Climate change rapidly warms and acidifies Australian estuaries

The study addresses how diverse estuary types respond to global climate change, a question that remains speculative due to high spatial and morphological variability across estuaries. Estuaries provide critical ecosystem services (nutrient cycling, carbon storage, fisheries and aquaculture support, and biodiversity). Eastern Australia hosts a wide range of shallow estuary morphologies (creeks, rivers, lagoons, lakes, and back dune lagoons) influenced by semi-diurnal tides and variable entrance conditions, with many systems periodically closed due to low rainfall and coastal sand transport. Existing remote sensing and large-scale oceanic/atmospheric models inadequately capture changes in shallow, morphologically complex estuaries. The authors present a 12-year monitoring programme across 166 estuaries along >1100 km of Australia’s east coast to quantify trends in summer water temperature, pH, and salinity and to identify drivers of change using random forest models. The overarching purpose is to determine the rates and drivers of estuarine warming, acidification, and salinity change, and to evaluate how morphology modulates these responses.

Previous research has documented warming in large estuaries and coastal regions in North America and Europe (e.g., Hudson River, Chesapeake Bay, Woods Hole, Narrow River, Helgoland Roads, and the Mediterranean). However, most existing studies focus on single or few systems, limiting generalizability to diverse estuary types and broader regions. Estuaries’ ecological and economic roles (nursery habitats, fisheries, nutrient processing) are well established, and morphological classifications for southeastern Australian estuaries emphasize the importance of entrance condition, retention, and flushing. Urbanization and catchment alteration are known to elevate water temperatures via reduced shading and heated runoff, and long-term pH trends in some estuaries (e.g., Neuse River) highlight influences of atmospheric CO₂, catchment conditions, and primary production. Yet, comprehensive, multi-estuary assessments across morphology gradients and large spatial scales have been lacking, particularly in regions with many intermittently closed systems like eastern Australia and other dry temperate zones (e.g., South Africa, Mediterranean).

Design: A synthesis of environmental monitoring data from 166 New South Wales (NSW) estuaries spanning 28°S to 37.4°S (~1100 km), covering all major estuary types (47 rivers, 43 lagoons, 28 lakes, 25 creeks, 23 back-dune lagoons). Sampling occurred every austral summer (November–March) from 2007–2008 to 2018–2019. Each summer, water parameters were measured six times at each of 2–3 central basin sites per estuary at 0.5 m depth using a calibrated YSI EXO2 sonde (temperature ±0.001 °C, pH ±0.01, salinity ±0.01 PSU). Samples were logged every second for 3–5 minutes; mean values were used. Sampling was stratified by estuary type and catchment disturbance, with regional rotations (North, Central, South) and sentinel sites sampled annually. Measurements encompassed multiple interannual climate phases (El Niño, La Niña, positive/negative Indian Ocean Dipole). Estuary Typology: Functional classification based on retention factor (ratio of estuary volume to runoff from a 10% annual inflow event) and flushing time (water turnover; tidal prism for open entrances; closure duration approximation for periodically closed systems). NSW estuaries were grouped into meta-types and types: tide-dominated rivers (drowned river valleys/barrier rivers), wave-dominated open rivers, wave-dominated intermittent lakes and lagoons, creeks, and back dune lagoons. Predictor Variables: Time (days elapsed since programme start; month), geomorphology (retention factor; latitude; catchment size; average depth; total flushing time; seagrass cover), human disturbance (percent catchment cleared; percent urbanized; proportional increase in nitrogen load since pre-1770). Estuary volumes were obtained from bathymetry (57 estuaries) or estimated via surface-area-to-volume regressions by type (r² 0.911–0.996). Surface areas were digitized; average depth computed as volume/area; seagrass areas compiled from mapping datasets. Catchment land use derived from ALUM-based mapping (1999–2007 imagery), aggregated for modeling nutrient/sediment export. Nitrogen load increases estimated via a catchment export model comparing pre-1770 native forest to current land use. Analyses: Simple linear models quantified trends in temperature, pH, and salinity over time for all estuaries and by type; model fit assessed via residual diagnostics. Random forest (RF) models (randomForest package) were built separately for temperature, pH, and salinity using relevant predictors (Table 1), with 1000 trees, bootstrap samples (~66%), and tuned mtry (temperature and salinity mtry=9; pH mtry=8). Validation used 20% withhold 10-fold cross-validation (Caret) to calculate RMSE, R², MAE. Additional randomization tests (1000 permutations) generated null distributions of % variance explained to assess model significance. Variable importance assessed via % increase in MSE upon permutation. Partial dependence plots were generated (Plotmo, PDP).

