Updated national climate projections for Aotearoa New Zealand

Introduction

The development of updated national climate projections for Aotearoa New Zealand was completed in mid 2024.

The data is available from the Ministry for the Environment (MfE) here.

In the coming months MfE intends publishing a map-based geospatial tool to display the climate projections data as well as insights into the data. You can read more about this work on MfE’s website.

Data has been generated for the recent historic period plus three future Shared Socioeconomic Pathway (SSP)-Representative Concentration Pathway (RCP) scenarios out to the year 2100. (You might find this explainer about SSPs useful). The three SSPs are: SSP1-2.6, SSP2-4.5 and SSP3-7.0.

NIWA is working on a further dataset of SSP5-8.5 projections, which will be complete by the end of 2024.

Background

The work is supporting Aotearoa’s transition to a low-emissions, climate-resilient economy. It helps New Zealanders better understand their climate-related risks by enabling access to the latest detailed climate projections for all New Zealanders. This will improve decision-making for climate adaptation and support resilience to extreme weather events.

The Working Group 1 contribution to the IPCC Sixth Assessment Report was published in August 2021 and was based on the latest suite of global climate model outputs – called the Coupled Model Intercomparison Project Phase 6, or CMIP6. CMIP6 includes around 100 models. See Aotearoa New Zealand climate change projections guidance (Ministry for the Environment) for more details.

NIWA has used six global models from this set and experimented with three regional climate models and sophisticated statistical/machine-learning approaches to develop this next of high resolution projections specific to New Zealand, via a process known as downscaling.

The six global models are: ACCESS-CM2, NorESM2-MM, EC-Earth3, GFDL-ESM4, AWI-CM-1-1MR and CNRM-CM6-1. We have selected these six global models based on their performance over our region (using model output for the period 1980-2015 compared to global observational reanalysis data and based on a large number of climate variables and indicators). We have also ensured that the models span a range of climate model sensitivities and are relatively independent from one another (i.e., they do not share major parts of the model code).

Published alongside this model suite will be downscaled projections from the New Zealand Earth System Model (NZESM), developed here in New Zealand through the Deep South Challenge to improve our understanding of the Southern Hemisphere processes that strongly influence our climate.

The work has been funded by the Ministry for Business, Innovation and Employment (with the contract managed by the Ministry for the Environment). This means the model outputs will be made available at no cost to all New Zealanders.

Project stages

Project stages include:

Completed: Selection of climate models from the CMIP6 suite (this has been done, with six global models and three regional models selected), setup and testing.
Completed: Regional climate model runs using NIWA’s High Performance Computing Facility.
Complete: Further data downscaling using statistical and machine learning approaches.
Complete: Bias correction.
Complete: Data post-processing and developing an appropriate mechanism to make datasets accessible to end users.
Underway: Downscaling, bias correction and processing of SSP5-8.5 projections.

The selection of the three SSPs for dynamical downscaling was heavily influenced by computing constraints, with the decision to not select SSP1-1.9 based on the recommendations of the international CORDEX downscaling community to prioritise the middle three SSPs. We also took into account the decision of the Deep South National Science Challenge to select the same three SSPs for the NZ Earth System Model runs and noted that this decision was based on a stakeholder survey.

The regional climate model runs took approximately 1 year to complete (June 2022 through June 2023). This aspect of the project (Project Stage 2) used approximately 70% capacity of one NIWA supercomputer (requiring over 12 million core hours in total). This computing power is equivalent to running more than 150 high end laptops non-stop for a year.

More detail can be found in reports on Methodology and Bias Correction (written by NIWA for the Ministry for the Environment).

Project outputs – core public dataset

For all the listed climate variables below, downscaled data (GIS-ready files at 5 km spatial resolution for all NZ) will be available for:

the average of the six global models outlined above
three SSPs (with SSP5-8.5 coming in late 2024 as detailed above)
three future periods (2021-2040, 2041-2060 and 2081-2100)
four global warming levels (1.5°C, 2°C, 3°C and 4°C) (* see below for more details about warming levels)
from two historical baselines (1986-2005 and 1995-2014).

These time slices, warming levels and baselines are consistent with the IPCC Working Group 1 Atlas (click on the Regional Information (Advanced) link).

Climate variable	Annual	Seasonal
Average daily air temperature: Tmin, Tmax, Tmean, Trange	x	x
Number of hot days (>25°C)	x	x
Number of frost days (<0°C)	x	x
Number of very hot days (>30°C)	x	x
Growing degree days (base 5°C & 10°C)	x
Cooling degree days (base 18°C)	x
Heating degree days (base 18°C)	x
Total rainfall	x	x
Number of dry days	x	x
Number of rainy days (>1mm)	x	x
Number of very rainy days (>25mm)	x	x
Heavy rainfall (99^th percentile)	x	x
Drought (PED) accumulation	x
Average wind speed	x	x
Number of windy days (>10m/s)	x	x
Strong wind (99^th percentile)	x	x
Average relative humidity	x	x
Average solar radiation	x	x

In addition to the above datasets there is already public access to projected changes in extreme rainfall via HIRDS (which NIWA will be looking to update) and to the 100-year ARI storm tide coastal inundation layers for current day and 10 cm increments of sea-level rise up to 2 m.

Other frequently requested information includes projected changes to river flows and flooding. NIWA is actively working on assessing changes to flood risk in their MBIE-funded Endeavour research programme and has begun work to update river flow projections.

* Notes about warming levels: Since the downscaled (i.e. CCAM, or Conformal Cubic Atmospheric Model) simulations begin in 1960, it was not possible to compute warming level change maps relative to a pre-industrial base period from CCAM. As such, the warming level change maps instead display the climate change signal relative to a modern base period (i.e. optionally 1986-2005 or 1995-2014). However, the future period associated with these change maps is based on when a global warming level is reached utilising data from the host GCM, defined as the 20-year window when the rolling mean reaches the corresponding global warming level relative to the pre-industrial mean (1850-1900), following the IPCC Atlas. For certain scenarios, a particular warming level may not be reached in all models. The multi-model mean for the corresponding warming level is therefore only included if at least four of the six models reach the warming level.

Example output

Animation showing hourly precipitation (yellow > red shading shows heavy precipitation) and cloud cover (white shading):

Animation showing hourly precipitation

Watch Video

NIWA's additional climate modelling

In addition to the dynamical downscaling, NIWA will once again be producing climate projections based on statistical downscaling (although for the CMIP6 update we will be using more sophisticated Machine Learning methods). Development of this procedure was recently outlined in a peer-reviewed journal article: High-resolution downscaling with interpretable deep learning: Rainfall extremes over New Zealand. Benchmark evaluations show that this new method vastly outperforms traditional statistical downscaling approaches for rainfall and extreme events over New Zealand.

This work will downscale many more CMIP6 global models (at least 20). This will complement the dynamically downscaled projections by providing information based on a wider range of models and scenarios. This two-pronged approach is consistent with NIWA’s previous downscaling work.