10 Topics from Chapter 10: Linking Global to Regional Climate Change
Carina Leighton
The IPCC AR6 Chapter 10 helps distill the methodology that connects global climate change impacts, such as temperature change increase, to more regional impacts on different regions. Other chapters of the AR6 report discuss more specific regional impacts, like extremes, but chapter 10 is dedicated to analyzing this difference in global and regional impacts and how we get there.
Chapter 10 makes several claims that it can back up during the chapter. The first major claim relates to the idea of having multiple pieces of evidence, such as combining ice core data with atmospheric data where the core was taken, will allow for a much higher confidence in the data that has been and will reduce the uncertainty of the data given. The second major claim states that global climate data turns out to be incredibly useful at the smaller, regional scale. These two major claims provide a baseline for the understanding of regional climate predictions and their relation to the global data collected over time. This leads to less uncertainty in future predictions, meaning humanity can be more confident in what to expect for any particular region.
Chapter 10 begins in 10.1 by laying the foundations for regional climate change by providing critical definitions as well as context that makes the understanding of regional climate change possible. The chapter then moves into 10.2, discussing observational techniques to understand regional climate change, including remote sensing data as well as things like quality control and homogenization. In chapter 10.3, the chapter moves into models, looking at how the models are created and what they’re used for, as well as how they’re used for specific phenomena such as fronts and mountain weather. 10.4 takes a look at the interplay between global anthropogenic change and natural regional changes to the climate, discussing how specific areas like the North American drought and the monsoons of multiple areas in the world are affected by this interplay. From there, 10.5 moves into the combination of what has been shown so far, discussing techniques for the distillation of regional climate change information as well as how different contexts impact humanity’s understanding of climate science. 10.6 rounds out the chapter by providing examples of how the data mentioned in 10.2 and 10.3 can be distilled and understood in a regional context, including areas such as India and the Mediterranean. The chapter is then finished with a few final remarks. Overall, Chapter 10 covers an important section of humanity’s current understanding of climate science and allows for direct comparisons of global and regional climate science and allows for a more thorough examination of the methods involved in studying regional climate science.
10.1
Chapter 10 begins with the baseline knowledge necessary for the understanding of what the rest of the chapter discusses and how it might be useful. This includes, but is not limited to, a distillation of general regional climate change drivers that make clear the impact the global scale has on the regional.
The section begins, however, with an infographic that explains how climatological data gets from the source to people’s homes. This graphic, figure 10.1, is a simplistic measure of how data moves in the climate field. The figure explains that different types of data, such as satellite data and process understanding, a way in which idealized environments can set up knowledge for later (like a model but not using real-world data), is synthesized then constructed into an easy-to-use user format that the public can easily understand.
The second of the 3 graphics marking the beginning of the chapter is quite simply a summary of the chapter. The 3rd graphic however, figure 10.3, is an important diagram showing how different atmospheric processes happen and different time- and distance scales, separating these out into separate climate models for easy linkage to each one. At the lower end, urban canopy (UC) models cover a scale of half a kilometer to 10 kilometers, lasting up to maybe a day. This area includes events such as tornadoes, thunderstorms, and convection. Moving up in scale, the figure shows the cloud resolving models (CRMs), which include events previously mentioned as well as ones such as squall lines. This increase in scale gives these sets of models a unique purpose, different from the UC models. As the scale increases, so too do the models change to reflect that, moving into the regional climate models (RCMs) and eventually into the global climate models (GCMs). This graphic is a useful guide for atmospheric scale as well as which models are associated with which scale.
10.1 then makes an important distinction that is useful for this chapter only. In this chapter, a region is not locked to the traditional AR6 definition, meaning it is not set on particular bounds. As the chapter explains, this allows for the scale to move between regional, like mountain ranges, all the way up to the size of a subcontinent, an important distinction that becomes important later in the chapter.
10.1.2 also makes the note that the reference periods in this chapter and all following chapters (not covered today) are not in the same scale as previous chapters! What this means is that the references to specific time periods should use what is listed and not the rest of the report. Due to the nature of regional climate change and its increased speed at which it is impacted, this is important to note.
10.1 then moves into regional climate variability, distilling information gone into detail in section 10.4. The chapter clearly indicates that regional climate changes are most likely caused by a combination of natural and anthropogenic forcings, as in, natural variability and feedback are exacerbating the issues that humanity has created. Of the natural forcings, there are many types that could influence how the climate changes in different ways.
