Research highlights 2023–24

Almost ever since the advent of computers, climate models have been developed to simulate the Earth's climate. First, models were developed to scientifically understand the climate system and later to also understand climate change and predict future climate depending on greenhouse gas and aerosol emissions. Early climate models simulated only the atmosphere, later models added the ocean, and modern climate models, called ‘Earth system models’, also simulate the cryosphere; biosphere, and chemical processes in the atmosphere and ocean. Many countries started developing one or more state-funded climate models through research institutes. Over time, hundreds of such models have been developed, often borrowing computer code from one another. The immense complexity and size of the code often make starting a new model from scratch prohibitively expensive and time-consuming. This has led to a maze of model relationships, which until now have not been properly studied.

Many climate studies today rely on results coming from a set of models called a ‘model ensemble’; for example, by averaging the results of multiple models. This serves as protection from the deficiencies of any single model, but also allows scientists to quantify the uncertainty in their results and capture a greater amount of variability due to our uncertainty about climate processes and how to simulate them on a computer, with each model representing them differently. Including multiple models that share code in such an ensemble then leads to a certain amount of dependency, and certain groups of models sharing substantial amounts of code can be over-represented in the ensemble. Ideally, this should be known and quantified, and over-representing models should be avoided.

In our study, we constructed a code genealogy, or 'family tree', of 167 atmosphere-only, atmosphere–ocean, and Earth system models spanning several decades, including all models participating in the Coupled Model Intercomparison Project (CMIP) phases 3, 5, and 6. We have found that all of the studied models trace their history back to about 14 ancestral models, forming 14 model families. Within each family, models often tend to produce similar results, such as global temperature, so-called ‘climate feedbacks’, and sensitivity of temperature to greenhouse gas concentrations. We have found that only 3 model families comprise about 70% of all models in the latest CMIP phase.

We propose statistical weighting which takes into account the code relationships between models, so that, for example, more balanced averages can be calculated from model ensembles. This is in contrast to a simple 'model democracy', where every model in an ensemble is given the same statistical weight, as in ‘one model, one vote’, which would be only truly applicable if all models were completely independent.

When constructing the genealogy, our focus was on the component of models simulating the atmosphere, but a similar genealogy could be constructed with a focus on the ocean-simulating component. This was motivated by the fact that cloud processes are currently the leading source of uncertainty in future climate projections.

The projected sensitivity of temperature to greenhouse gases has increased on average in the latest generation of CMIP models. Our results contribute to explaining a part of this increase. Models in families with high sensitivity have proliferated more in the latest phase of CMIP compared to other model families and therefore have a greater weight in the overall result, even though they are often closely related. Our proposed weighting, accounting for dependencies, reduces the difference between the latest and the previous generation of CMIP models.

Future research could extend our analysis to the ocean component, quantify the strength of code relationships between models, or look at how the identified model families differ in their simulations of different aspects of the climate system. Overall, this could help explain why climate models do not always agree on their results, and thus reduce our uncertainty in future climate projections.

Our work was done at the Department of Meteorology and Bolin Centre for Climate Research at Stockholm University, and funded by the EU projects FORCeS and NextGEMS and the Swedish e-Science Research Centre. The results have been published in the Journal of Advances in Modelling Earth Systems (https://doi.org/10.1029/2022MS003588).

Wetlands are experiencing intense decline due to various factors, including climate change and human activities such as agricultural expansion, fragmentation by road infrastructure, water impoundment, and freshwater withdrawals for irrigation. Historically, research has focused on local and immediate threats to wetlands, such as pollution, direct land conversion, and water diversion. However, there is growing recognition that remote land-use changes, particularly those affecting upwind atmospheric moisture supply, can significantly influence wetland hydrologic regimes. Understanding these impacts is crucial, as they can alter precipitation patterns and moisture recycling, leading to changes in water availability and, consequently, wetland functionality.

A study by Fahrländer et al. (2024) evaluates the vulnerability of 40 wetlands protected by the Ramsar Convention and their hydrological basins to upwind precipitation shifts caused by land-use and hydroclimatic changes. The study investigates the precipitation source regions of these wetlands with a comprehensive approach that maps where precipitation predominantly comes from, its relationship to historical global land use change, and the hydroclimatic changes recently experienced by the wetlands. This approach allows us to assess how changes in upwind regions influence wetland moisture supply, which is a critical factor in maintaining their hydrologic balance and overall health Fahrländer et al. (2024) use a dataset of atmospheric moisture trajectories generated with the Lagrangian atmospheric moisture tracking model UTrack to trace the sources of atmospheric moisture contributing to precipitation in selected wetlands. They combine this model output with atmospheric reanalysis data, land use, and modelled evaporation and precipitation data to map the precipitation sources given their current land use and analyse changes in moisture supply through anthropogenic land use change. The results are then combined with trends in runoff, expressed as P–E, in a vulnerability indicator to assess the wetlands hydroclimatic vulnerability to upwind land use change.

