Climate data for water modelling
Rainfall and evapotranspiration are the 2 main climate factors that affect NSW water resources. We use hydrological modelling to explore how these factors affect water availability in NSW. Some models also consider temperature, which has an important influence on water availability in the cooler regions of NSW that experience regular winter snowfall.
Along with these climate factors, hydrological modelling considers other factors such as the physical properties of river channels, the effects of river regulation and management rules, and water demand and use.
To provide meaningful outputs for water planning and decision-making, modelling needs to capture both natural variability and the effects of climate change on NSW weather patterns. This helps us to better understand how water availability may change across NSW.
There are 2 data sources needed to do this:
- The paleo-stochastic baseline – an extended dataset that captures naturally occurring wet and dry periods. It helps describe natural climate variability, but because it’s based on past conditions, it doesn’t represent the climate we experience today or expect to experience in the future.
- NARCliM2.0 climate projections – modelled data that estimate how rainfall, temperature and other climate variables may change in the future (under different future greenhouse gas emissions scenarios). These projections show climate trends but don’t capture the full range of natural interannual to multiple decade variability.
This page outlines how these 2 data sources support climate risk management in line with best-practice approaches, and how they can be used in similar contexts.
Climate change and climate variability
Climate change refers to long-term changes in climate patterns over decades or longer and is primarily driven by human activities. Climate change is already affecting NSW. Projections show further increases in temperature and potential evaporation and changing rainfall patterns across NSW. Although the direction and scale of average changes to rainfall vary considerably, most climate projections for NSW reflect a future where it's very likely (90–100% probability) that dry periods are drier and more frequent, and wet periods are more intense when they occur.
Climate variability refers to the natural fluctuations in climate that occur over time, such as differences in rainfall and temperature from season to season, year to year, or over several years. For NSW, the main climate drivers that affect climate variability are:
- east coast lows
- El Niño / La Niña
- Southern Annular Mode
- Indian Ocean Dipole (IOD)
- the Interdecadal Pacific Oscillation (IPO).
These are a natural part of the climate system and have always shaped how much water flows into rivers and storages, regardless of human activities.
Climate variability is measured as the changes from an average value. For example, how much more or less rainfall occurs, compared to the average (over hours, days, months, years or decades). Climate change is measured as a long-term shift in average conditions or variability (see Figure 1 below).
Recommended data sources
The paleo-stochastic baseline is the recommended dataset to use as a baseline for climate variability because it reveals a wider range of natural variability than the observed past records. Using this baseline for analysis and planning means a broader range of climatic conditions can be considered, to improve resilience and decision making.
The paleo-stochastic dataset is endorsed as a Common Planning Assumption and offers a consistent foundation for testing water systems’ performance under conditions that are rare but still plausible. Paleo-stochastic baseline data are available for different regions across NSW. Find the data on the NSW Government SEED portal.
Climate projections from the NSW and Australian Regional Climate Modelling (NARCliM) program are the recommended data to use for understanding future climate change. NARCliM2.0 is the official New South Wales climate projection dataset and is endorsed as a Common Planning Assumption for government and infrastructure planning. Find NARCliM projections on the NSW Open Data SEED Portal and the AdaptNSW website.
How the paleo-stochastic dataset captures climate variability
The paleo-stochastic baseline integrates detailed climate records from instrumental measurements with long-term climate patterns informed by paleoclimate data. This creates a dataset representing 10,000 years of possible daily rainfall and evapotranspiration patterns in river valleys across NSW.
This extended dataset captures a broader range of climate variability than historical records alone. It also provides insights into the frequency, length and distribution of wet and dry periods that we know the climate system is capable of (see Figure 2 below).
The paleo-stochastic baseline shows that NSW has experienced droughts and wet periods that are far more extreme than those captured in instrumental records. Although rare, events of this scale could have major impacts on our rivers and water availability.
The baseline aims to generate many plausible sequences of natural climate variability to use in hydrological models for stress-testing, risk assessment and sensitivity analysis.
We analysed data recorded from climate gauges to identify recent daily and broader-scale climate patterns. This showed us:
- daily, seasonal and yearly variations in temperature, rainfall and evapotranspiration for the past 130 years
- the broader climate patterns caused by the main climate drivers that affect NSW – El Niño /, La Niña, IPO, Southern Annular Mode, Indian Ocean Dipole and east coast lows, along with tropical storms in northern NSW.
This analysis provides insight into the variability of the climatic system. But 130 years is a very short period in the long history of our climate. Instrumental records alone are not enough to completely understand the frequency and severity of extreme events, especially multiyear droughts.
