the large massive faviid colonies characteristic of Bramston Reef, which is slightly sheltered by Stone Island, remained unaffected by strong waves Third, the disturbances that occurred in 20th Century resulted in drastic modification of the reef flat and a loss of structural complexity, thus hindering coral re-establishment. Evidence of this comes from Hedley 2 , who described the reef as though all coral had 'been planed away by the waves as if some huge razor had shaved off the coral growth down to the low tide level'.

Catchment modification within Edgecumbe Bay
Soon after settlement in 1861, dredging of Port Denison began (c.1880) to maintain accessibility for shipping and was likely to have resulted in the local re-suspension of fine sediments. Other major industries followed, including the Bowen Saltworks (est.1925) and Cokeworks (est.1933), where little is known about what waste products were, and still are, being discharged directly into coastal waters 3 . Other point-sources of pollutants include several large aquaculture farms and the Bowen wastewater treatment facility (est. late 1950s), with the latter only upgraded to secondary treatment in 1989. In 2010, the wastewater facility did not comply with the Department of the Environment's total nitrogen and total phosphorus limits and only 20% of the effluent was being recycled each day. The remaining four megalitres of effluent that was not recycled was being discharged directly into northern Edgecumbe Bay from Dalrymple Point 4 (Fig. 1b).
Grazing, horticulture and sugar cane farming are also major contributors of diffuse sources of pollutants including sediments, nutrients, herbicides and insecticides that enter waterways in runoff during high rainfall events. The effects of these pollutants on corals and coral reef environments are well documented [5][6][7] . Yet despite the various threats to marine environments within this area, water quality data for the region is either limited or results are not available to the public 3,8 . This deficiency in estuarine and offshore monitoring makes it difficult to assess any downstream impacts 8 .
From the upstream data that is available for several rivers and creeks, levels of atrazine and the herbicide diuron exceed trigger values developed by the National Water Quality Management Strategy (NWQMS) 3 . Surveys of the Gregory River in 2000-2002 also revealed total nitrogen, total phosphorus and total suspended solids to exceed trigger values 8,9 . The amount of sediment being delivered to the region from the neighboring Mackay-Whitsunday and Burdekin catchments is estimated to have increased 6.0 and 7.8 fold to 1.5 and 4.7 million tonnes compared to pre-European conditions, respectively, of which ~87% can be attributed to human activities 10 . Other tell-tale signs of the effects of poor water quality come from recent reports of increased marine turtle stranding events attributable to the high loading of heavy metals and toxicants in Edgecumbe Bay 3 .

Methods
The high and low density band couplets were further verified as one year of growth by measuring Sr/Ca, Mg/Ca and U/Ca ratios at approximately monthly resolution in the P. verweyi sample Bram_R8. This sample was chosen for analysis as enough material was available to represent several years of growth. Moreover, the X-ray images revealed clear density banding patterns. Approximately 0.001 g of skeletal material was carefully milled at ~1 mm increments (approximately monthly resolution) along the coral's main growth axis using a hand held drill and diamond drill bits. Powdered material was collected on weighing paper and each drill bit replaced between samples to prevent cross-contamination.
A stock solution of each sample was prepared by dissolving ~0.0002 g of powdered material in 12 ml of 2% HNO3 in a pre-cleaned HDPE tube to achieve a dilution factor of 60,000 times. The solution was then shaken to ensure complete homogenisation and centrifuged at 4,000 rpm for 10 minutes. JCP-1 coral standard and an in-house coral standard were used as external monitors to correct for interference and prepared by dissolving powdered material in 2% HNO3 to achieve a final dilution factor of 60,000 times. An internal standard solution containing known concentrations of Sc, V, 206 Bi, 235 U was prepared by diluting each element to 60 ppb in 2% HNO 3 . The 2% HNO3 solution used to dissolve the samples was used as a blank.
Samples were then measured fully automatically, with the external monitors, internal standard and blank solution introduced separately to an Agilent 7900 inductively coupled plasma mass-spectrometer (ICP-MS) at the Radiogenic Isotope Facility, The University of Queensland. Samples were measured twice to assess reproducibility among samples. Samples were corrected for blank contamination and internal drift using the 2% HNO 3 blank solution and internal standard solution respectively. Elemental concentrations for Sr, Ca, and Mg were calculated using the software MassHunter 2015. Elemental ratios for Sr/Ca, Mg/Ca, U/Ca and Ba/Ca were converted to mmol mol -1 using JCP-1certified values