The first of these forcings the chapter mentions is the solar cycle, the 11 year change in solar radiation. This is up for debate, as studies have both shown correlation as well as not. What this means is that the solar cycle might impact the amount of radiation that reaches the planet which could in theory change how warm the Earth gets on an 11 year cycle, but it’s still being contested. The report states that it has a higher likelihood of affecting winter circulation more than summer, most likely due to a higher gradient of temperature between the equator and the poles.
Another source of climate forcing mentioned is changes to stratospheric ozone, triggered by human activity. This has changed a variety of different atmospheric circulation structures, which has led to the movement of locations of precipitation. If ozone continued to change, humanity would likely see greater changes to these circulation patterns.
More importantly than that, however, is the influence of aerosols on regional climate forcing. Their nature makes it easy to understand why they’re so important, primarily due to their regional emission. Because they rely on the atmosphere to transport, they don’t affect the world on a global scale, but enough can alter the precipitation and temperature of a given area given their ability to scatter radiation. It is much harder to understand or control the natural aerosols produced, such as minerals and volcanic eruptions, but anthropogenic aerosols do play a key role in the dysregulation of our climate in their forcing. Volcanos, despite their inability to be controlled, can affect multiple regions if a volcanic plume reaches the stratosphere, as the particles can be left up there for years. This can affect global temperatures, just as anthropogenic aerosols affect regional ones.
There are also internal drivers of climate variability, mostly surrounding ocean circulation. Circulation in the north Atlantic impacts the European continent, and (multi)decadal variability in the Atlantic and Pacific oceans impacts the regional climate of the areas surrounding their respective oceans. Of course, interactions with the oceans and other forms of variability can lead to non-linear interactions and cause changes to their impact over time.
Rounding out 10.1.3, there is uncertainty on the impact of these forcings. Basically, it is treated in the same capacity that it’s treated for global climate change variability. Using process understanding and systematic comparison of data, this is attempted to be mitigated as much as possible. Model uncertainty is managed at multiple places in the process, mitigating each of the uncertainties associated with the models or the pathways through similar techniques to non-model components.
10.1.4 shows the ways in which humanity gets information surrounding regional climate change. The section states that these methods are diverse and selection of sources changes based on what is needed. This changes based on who is asking for the information, as well as where the information is being gathered from. The data is part of a linear supply chain ranging from data to user, but it has some issues related to assumptions made about the data as it is passed along, similar to the children’s game ‘Telephone.’
Section 10.1 ends by looking at what regional climate information is used in the AR6 and how it compares to AR5. Most of the specifics are dedicated to Chapters 11 and 12 of the AR6, as well as the Interactive Atlas. Other chapters in AR6 took regional components as well, based on specific contexts needed for those chapters.
Box 10.1 touches on the comparison to the AR5 component by showing the data already known and what the current AR needs to address in its upcoming sections and chapters. The first of the major issues with AR5 was found to be in relation to its ability to properly assess regional climate models, or RCMs, as studies were only just coming out, as well as the inherent variability in their data. The Working Group II Chapter 21 found issues with the use of global datasets to inform regional climate data, which is a mistake as data from there can be misaligned to the local scale. The same chapter also looked at the context in which a study was produced, as the two major contexts could influence the biases in them. The Special Report on Climate Change and Land found that heat waves would increase with drier soil moisture (which would help explain the heatwave in the northwest US in 2021), as well as the fact that water management was not accounted for in the CMIP5 (global climate model). The Special Report on the Ocean and Cryosphere showed that it is easy to connect ice mass loss in Antarctica and Greenland to anthropogenic climate change, and that it will continue on humanity’s current path.
The Special Report on Global Warming of 1.5C reveals major differences in the climate between 1.5oC and 2oC and that pathways to 1.5oC may involve passing through it at some point.
Cross Chapter Box 10.1 spends time looking at the Arctic’s impact on mid-latitude climate. It shows little impact of Arctic changes to mid-latitude climate. There is contrasting evidence that cannot yet be reconciled as of today.
10.2
Section 10.2 is dedicated around observations and regional climate information. More specifically, it is about how one can use them to construct said information. The section begins by discussing different observational types and how they can be used, the first type of which is surface in-situ observations. These observations come from multiple groups, such as climate reference networks, among others. The data found is used to understand differences in regional climate as well as record major changes to it.