Paleoclimate data refers to natural records of climate variability found in sources such as:
- tree rings
- cave deposits (speleothems)
- corals.
The paleo-stochastic baseline uses records from tree rings and corals, accurately dated to annual and seasonal scales. The patterns in these data are used to model the IPO periods. We know enough about the relationship between our climate and the patterns in these sources to reconstruct a record of what our climate was like before instrumental records began. This extends our climate record to more than 500 years.
Stochastic modelling is used to generate synthetic climate sequences that reflect the natural variability shown in the instrumental and paleoclimate datasets. It draws on the statistical properties of observed rainfall and evapotranspiration and introduces controlled randomness to create realistic daily and seasonal patterns. This approach also captures long term climate variability, including the influence of large scale cycles such as the IPO, allowing exploration of a wider range of possible climate conditions than the historical record alone can show.
The paleo-stochastic baseline is suitable for water modelling across NSW systems and sectors. It provides a consistent foundation for testing how water systems would have performed under past climatic conditions. To understand how it may perform under future climate conditions we need to combine this data with climate change data.
Using the paleo-stochastic baseline with NARCliM climate projections
The paleo-stochastic baseline improves our understanding of past extremes but does not account for future climate change. To assess future risks, it needs to be used alongside climate projections. NARCliM2.0 is the official NSW climate projection dataset. It shows how conditions such as rainfall, temperature and evapotranspiration may shift over coming decades.
To apply climate projections, the climate variables in the paleo-stochastic baseline must be scaled according to the projected changes. This approach keeps the realistic daily and seasonal patterns of the baseline, while reflecting future changes. The resulting datasets represent plausible future climates, designed to test how water systems perform across a much wider range of conditions.
NARCliM2.0 provides multiple climate scenarios based on different combinations of parameters. For water planning, some of these parameters are set based on water modelling practices, and others are chosen based on the aim of your planning.
Climate variables
NARCliM2.0 provides projections for 15 core variables and approximately 150 variables in total. The variables you choose will depend on your project. For water modelling, required variables include rainfall, temperature and evapotranspiration.
Spatial resolution
NARCliM2.0 provides a spatial resolution of 4km across the NARCliM domain, which covers south-eastern Australia. This fine resolution allows rain-generating processes to be modelled with greater accuracy compared with previous generations of NARCliM that provided a 10km resolution across south-eastern Australia.
Time horizons
The time horizon refers to how long a decision, policy or asset needs to remain effective. This often matches the life of infrastructure or the review cycle of a planning document. Sometimes, a mid point or future trajectory horizon is also included, to see how outcomes might change over time. NARCliM2.0 provides continuous climate projections to 2099.
Emissions pathways
Emissions pathways represent different possible futures for greenhouse gas emissions, global development and the amount of warming the climate may experience. NARCliM2.0 covers 3 different emissions pathways:
- SSP1-2.6 – a low-emissions pathway
- SSP2-4.5 – a medium-emissions pathway
- SSP3-7.0 – a high-emissions pathway.
Select a pathway that suits your planning purpose and your tolerance level for future climate risks. By considering multiple emissions pathways it helps:
- understand how sensitive a planning decision is to more or less extreme climate change
- identify the decisions that work under multiple scenarios, so we can be prepared even if the future turns out differently than expected.
Climate model ensemble
NARCliM2.0 generates NSW-specific climate projections by combining global climate models (GCMs) with regional climate models (RCMs). GCMs capture global climate patterns and RCMs are used to downscale the GCMs to the NSW context. This combination of models is referred to as an ensemble.
Each model in the ensemble is built differently and makes slightly different assumptions. These differences mean each model produces a slightly different picture of how the climate may change. Looking at the whole ensemble helps capture a broader range of plausible future climates and gives a better understanding of the uncertainty in climate modelling.
How the NARCliM ensemble is used depends on the purpose of your planning and the level of risk involved. Different contexts may require different ways of working with the ensemble, from exploring the full range of model outcomes to focusing on a smaller subset that represents particular conditions of interest.
Choosing how to use the ensemble involves thinking about factors such as the:
- lifespan of an asset or decision
- consequences of underestimating risk
- degree of uncertainty that can be tolerated
- resourcing available to support the analysis.
Scaling is used to adjust the paleo-stochastic baseline according to your chosen climate projections. Scaling preserves the natural variability and sequences of the baseline, which is essential for understanding system behaviour and stress-testing water systems.
Scaling ensures that:
- projected climate datasets retain realistic relationships between climate variables, which is needed to model water availability, storage performance and reliability under changing conditions
- there is a transparent and repeatable way to apply projections, while reducing the risk of introducing unrealistic patterns or losing important climate signals.
