Here, in agreement with results from a study based on the Great Barrier Reef29 and results from our previous study18, the Ba/Ca average values in each coral record showed an expected gradient coherent with the distance from the main input of Ba in the coastal environment: the Baram River mouth. However, the differences in average Ba/Ca values between adjacent colonies (Eve and Anemone, Anemone and Siwa) were not coherent with the distance from the river mouth to Anemone compared to the distance to Eve. The same can be said for Siwa when compared with Anemone or Eve. This difference points towards non-linear dissolution of Ba in the coastal waters, which is highly dependent on the mixing of water masses with different salinities and Ba concentrations30, as well as other processes such as adsorption/desorption to/from suspended particles31, or precipitation32. As such, the Ba concentration in coral skeletons, and thus, in surrounding seawater cannot be accurately predicted based on distance from the input source alone without extensive hydrological modelling, even if monthly riverine Ba records were available.
Similar to results from the initial Ba/Ca study18, only one of our records was significantly correlated with the precipitation record. This may be somewhat surprising given that the available uninterrupted periods of local rain gauge data from the Miri airport station were located less than 13 km away from Eve’s site. In contrast, a study on the Great Barrier Reef (GBR) that used Ba/Ca records to track terrestrial discharge and drought-breaking floods found significant correlations with precipitation33. We suspect the lack of correlation in Miri is due to the close proximity to the river mouth (unlike the GBR study where cores were several tens of kilometres from the nearest rivers), as well as direct exposure to the river plume during the winter monsoon season. Indeed, some studies with catchments similar to that of the Baram River’s have also found that precipitation did not scale linearly with river discharge34,35, thereby supporting our findings. This disconnect between precipitation and river discharge poses a significant challenge when attempting to track precipitation using coral Ba/Ca records, even if they are considered ideal proxies for river discharge.
Conversely, Anemone, Eve, and C1 showed a strong capacity to record river discharge on annual scales, with stronger results closer to the river mouth at Eve’s Garden reef, a similar result was found in a study from the GBR that included several records from coral colonies located at different distances from the river mouth36. Such a gradient does, however, contrast with a previous study by Lewis et al showing the opposite trend in the same reef system. However, in this study, Lewis et al.29 only involved annual Ba/Ca peak maxima and coral colonies were influenced by river runoff to the north and south of the reefs, making comparison with our results difficult. Here, the use of a composite, annual data record improved the signal-to-noise ratio by displaying the best fit with river discharge out of the all records. This demonstrates that, although Anemone had weaker (and sometimes non-significant) correlations than Eve, it still has value for tracking sediment in river discharge when combined with other proxy records, despite the increased distance29.
Using monsoon seasonal records and thereby taking into account the impact of seasonally changing currents on the direction of the river plume, we confirmed that our coral Ba/Ca records trace sediment in river discharge best during the winter monsoon. We expected significantly stronger correlations during the winter monsoon as river waters are pushed towards the coral colonies by north-easterly winds on a more regular basis. The increased correlation strengths using winter monsoon records for both Anemone and C1, even when compared to three-year-averaged annual data (except Eve, 0.02 difference in correlation strength), as well as the stark contrast with correlations using the summer monsoon records, stresses the importance of looking beyond correlation strength (and significance) of non-transformed monthly or annual data37,38. Interestingly, despite not being the closest to the river mouth, the Ba/Ca record from the Anemone colony showed higher correlations during the winter months than Eve or C1 during the same period, perhaps pointing towards an optimal distance to best record sediments in river discharge, as suggested by Lewis et al.29, due to the way freshwater and ocean water mix as well as sedimentation settling rate. Based on this data alone, we have no way of confirming if the 36.4 km separating Anemone from the river mouth is the optimal distance. This would require collecting additional coral records from colonies at sites other than Anemone, Eve, and Siwa to establish such a distance.