Satellite data also plays an important role in data collection. Most effectively used over areas where in-situ data is unavailable, their use of multiple portions of the electromagnetic spectrum allows them to prove incredibly useful. Specifically, the Low Earth Orbit (or LEO) satellites have been able to record data on a variety of different measures, ranging from aerosol movement to clouds and precipitation data, allowing for more precise weather prediction and more accurate understanding of the climate. Precipitation especially has seen benefit, as using these satellites, it is possible to see day-to-day changes in precipitation. However, a major gripe found with such a design is that the resolution (how far down a satellite can see into the atmosphere, i.e., 1.5 km) does change, causing issues when trying to look at soil or precipitation.
Derived products also play an important role. These products take data already known and combine it with mathematical theory to create a new analysis. Regional reanalyses are useful to represent the extreme ends of many critical categories, like wind and temperature. They do come with high uncertainty due to a low resolution.
There are also challenges to assessing regional climate change, the first of which is quality control. Quality control is a check to make sure that the data given is actually accurate and that there aren’t any uncontrolled factors influencing it. If any bad data is shown, it is simply removed. There is now wide use of QC data on the regional scale, but its use is challenging.
Other challenges to climate data include homogenization (making the data flow evenly through time and space, so there’s data for all space and time), data scarcity, gridding (making data that is produced on a grid, which feeds into homogenization), and obtaining mountain data (Section 10.2.2). These four issues are integrally connected as they each require each other in order to make progress in humanity’s understanding of regional climate change. Each of them have unique ways to solve them, and are currently being worked on today.
Humanity is also using regional climate data for calibrating statistical methods and to help assimilate paleoclimate data. The former is used to derive climate information from climate simulations, while the latter is a technique used to help humanity understand the past and sync it up with the current day. Using these techniques humanity has the ability to make better predictions about what to expect in the future, whether it be by interpreting models or understanding the paleoclimate.
In the future, the goal is to continue the recovery of archives of meteorological data from years past, as well as try and provide an easier way to better understand the high resolution processes that make up regional data. Hopefully this will provide humanity with a more precise understanding of the world we live in and allow humanity to adapt more quickly.
Using Models for Constructing Regional Climate Information
A summary of Section 10.3 of the IPCC AR6 WG1 Draft Report
Leslie Nguyen
Introduction
A variety of methods exist to generate regional climate data. Regardless of the model type, performance of models in simulating regional climate can be measured using the same set of factors such as: temporal and spatial averages of precipitation or temperature; seasonal cycles of climate in a region; representation of regional processes and phenomena, feedbacks, drivers, and forcings.
The different methods and model types used for simulating regional climate and climate change discussed in detail in Section 10.3 include: global climate models, regional climate models, and statistical models, which include perfect prognosis, bias adjustment, and weather generators. Model experiments can also be used to examine regional climate information and are most often used to understand specific instances of climate disturbances, like volcanic eruptions, or observed past trends. Figure 10.6 in the IPCC report is a visual representation of how different models provide different results for the same region. Given that different models can produce varying climate information, understanding when one model type is preferred over another is important for developing confidence in regional climate projections.
The remainder of this essay will evaluate how these different types of models can lead to “improvement” in representation of regional climate in simulated results and the “added value” of using different types of models. In the context of this essay, “improvement” means new or more accurate information about a particular region or climate phenomena, and “added value” will refer to when a model provides information about a particular region or climate phenomena that is not observed using other models.
Global Climate Models
Global climate models can simulate both past and future climates at the global scale and are generally used to extract climate information for the entire Earth or at the continental level. Global climate models typically have spatial resolutions between 100-200 km. Regional climate data can be obtained from these global models by looking at specific regions of interest. Because global climate models simulate climate across the whole planet, there are limitations to generating credible regional projections from these modeling results.