Emissions of the trace gases carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O) from the coastal ocean, particularly from the shallow inshore areas, have generated strong interest regarding marine emissions. This because these coastal areas are highlighted as potential blue carbon sequestration areas, but are also among the most anthropogenically affected areas of the world thanks to human development and industrial use. RT2 has focused its research activities on a range of topics in the coastal environment by conducting local investigations in the Baltic Sea in collaboration with the Baltic Sea Centre and remote studies in the Arctic.

One of the highlights was the first-ever investigation of the shelf-wide assessment of N2O emissions from the Siberian shelf, looking at the relationship between N2O emissions and the land-derived discharge of nitrogen from large Russian rivers and coastal erosion in a rapidly warming region of the world, published by Wild et. al (2023). Our data shows that concentrations of dissolved nitrogen in the water and of total nitrogen in sediments increase toward large river mouths. On average, water-air N2O fluxes were low, but strong N2O sources and sinks were observed locally. An increase in water temperature could substantially reduce N2O solubility in ocean water, and add to biogeochemical and physical changes. This could alter N2O production in the shelf sediment, and convert the Siberian shelf to a N2O source in the future.

Coastal ecosystems can efficiently remove CO2 from the atmosphere and are therefore promoted for nature-based climate change mitigation. Natural CH4 emissions from these ecosystems may counterbalance atmospheric CO2 uptake. Still, knowledge of mechanisms sustaining such CH4 emissions and their contribution to net radiative forcing remains scarce for globally prevalent macroalgae, mixed vegetation, and surrounding depositional sediment habitats. A 2023 publication by Roth et. al (2023) showed that these habitats emit CH4 in the range of 0.1–2.9mg CH4m−2d−1 to the atmosphere. This was revealed by in situ CH4 emissions from macroalgae which were sustained by conventional methanogenic archaea, but also potentially associated with photosynthetic processes. The key finding was that over an annual cycle, CO2-equivalent CH4 emissions offset 28% and 35% of the carbon sink capacity attributed to atmospheric CO2 uptake in the macroalgae and mixed vegetation habitats offsetting the sink capacity of these potential blue carbon habitats. A multi-method, multi-instrument approach to tackle the complexity of trace gas emissions and carbon sequestration potential with an eye on coastal habitat diversity has risen to the forefront of RT2 investigations for the coming year.

RT3 member Liping Liu and her team has explored how modern zoogeographic regions in Asia developed by examining the relationship between the dental traits of mammalian communities and climatic conditions. Using a method called redescription mining, Liu et. al (2023) have identified patterns of association between these traits and climate variables, providing insights into the history and distribution of mammal species in Asia.

Redescription mining methodology

Redescription mining is a key methodology used in this study. Redescriptions are pairs of logical statements about data variables that capture the interplay between the dental traits of mammals and their climatic conditions. Each logical statement, or query, selects a subset of localities where specific conditions are met, such as temperature or specific dental traits. The similarities between two subsets (one based on climate, the other on dental traits) is measured using the Jaccard coefficient, indicating the strength of their association.

Characterizing modern zoogeographic regions

The study identified nine significant redescriptions (patterns) that highlight the correlation between mammalian dental traits and climatic conditions across Asia, as seen in Figure 1. These redescriptions depict areas corresponding to prominent ecoregions and notable mammalian distribution patterns. This detailed characterisation helps us understand how specific regions are defined by both climate and the evolutionary traits of the mammals which inhabit them.

Emergence of modern zoogeographic patterns

Researchers traced the development of these patterns over five intervals of the Neogene period, which spans over the last 22 million years. They identified both north–south and southeast–northwest zoogeographic divisions in Asia. These patterns reflect changes in climate and tectonic activities, such as mountain ranges, which have significantly influenced the distribution and evolution of mammalian communities.

Impact of paleoclimate models

Various paleoclimate model simulations were used to understand the effect of different climate conditions on mammalian distribution. These models varied in terms of parameters like mountain elevation and climate data from different geological periods. By integrating these models, the researchers could analyse how past climate conditions impacted mammal distributions, and how these conditions have shaped current zoogeographic patterns.

Unique intergration of paleontology and climate science

One of the most interesting highlights of this study is its integration of paleontological data with climate science. The study combines data from fossil records with modern climate models, providing a comprehensive view of how mammalian communities have evolved in response to climatic changes. This interdisciplinary approach enhances our understanding of the historical context of modern zoogeographic regions.

