Search

The modelling of small-scale processes is a major source of error in climate models, hindering the accuracy of low-cost models which must approximate such processes through parameterization. Red noise is essential to many operational parameterization schemes, helping model temporal correlations. We show how to build on the successes of red noise by combining the known benefits of stochasticity with machine learning. This is done using a physically-informed recurrent neural network within a probabilistic framework. Our model is competitive and often superior to both a bespoke baseline and an existing probabilistic machine learning approach (GAN) when applied to the Lorenz 96 atmospheric simulation. This is due to its superior ability to model temporal patterns compared to standard first-order autoregressive schemes. It also generalises to unseen scenarios. We evaluate across a number of metrics from the literature, and also discuss the benefits of using the probabilistic metric of hold-out likelihood.

A new model is presented for multisite statistical downscaling of temperature and precipitation using convolutional conditional neural processes (convCNPs). ConvCNPs are a recently developed class of models that allow deep-learning techniques to be applied to off-the-grid spatio-temporal data. In contrast to existing methods that map from low-resolution model output to high-resolution predictions at a discrete set of locations, this model outputs a stochastic process that can be queried at an arbitrary latitude–longitude coordinate. The convCNP model is shown to outperform an ensemble of existing downscaling techniques over Europe for both temperature and precipitation taken from the VALUE intercomparison project. The model also outperforms an approach that uses Gaussian processes to interpolate single-site downscaling models at unseen locations. Importantly, substantial improvement is seen in the representation of extreme precipitation events. These results indicate that the convCNP is a robust downscaling model suitable for generating localised projections for use in climate impact studies.

Ensembles of chemistry-climate models (CCMs) are fundamental for the exploration of the chemistry-climate system. A particular focus of chemistry-climate modelling is stratospheric ozone, whose concentrations have been decreased by anthropogenic releases of ozone depleting substances. In conjunction with observational data, CCM ensembles have been relied upon to simulate historic effects of ozone depletion and to project future ozone recovery. However, many widely used ensemble analysis methods are simplistic and are based upon incorrect assumptions about the design of the ensemble. Multi-model means used to construct future ozone projections do not account for variable model performance or similarity and therefore give biased and inaccurate projections. Similarly, simplistic linear regression methods used to infill historic ozone records underestimate interannual variability and are inaccurate in regions of sparse data coverage. Moreover, given advances in machine learning and data science and their increased use in environmental science, it is timely to apply more advanced tools to CCM ensembles. To address this methodological deficit, this thesis presents a set of novel tools to improve the predictions and projections from CCM ensembles of stratospheric ozone. A process-based weighted mean is developed which accounts for model performance and similarity in CCM ensembles. This improvement over pre-existing methods was used to generate accurate ozone hole recovery projections. This thesis also developed a Bayesian neural network (BNN) which fuses together CCMs with observational data to produce accurate and uncertainty-aware predictions. The BNN framework was used to produce historic continuous datasets of total ozone column and vertically resolved ozone, and represents a significant improvement in methods used to ensemble models. Though designed for CCM ensembles these flexible tools have the potential to be applied to other environmental modelling disciplines to

The 880 million agricultural workers of the world are especially vulnerable to increasing heat stress due to climate change, affecting the health of individuals and reducing labour productivity. In this study, we focus on rice harvests across Asia and estimate the future impact on labour productivity by considering changes in climate at the time of the annual harvest. During these specific times of the year, heat stress is often high compared to the rest of the year. Examining climate simulations of the Coupled Model Intercomparison Project 6 (CMIP6), we identified that labour productivity metrics for the rice harvest, based on local wet-bulb globe temperature, are strongly correlated with global mean near-surface air temperature in the long term (p ≪ 0.01, R 2 > 0.98 in all models). Limiting global warming to 1.5 °C rather than 2.0 °C prevents a clear reduction in labour capacity of 1% across all Asia and 2% across Southeast Asia, affecting the livelihoods of around 100 million people. Due to differences in mechanization between and within countries, we find that rice labour is especially vulnerable in Indonesia, the Philippines, Bangladesh, and the Indian states of West Bengal and Kerala. Our results highlight the regional disparities and importance in considering seasonal differences in the estimation of the effect of climate change on labour productivity and occupational heat-stress.

Arctic sea ice forecasting is a major scientific effort with fundamental challenges at play. To address such challenges, we have developed a physics-informed, data-driven sea ice forecasting system, IceNet, which outperformed a leading dynamical model (ECMWF SEAS5) in monthly-averaged forecasts of pan-Arctic sea ice concentration. IceNet adopted a U-Net deep learning architecture and was trained on over 2,000 years of CMIP6 climate simulation data. Despite its state-of-the-art seasonal forecasting skill at lead times of 2-6 months, IceNet has two main limitations. First, it could not outperform the dynamical model in short-range (1-month) forecasts. This is partly caused by IceNet operating on monthly-averages, which smears the initial conditions and weather phenomena that can dominate predictability at short time scales. Second, IceNet is afflicted by the ‘spring predictability barrier’ that affects all long range forecasts of summer. This predictability barrier arises primarily due to the importance of melt-season ice thickness conditions on summer sea ice. Here we present our early findings from IceNet2, which attempts to alleviate these issues by operating on daily-averages and including sea ice thickness as an input variable. IceNet2 paves the way for our efforts to aid the Arctic conservation community by developing the first public, operational sea ice forecasting AI.