Global climate models can be used to model large-scale climate phenomena that also control regional climate patterns. El Niño Southern Oscillation (ENSO) is a tropical phenomenon that occurs in the tropical Pacific Ocean, which has an impact on climate at both the global and regional scale. This phenomenon highlights the interconnectedness between the ocean and the atmosphere, and has three phases: neutral, La Niña, and El Niño. The neutral phase can be considered “normal conditions,” and in this phase, the western Pacific waters are warmer than eastern Pacific waters. Dry, cool air sinks above the colder eastern Pacific waters and moves west towards warmer ocean waters. As these winds originate from the east, they are called easterly winds. Air rises above the warmer waters in the western Pacific, which results in rainfall over Indonesia and other western tropical Pacific areas. During La Niña, the temperature differential between the eastern and western tropical Pacific Ocean is greater, leading to stronger easterly winds, increased convection, and significantly more rainfall over Indonesia and eastern Australia. El Niño events occur when the eastern Pacific waters are warmer than normal and western Pacific waters are cooler than normal, leading to a reversal of wind direction in the tropical Pacific. Cooler air moves from the west to the east, causing less precipitation in Indonesia, as there is more dry air sinking above the western Pacific, and increased rainfall over the eastern tropical Pacific Ocean.
A more detailed description of model performance in simulating ENSO is provided in Chapter 3 of the IPCC report. However, in terms of modeling regional impacts of ENSO, global climate models can simulate certain expected features. The models correctly simulate the Arctic stratospheric response to ENSO, with the polar vortex weakening during El Niño events and strengthening during La Niña. The Northern Hemisphere polar vortex consists of strong westerly winds in the stratosphere above the Arctic during winter. During El Niño, normal Pacific Ocean surface winds are reversed and come from the west, resulting in upper troposphere convection movement to be easterly. When these upper troposphere convection waves travel up to the stratosphere, it can disrupt the westerly winds of the Arctic polar vortex, thus weakening it.
Because La Niña events strengthen the usual easterly winds, it can also strengthen the polar vortex. One response to ENSO that has not been well captured in global climate models is the observed weakening of the Southern Hemisphere polar vortex during El Niño events in the central Pacific. Additionally, the latest global models have not shown any improvement in spatial patterns of ENSO’s influence on winter precipitation. Even when certain large-scale features are well represented, the resulting climate projections may not be accurate.
Other biases can be generated in global climate models, which affect regional climate representations. For example, in the mid-to-high latitudes, there is a large-scale meteorological phenomenon called atmospheric blocking, which blocks and diverts certain cyclones and leads to cold winters and warmer summers. This blocking essentially defines the seasons for certain regions, and accurate representation of this phenomenon is critical for developing regional climate data. However, as explored in Section 3.3.3.3 of the IPCC report, the latest global climate models still show blocking biases in terms of location of blocking, frequency, and persistence (Figure 10.7).
Global models have not performed well when there is complicated regional processes, feedbacks, or nonlinear scale interactions. One area where this is evident is with simulating regional feedbacks, such as snow-albedo and soil-moisture feedbacks. Soil-moisture feedbacks affect both temperature and precipitation. As temperatures increase, the air’s capacity to hold water will also increase, and the potential evapotranspiration will increase. Therefore, if soil moisture is high, more moisture will evaporate out of the soil, increasing moisture content in the atmosphere and amplifying the temperature increase.
In climate models, preceding precipitation is used as a proxy to estimate soil moisture content. Global climate models do not realistically represent the soil-moisture feedback described above; some models overestimate the influence of preceding precipitation on temperature while others unrealistically model the impact of evaporation on temperature extremes in wet areas of Europe and the US. This is likely because the effects of soil moisture on temperature is observed on a local scale, which global models have difficulty simulating at the lower resolution.
Additionally, it has been suggested that the effects of soil-moisture on precipitation may vary depending on the region, including whether there is a positive or negative feedback. Therefore, trying to represent this phenomenon on the global scale does not make sense. In situations like this, the IPCC report recommends using variable resolution global models (described below) or regional climate models (described in the next section) which may be better suited to resolve these specific areas of interest.
Variable resolution global models are a special class of high-resolution global models that have the potential to produce improved regional climate data as compared with standard global models. These models have increased resolution such that regional climate phenomena are better represented, while still resolving global-scale climate processes. However, increasing resolution is not the answer to all biases. For example, the interaction of Asian monsoon rainfall with mountains in the area are better represented with higher resolution, but there is still a major dry bias in the results.
Regional Climate Models
Regional climate models can simulate dynamical climate elements, like global models, but are only applied over a specific region and at a higher resolution scale. The spatial resolution of these models is typically between 10-50 km, but certain models are also run at “convection-permitting” resolutions, which is considered 4 km or less. Regional models can be driven by global models, where the outputs of global models are inputs into the regional model. This is known as dynamical downscaling. Regional models can improve upon global climate models in representing regional weather and climate phenomena. This is especially true in regions with complex terrain and local-scale phenomena, because of the use higher spatial resolution, which can lead to better representation of certain regional climate phenomena, as explored below.