Results
Sr/Ca ratios display a clear cyclical pattern with values ranging from 8.077 ± 0.011 (1sd) to 9.051 ± 0.054 mmol mol -1 . This is a slightly lower range compared to values reported in massive Porites corals elsewhere in the GBR (ranging ~8.5 to 9.6 mmol mol -1 ) that are routinely used for climate and environmental reconstruction [11][12][13] . The time series represented by the Sr/Ca data spans approximately three complete cycles (trough to trough), with each cycle representing one year of growth. High Sr/Ca ratios were found to correspond to high density bands, while low Sr/Ca ratios corresponded with low density bands. Mg/Ca ratios ranged from 3.6051 ± 0.0058 to 5.467 ± 0.058, also producing three cycles between high and low values. Mg/Ca data also followed a similar pattern to U/Ca ratios which ranged from 0.9999 ± 0.0019 to 1.2 ± 0.01 and also producing three cycles. Reproducibility of Sr/Ca for the inhouse external monitor and JCP-1 was 0.44 (N=9) and 0.49% (N=9), respectively. Reproducibility of Mg/Ca for the in-house external monitor and JCP-1 was 0.51 (N=9) and 0.38% (N=9), respectively, and 1.23 (N=9) and 0.96% (N=9) for U/Ca.

Supplementary Figures
Supplementary Figure S1 | Historical photographs of Bramston Reef. Photographs taken by William Saville-Kent 14 in c.1890. While a geographical landmark is absent in these photographs, comparisons between the type of coral communities existing in the area can still be made with modern photographs to assess any changes in the presence or absence of coral genera.  3.0 2.5 ± 1.5 -4661.5 ± 29 Ratios in parentheses are activity ratios calculated from atomic ratios using decay constants of Cheng et al. 15 . All values have been corrected for laboratory procedural blanks. All errors reported as 2σ. Uncorrected 230 Th age (a) was calculated using Isoplot/EX 3.0 program 16  Th ages corrected using a model two-component correction value based on the equation from Clark et al. 17 : Th where 232 Thdead is the measured 232 Th value (ppb) in the non-living coral sample. 232 Thlive is the mean measured 232 Th value (ppb) and 230 Th/ 232 Thlive represents or approximates the isotopic composition of the hydrogenous component in the dead coral skeleton with an atomic value of 5.76 × 10 -6 ± 20% (which corresponds to an activity value of 1.066 ± 20%) based on live Porites corals collected from the GBR 18 . 230 Th/ 232 Thsed is the detrital component represented by a mean atomic value of 4.38 × 10 -6 ± 20% (which corresponds to an activity value of 0.81 ± 20%) from isochron derived initial 230 Th/ 232 Th values obtained from two dead faviid coral skeletons. § For samples where no X-rays were available or where annual density banding was ambiguous, the number of potential annual growth bands above the sampling location was calculated by dividing the sampling distance from the top of the colony (cm) by: i. An average linear extension rate of 7.3 ± 0.5 mm yr -1 (1sd) [based on the length of annual density bands (N=8) observed in X-rays taken of two dead G. aspera colonies collected in this study] for G. aspera colonies ii. An average linear extension rate of 8.8 ± 2.3 mm yr -1 (1sd) [based on the length of annual density bands (N=37) observed in X-rays taken of five dead P. verweyi colonies collected in this study] for P. verweyi colonies.
iv. An average linear extension rate of ~12 mm yr-1 for massive Porites spp. colonies 19 v. An average linear extension rate of 7.2 mm yr -1 (range 6.6-8.9 mm yr -1 ) for C. seralia 20 (Average linear extension rate for M. turgescens is unknown. The large 230 Th age error for the two M. turgescens colonies sampled is likely to encompass any age uncertainty associated with sampling location). Each uncertain layer was assigned a ± 0.5 year uncertainty similar to Rasmussen et al. 21 and Clark et al. 18 . ^ denotes samples where annual density banding above sampling location was clear. ¶ Surface age = Corrected 230 Th growth band age + number of growth bands above sampling location. Surface age error calculated using addition rule for error propagation: ∆ ∆ ∆ where Δa = surface age error (years), Δb = 2σ 230 Th age error (years), Δc = uncertainty associated with annual density band counting (years)