Common scaling techniques include monthly change factor scaling and quantile scaling, both of which maintain realistic daily and seasonal behaviour.
Guidance on these and other methods is available from the Climate Change in Australia website.
An example of a scaling applications can be found in NARCliM projections and stochastic simulations – Southern Basin.
A note on reference periods
Climate projections give us information on what the future climate might be like. To understand how this future picture differs from what we experience now, we need to compare the projections against a baseline. We do this by using a defined block of time called a reference period. In hydrological modelling, a longer reference period of 30 years is commonly used to:
- capture the full range of natural variability
- reduce the risk of the baseline being skewed by an unusually wet or dry phase
- better reflect long term average conditions.
It is important to be transparent about which reference period is used, because different periods can highlight different aspects of the climate system.
The paleo-stochastic baseline does not forecast future climate. Rather, it represents many possible sequences of natural variability. When the baseline is combined with climate projections, it helps explore how future conditions could affect rivers and storages.
The baseline is intended for long-term planning and risk assessments. Typical applications include:
- using extended input sequences in hydrological models to assess storage behaviour, system reliability and, environmental outcomes under a wide range of natural variability
- stress-testing management rules and infrastructure options, including sensitivity to prolonged dry or wet clusters
- estimating frequency, duration and sequencing metrics for multiyear droughts and extended wet periods beyond the instrumental record.
Case study: Using paleo-stochastic data to understand groundwater and climate change
Groundwater is a critical resource for communities, agriculture and ecosystems, especially during extended dry periods. While groundwater systems respond more slowly to climate drivers than surface water, they are still vulnerable to long-term changes in rainfall, temperature and evapotranspiration.
A CSIRO study assessed how a projected drying climate could affect groundwater recharge across NSW. Using a 10,000-year paleo-stochastic baseline scaled with NARCliM1.0 projections to simulate streamflow, the study found that:
- diffuse recharge is expected to decline across most of NSW, with an average reduction of 14% and some areas seeing decreases of more than 50%
- localised recharge from losing streams and overbank flooding may also decline significantly, with reductions in some areas exceeding 90%
- groundwater sources reliant on localised recharge, such as inland alluvium systems, are particularly vulnerable and require further investigation.
The use of a paleo-stochastic baseline highlighted the need to improve recharge estimation methods and groundwater models to better support planning under future climate scenarios. Understanding these risks is essential for protecting water security and environmental values.
The paleo-stochastic baseline has important limitations to consider when using it for water planning or risk assessment purposes.
Data are within accepted bias tolerance levels, but biases will exist
All data are within accepted bias tolerance levels (generally ±10% of mean). However, because of the size and complexity of the baseline dataset, it's not possible to ensure unbiased data at all sites, and across all timescales. Check that the data suits your purpose.
It shows baseline climate variability with no climate change
The dataset is considered a baseline representation of natural climate variability and assumes no influence from human-driven climate change. It reflects observed climate conditions up to 2018. For future assessments, this dataset should be used in conjunction with climate projections and appropriate scaling methods to account for expected anthropogenic impacts.
Consider uncertainty – especially for events with long return periods
Using paleoclimate data to model the influence of the Interdecadal Pacific Oscillation (IPO) better represents the duration of wet and dry periods. But some uncertainty will always remain. It's unlikely, but possible, that events outside the range of the paleo-stochastic baseline occur. Always consider uncertainty and treat conclusions about events with long return periods, such as an extreme wet or dry period that occurs once every 1,000 years, with caution.
Fine-scale relationships are assumed to be stationary – unless evidence shows otherwise
The daily distributions, seasonality and spatial correlation structures derived from climate records are assumed to hold across the modelled paleo-stochastic baseline dataset. Non-stationarity exists in other parts of the baseline – acknowledge any instances of non-stationarity in your model design and communication.
The minimum temporal resolution is daily
The finest temporal resolution of the data is daily – you can use daily or any scales above. Sub-daily extremes cannot be reliably represented.
Relationships between climate variables are preserved at catchment scale
Finer-scale topographic effects and local microclimates may not be fully resolved. Exercise care for point-level applications or highly diverse terrain.
Appropriate scaling is needed to apply climate projections
The choice of scaling method can influence wet-day frequency, extremes and sequencing. Document and test the sensitivity of the model outcomes to different scaling methods to validate the scaling approach. Maintain internal consistency between rainfall, temperature and evapotranspiration when applying changes.
Information and resource hub explaining the methods, data and reviews underpinning the NSW paleo‑stochastic climate baseline.