The most mature aspect of applying artificial intelligence (AI)/machine learning (ML) to problems in the atmospheric sciences is likely post-processing of model output. This article provides some history and current state of the science of post-processing with AI for weather and climate models. Deriving from the discussion at the 2019 Oxford workshop on Machine Learning for Weather and Climate, this paper also presents thoughts on medium-term goals to advance such use of AI, which include assuring that algorithms are trustworthy and interpretable, adherence to FAIR data practices to promote usability, and development of techniques that leverage our physical knowledge of the atmosphere. The coauthors propose several actionable items and have initiated one of those: a repository for datasets from various real weather and climate problems that can be addressed using AI. Five such datasets are presented and permanently archived, together with Jupyter notebooks to process them and assess the results in comparison with a baseline technique. The coauthors invite the readers to test their own algorithms in comparison with the baseline and to archive their results.

Multi-output Gaussian processes (MOGPs) leverage the flexibility and interpretability of GPs while capturing structure across outputs, which is desirable, for example, in spatio-temporal modelling. The key problem with MOGPs is their computational scaling O(n3p3), which is cubic in the number of both inputs n (e.g., time points or locations) and outputs p. For this reason, a popular class of MOGPs assumes that the data live around a low-dimensional linear subspace, reducing the complexity to O(n3m3). However, this cost is still cubic in the dimensionality of the subspace m, which is still prohibitively expensive for many applications. We propose the use of a sufficient statistic of the data to accelerate inference and learning in MOGPs with orthogonal bases. The method achieves linear scaling in m in practice, allowing these models to scale to large m without sacrificing significant expressivity or requiring approximation. This advance opens up a wide range of real-world tasks and can be combined with existing GP approximations in a plug-and-play way. We demonstrate the efficacy of the method on various synthetic and real-world data sets.

If financial markets are to realign towards truly sustainable development the financial sector needs to differentiate commercial actors more accurately on their climate and environmental performance. A potential breakthrough to help in this challenge is the emerging field of ‘Spatial Finance’, the independent assessment of the location of a company’s or a country’s assets and infrastructure using ground data, remote sensing observations and modelled insights, offers a potentially transformative means to gain improved quantitative ESG insights. Rapid development is required if spatial finance is to deliver in the short term. In this report, authors outline a possible taxonomy and hierarchy for spatial finance, showing how discrete forms of technology, approaches and data can be considered within a single consistent framework. Using this framework, spatial finance could provide insights at differing scales for different applications from the asset-scale for project finance, to company-scale for investment, to country scale for sovereign debt. Throughout the document authors provide insights into current cutting-edge developments within the field, illustrated with case studies from practitioners and data providers, and explore potential future developments.

An important source of polar stratospheric clouds (PSCs), which play a crucial role in controlling polar stratospheric ozone depletion, is the temperature fluctuations induced by mountain waves. These enable stratospheric temperatures to fall below the threshold value for PSC formation in regions of negative temperature perturbations or cooling phases induced by the waves even if the synoptic-scale temperatures are too high. However, this formation mechanism is usually missing in global chemistry–climate models because these temperature fluctuations are neither resolved nor parameterised. Here, we investigate in detail the episodic and localised wintertime stratospheric cooling events produced over the Antarctic Peninsula by a parameterisation of mountain-wave-induced temperature fluctuations inserted into a 30-year run of the global chemistry–climate configuration of the UM-UKCA (Unified Model – United Kingdom Chemistry and Aerosol) model. Comparison of the probability distribution of the parameterised cooling phases with those derived from climatologies of satellite-derived AIRS brightness temperature measurements and high-resolution radiosonde temperature soundings from Rothera Research Station on the Antarctic Peninsula shows that they broadly agree with the AIRS observations and agree well with the radiosonde observations, particularly in both cases for the “cold tails” of the distributions. It is further shown that adding the parameterised cooling phase to the resolved and synoptic-scale temperatures in the UM-UKCA model results in a considerable increase in the number of instances when minimum temperatures fall below the formation temperature for PSCs made from ice water during late austral autumn and early austral winter and early austral spring, and without the additional cooling phase the temperature rarely falls below the ice frost point temperature above the Antarctic Peninsula in the model. Similarly, it was found that the formation potential for PSCs m

The current method for averaging model ensembles, which is to calculate a multi model mean, assumes model independence and equal model skill. Sharing of model components amongst families of models and research centres, conflated by growing ensemble size, means model independence cannot be assumed and is hard to quantify. We present a methodology to produce a weighted model ensemble projection, accounting for model performance and model independence. Model weights are calculated by comparing model hindcasts to a selection of metrics chosen for their physical relevance to the process or phenomena of interest. This weighting methodology is applied to the Chemistry-Climate Model Initiative (CCMI) ensemble, to investigate Antarctic ozone depletion and subsequent recovery. The weighted mean projects an ozone recovery to 1980 levels, by 2056 with a 95 % confidence interval (2052–2060), 4 years earlier than the most recent study. Perfect model testing and out-of-sample testing validate the results and show a greater projective skill than a standard multi model mean. Interestingly, the construction of a weighted mean also provides insight into model performance and dependence between the models. This weighting methodology is robust to both model and metric choices and therefore has potential applications throughout the climate and chemistry-climate modelling communities.