Previous studies have used Ba/Ca records to track changes in land use22,39, deforestation39,40, flood events33,38 and in population density and settlements22 in Madagascar, Indonesia, and the GBR. Similarly, we argue here that our Ba/Ca time series are recording an increase of river-bound sediments attributed to land use changes related to the onset of deforestation starting before the first half of the 20th century. Indeed, all three Ba/Ca time series showed trends of increasing average values throughout their records and when shortened to match the record length of Anemone. However, the scale of their trends was not similar, similarly to coral Ba/Ca records from the Gulf of Chiriquí38. Contrasting with the trend analysis performed on the shorter version of Anemone and Eve records showing slopes of 0.067 and 0.072, respectively18, results here showed weaker trends, indicating a nonlinear increase of Ba/Ca throughout the records with a stronger increase towards more recent periods41. The difference in trend strengths between Anemone, Eve, and Siwa noted here reinforces our previous study’s hypothesis that either rates of change do not have any relationship with distance from the river mouth, or there are additional factors at play, related to local hydrodynamics that cannot be accounted for in the scope of this study18. We also observed an increasing lag in the position of the changepoint in time with increasing distance from the main source of Ba input (the Baram River). This pattern could potentially be due to a threshold effect related to local hydrodynamical conditions that prevented the riverine Ba signal from reaching Anemone and Siwa in amounts proportional to the distance from the river mouth41. Such a threshold could mean that a 61% increase in riverine Ba input was recorded by the Eve core but either was not recorded by the Anemone coral or was recorded as a lesser increase than the 28.5% greater distance from the river mouth would suggest41. Our results from the three coral colonies are consistent with the second option, as records further away from the river mouth need a stronger riverine Ba increase in the river discharge output to record a significant increase in their Ba/Ca signature. Further, the percentage of increased average value associated with the changepoint increased with distance from the river mouth (Fig. 3). We argue here that the mean value difference is due (at least in part) to the substantially lower baseline values found in Anemone and Siwa compared to Eve.
As such, whether the recorded increase in Ba/Ca is due to increased concentrations of sediment in river runoff or increased river discharge, or both, is currently unknown. However, for the period of overlap with instrumental data starting in 1989, we have indications for increased river runoff as a major driver of increased Ba delivery over the recent decades18.
Logging data from the Miri region in Sarawak from 1955 and 1961 shows an increase consistent with the findings of our geochemical records27. Although such data can hardly be transformed to approximate the deforested area without making several significant assumptions, a 100% increase in logged timber is indicative of increased deforestation in the area, even over the short time scale between the two data points (6 years). Such a short timescale does not allow for timber wood regrowth, which has previously been observed to take between 12 to 15 years42. Comparably, estimated land use data across the Southeast Asia region showed a decrease in forested areas corresponding to increased deforestation and land conversion starting as early as 1850, and accelerating in the 1900s and 1950s28. We argue that just before 1950, while deforestation was ramping up across the Maritime Continent28,43, the increased sediment load resulting from deforestation in the catchment reached the threshold required to prompt a change in the Ba/Ca ratio recorded by coral colonies. The inferences we draw from geochemical data align with historical records indicating that approximately one-third of Southeast Asia's forested areas had been cleared prior to World War II, and that this trend continued with only a gradual decline in forest cover post-195044. Even though deforestation is mainly driven by commercial wood extraction, cultivation, livestock grazing, or infrastructure development in most tropical regions globally, Southeast Asia’s deforestation was primarily driven by timber logging as well as swidden cultivation and permanent agriculture during the 20th century44. By the mid-1980s, over one-third of South-East Asia’s forests had been cleared, contrasting with South America where rapid increases in deforestation occurred earlier in the 1900s. Forest clearing in Southeast Asia does, however, show similar trends and beginnings with forests in Africa where the extraction of exotic timber was started by European settlers around 160044. Although the scale of deforestation was not comparable to that undertaken in the 20th century, these results indicate that by the turn of century the forested area had already been significantly reduced, contradicting previous notions of the extent of forested areas45.
It is imperative that we identify the onset times of deforestation to better understand rates of decline in forested areas worldwide, but particularly in tropical areas where deforestation is the most intense1,12,46. These efforts are essential to understand the myriad of effects that deforestation has on local and regional land and coastal ecosystems aided by a quantified changepoint corresponding to an onset date of massive deforestation. As such, results like ours should encourage further research on the impact of deforestation and land use in Malaysian Borneo had on coral reef organisms’ health to better develop science-based policy guidelines that would reduce deforestation and protect remaining ecosystems. Curbing rates of deforestation and encouraging afforestation would also reduce flood risks and enhance water quality in the coastal environment following reductions in sediment loads entering river waters47,48. Finally, the methods employed here should be applied to other catchments where historical instrumental data is lacking to understand the impact of anthropogenic development and uncover deforestation onsets in other tropical coastal regions.