Redescription mining as a tool

The adaptation of redescription mining, typically used in data science, to biological and environmental studies showcases its versatility in uncovering complex patterns in large datasets. This method allows researchers to identify and analyse associations between different variables, offering a powerful tool for exploring ecological and evolutionary dynamics.

Visual mapping of data

The study's results are visualised through maps, as seen in Figure 2, where different colours indicate whether the dental traits and climate conditions match the queries. This visual representation makes it easier to see the geographical distribution and overlap of these traits, providing a clear and intuitive understanding of the data.

Despite considerable differences between historical and modern agriculture, relationships between climate variability and harvest variations of past centuries can be informative when it comes to understanding general climatic risks to crop yields. However, such studies remain rare. A team of researchers representing several disciplines studied the climate signal in long harvest series from parts of Sweden, Switzerland, and Spain from the 16th to the 18th centuries. They found regionally consistent climate–harvest relationships of similar strength and character to modern ones for the same regions and crop types.

The understanding of associations between climate variability and grain harvest variations in early modern Europe (c. 1500–1800) has long been hampered by a lack of collaboration between historians and palaeoclimatologists and by a predominant focus on extreme climate events rather than long-term relationships. A group of historians and climate researchers, with experience in working on the climate–history interface collected a large dataset of already-published data covering grain harvests from the 16th and 18th centuries in central agricultural districts of east–central Sweden, northern Switzerland, and throughout Spain. The choice of regions was motivated by data availability as well as by a wish to obtain a north–south transect across Europe. State-of-the-art palaeoclimate reconstructions of seasonal temperature, precipitation, and soil moisture, based on tree-ring data and documentary sources, were used along with several statistical techniques to investigate the relationships between climate variability and grain harvest variations for barley, rye, oats and wheat. Different methods were used to investigate climate effects on inter-annual and longer timescales.

Grain harvests in east–central Sweden were found to have benefited mainly from wetter summers, in northern Switzerland from dry and warm winters, and in Spain from cooler springs and annual mean temperatures. The strengths of these relationships are comparable to modern climate–harvest relationships at comparable spatial scales. Spain exhibited particularly heterogeneous climate–harvest relationships due to its complex topography. For the summer season in Sweden and Switzerland, the relationship between temperature and harvest was found to be timescale dependent. Warmer summers reduced harvests (presumably through droughts, sub-optimal crop development rates, and plant pathogens) on inter-annual timescales, while this adverse effect of warm summers was not detected when considering longer timescales. Longer periods with warm summers instead increased harvests, at least in Sweden. Despite the northern location, the grain harvests from east–central Sweden (around Lake Mälaren and in Östergötland) were found to be more drought sensitive than those in Switzerland or even Spain. The reason for this is partly the low precipitation in spring and early summer in east–central Sweden, coinciding with the early part of the grain growing season, and the long hours of sunshine during this season. Grain in most of Spain, with its very hot and dry summers, has a key growing season from March to May, rendering the summer drought season largely irrelevant for crop harvests.

The study revealed that the relationship between climate variability and regional harvest variations during the 16th to 18th centuries were weaker than the relationship between climate variability and grain prices during the same period. This is because grain prices, through both trade and storage, represented an average of harvests from large parts of Europe for up to several years, leading to diffused impacts of local weather on harvests. Furthermore, climate variability only explains a portion of harvest variations. This is partly because poor harvests frequently resulted from single days or weeks of unfavourable weather conditions (and not seasonal climate averages). Harvest variations were partly due to factors related to seed quantity and quality, manure availability, access to draft animals, the availability of labour, and pest outbreaks.

While the strength and direction of the relationship between climate variability and grain harvest variations from the 16th to 18th centuries are surprisingly similar to those of today for the same regions and crops, the comparison between historical and modern agriculture is not without its limitations. The genetic variation of early modern grain types was much greater than it is today. Local varieties, suitable for local growth conditions, were used – particularly for barley and rye – complicating the comparisons of the relationship through time and between regions. More importantly, the longer time required to plant and harvest, and the limitations in technology used to dry the grain after harvest, posed other climatic hazards to historical grain production. Modern mechanisation, chemical fertilisers, pesticides, and higher-yielding seed varieties (which have resulted in multiplied harvest yields) have obviously changed the relationship between climate variability and grain harvest variations since pre-industrial times. Finally, while the relationships still remain similar to those of historical times, they are expected to change dramatically during the course of this century across Europe as a result of ongoing climatic warming and potential changes in precipitation patterns and their seasonality.

An investigation led by RT4 member Caroline Greiser, researcher at the Bolin Centre for Climate Research and the Physical Geography Department looked at the mechanisms which create cool microclimates beneath forest canopies during warm and dry summer days. The study reveals the importance of soil moisture in creating cooler forest microclimates compared to temperatures outside forests.