• Overall trends (summer data): Estuaries warmed by 2.16 °C over 12 years (0.2 °C year⁻¹; overall slope 0.00052 per day; Table 2 “ALL” ≈0.192 °C year⁻¹). pH decreased by 0.51 units over 6 years (−0.0978 pH units year⁻¹). Salinity decreased slightly overall (−0.97 PSU over the full period; −0.0861 PSU year⁻¹). - By estuary type (Table 2): Temperature change (°C year⁻¹): Lagoon 0.325 (total +3.65 °C), River 0.248 (+2.79 °C), Back Dune Lagoon (BDL) 0.117 (+1.31 °C), Lake 0.0954 (+1.07 °C), Creek not significant (NS). pH change (units year⁻¹ over 6 years): Creek −0.101 (−0.53 total), Lagoon −0.0888 (−0.46), River −0.0612 (−0.32), BDL −0.0607 (−0.32), Lake −0.0534 (−0.28), ALL −0.0978 (−0.51). Salinity change (PSU year⁻¹): Creek −0.6 (−6.74 total), Lagoon −0.479 (−5.38), River +0.238 (+2.68), Lake NS, BDL NS, ALL −0.0861 (−0.97). - Drivers and morphology effects (RF and partial dependence): Temperature increased over time; strongest predictors: month (January–February hottest), days since sampling began, latitude. Morphological controls: shallower average depth, longer flushing time, and longer retention associated with higher temperatures; effects strengthened in latter half of the study. pH declined over time; strongest predictors: month, latitude (greater declines at higher latitudes), and days since sampling. Biological and catchment factors: higher seagrass cover and increased nitrogen load were associated with higher pH (less acidification), indicating photosynthetic buffering and linkages with pelagic production; longer flushing/retention also associated with higher pH. Salinity showed high variability; key predictors: days elapsed, month, retention. Type-specific patterns: creeks and lagoons freshened (due to freshwater accumulation during closed entrances), rivers became more saline (greater oceanic intrusion under reduced freshwater inflow). Salinity tended to increase with latitude (more saline in southern systems), consistent with greater summer rainfall in northern regions. - Model performance (Table 3): Temperature RF: 82.59% variance explained (OOB), R² 0.794, RMSE 1.223–1.344 °C (OOB/CV), MAE 0.934, p<0.001. pH RF: 81.02% variance explained, R² 0.793, RMSE 0.238–0.251, MAE 0.164, p<0.001. Salinity RF: 85.65% variance explained, R² 0.839, RMSE 4.345–4.616 PSU, MAE 2.95, p<0.001. - Comparative context: Estuary warming rates substantially exceed concurrent regional air (+1.5 °C decade relative to 1961–1990 average) and sea surface temperature increases (~+1 °C) and exceed global model projections for air/ocean, indicating rapid amplification in shallow, restricted systems. pH declines (~0.098 units year⁻¹) exceed open-ocean projections for 2100 under business-as-usual (~0.00625 units year⁻¹).

The findings demonstrate that climate change is rapidly altering estuarine water quality across a broad suite of morphologies, with warming, acidification, and freshening occurring at rates greater than those predicted by large-scale oceanic or atmospheric models. Morphological and hydrological traits (shallow depth, long flushing and retention times, restricted entrances) amplify warming and acidification, explaining why lagoons and rivers warm fastest and why systems with greater seagrass cover acidify less. These results validate the hypothesis that responses depend on estuary-specific attributes and regional climate. Reduced rainfall and streamflow in eastern Australia over the last decade increase retention times and decrease average depths, enhancing warming; concurrently, intermittently closed systems freshen due to trapped runoff while open rivers experience saline intrusion. Interannual climate drivers (SOI/ENSO, IOD) modulate weather but are superimposed on a clear warming trend. The ecological implications are significant: accelerated warming and acidification can increase energetic demands and reduce performance across taxa; combined stressors (temperature, pH, salinity) may narrow tolerance ranges, affecting seagrasses, invertebrates, and fish. Potential poleward range shifts and tropicalisation are likely for open estuaries, while periodically closed estuaries may freshen and experience algal blooms and hypoxia under low-flow conditions, with consequences for biodiversity and fisheries. Seagrass acts as a partial buffer against acidification, but global seagrass loss threatens this mitigation. The study underscores the need to incorporate estuary morphology, hydrology, and biological attributes into regional models to more accurately forecast estuarine change and inform management. The patterns observed are relevant to other dry temperate regions (e.g., South Africa, Mediterranean, parts of Western Australia) with similar estuary types and climatic trends.

Estuaries along >1100 km of southeast Australia are rapidly warming (~0.2 °C year⁻¹), acidifying (0.098 pH units year⁻¹ over 6 years), and slightly freshening (−0.086 PSU year⁻¹), with the fastest changes in shallow, restricted systems (lagoons, rivers) that have long flushing/retention times and limited ocean exchange. These rates exceed projections from large-scale oceanic/atmospheric models, showing that estuary-specific morphology and hydrology drive amplified responses. The study provides a continental-scale, multi-estuary assessment and identifies key drivers (time, depth, flushing/retention, latitude, seagrass cover, catchment disturbance), offering a framework to improve regional climate impact models and guide mitigation. Future research directions include: integrating estuary morphology and biological components (e.g., macrophyte dynamics) into coupled climate–hydrodynamic–biogeochemical models; expanding monitoring to year-round and additional regions to capture seasonal and geographic variability; resolving mechanistic links between catchment management (clearing, urbanization, nutrient loads) and estuarine pH/temperature; and assessing ecosystem thresholds and species-specific vulnerabilities to combined stressors to inform adaptation for fisheries and aquaculture.

• Temporal coverage: pH was measured only for 6 of the 12 monitoring years; temperature and salinity cover 12 years but only in summer (November–March), potentially missing seasonal dynamics. - Sampling design: Creeks were under-represented and are highly variable, reducing power to detect trends; sampling times within days and across summer were haphazard, introducing diel/seasonal variability. - Spatial scope: Study limited to NSW estuaries (temperate–subtropical); generalization to other regions assumes morphological/climatic similarity. Bays were not sampled; barrier rivers and drowned river valleys were grouped. - Measurement depth: Single-depth (0.5 m) measurements assume vertical mixing; although justified for shallow systems, stratification events may be missed. - Derived variables: Many geomorphological and catchment inputs (volumes for most estuaries, nutrient loads, land-use classes) were modeled or estimated (e.g., area–volume regressions), introducing uncertainty. Land-use mapping dates to 1999–2007; subsequent changes may not be fully captured. - Salinity dynamics beyond study window: Some salinity increases in closed estuaries observed in 2019–2020 were not included. - Statistical modeling: Simple linear trends may not capture nonlinearities; RF models, while robust, infer associations rather than causation and may be influenced by collinearity among environmental drivers.