In general, increasing the resolution of models can lead to better representation of local topography, which plays an important role in driving wind patterns and thus regional climate. In particular, mountains and bodies of water have been shown to impact diurnal climate patterns. In areas like the Rocky Mountains, mountain slopes warm during the day significantly more than the air surrounding the mountains. This temperature gradient means that warm air near the mountain surface will rise. Because mountain surfaces are sloped, the warm air rises along the mountain slope, almost like a chimney. This phenomenon is called an upslope wind. However, at night, the reverse happens, where the mountain slopes are cooler, the air along the slopes are cooler, and cold air drops down towards the base of the mountains. Accurately modeling this convection phenomenon is only possible when the mountains are highly resolved. For the Rocky Mountains, regional models with spatial resolutions of around 4 km have been able to simulate this diurnal temperature and wind cycle, but an even finer resolution of 1 km is needed to simulate this effect in the Alps.
Regional climate of coastal areas and land near large lakes can be difficult to model due to a combination of factors, including complex terrain and the heat capacities of land and water. As with mountains, the terrain near bodies of water makes resolving local convection difficult. However, regional models using resolutions ranging from 8-50 km have shown to improve model performance in these coastal and shoreline areas. One important coastal climate phenomenon is land and sea breezes. Land and sea breezes are a result of differences in heat capacity of land and water. During the day, the land warms up more quickly than the water, and thus air above land rises. This allows for cooler air above the water to move inland; this is called the sea breeze. At night, the air flow is reversed – cooler air from over land flows out towards the sea and is referred to as the land breeze. This diurnal movement of air also impacts precipitation in coastal areas. Moist air above the ocean is transported over land, warms up over the land, and then rises. Precipitation requires water vapor and rising motion, both of which are present in these coastal areas due to land and sea breezes. Therefore, improper representation of land-sea breezes in climate model simulations can lead to misrepresentation of precipitation in coastal areas. Higher spatial resolution climate models help to better resolve these small-scale convection mechanisms that are important drivers of regional climate.
At convection-permitting resolutions (4 km or smaller), fine-scale precipitation data can help improve understanding of regional cloud cover and its radiative effects. Clouds are relatively small-scale features, and global models with larger resolutions cannot represent these features accurately, which can lead to the largest differences in outcomes between global models. With finer resolutions, cloud cover is better represented, and the feedback mechanisms of clouds can be observed. Depending on the type of cloud, climate feedbacks can be either positive or negative. Clouds have relatively high albedo and can reflect incoming solar radiation, leading to cooling of the atmosphere. However, clouds are made of water, which has a high heat capacity, and therefore, clouds can also trap heat effectively, leading to the atmosphere retaining longwave radiation from the Earth instead of releasing it into space. These competing effects of clouds mean that clouds at different heights in the troposphere can result in either a positive or negative climate feedback. Being able to resolve where these clouds appear using higher resolution regional climate models can add value to regional climate simulations. Additionally, being able to accurately model convection at a local scale can lead to improved modeling of tropical cyclones. Models using horizontal resolutions between 1-8 km have been able to represent tropical cyclone features such as eyewall structure (Figure 10.8). However, modeling at this resolution also requires significant computational power, and is not used currently for routine climate change modeling.
Biases in regional model outputs can be a result of biases inherited by the driving global model. Depending on the type of biases, this can lead to either uncertain or improved regional climate simulations. In the case of atmospheric blocking, the biases observed in global models (as described in the Global Climate Model section above) are essentially transferred to the regional model, leading to regional blocking frequencies to be misrepresented in the regional climate data. Therefore, in regions where large-scale phenomena like atmospheric blocking is an important driver of regional climate, global models may not be the best source of data to downscale to this particular region. However, when looking at modeling ENSO teleconnections, although global models have difficulty predicting precipitation, regional models have been shown to accurately reproduce regional precipitation responses to ENSO. For regions where climate is controlled by ENSO, it may be appropriate to use dynamical downscaling of global climate models because the regional climate model can add value in representing precipitation patterns due to ENSO.
Statistical Models
Statistical models are an alternative to dynamical downscaling to obtain regional climate projections. Statistical approaches analyze the results from global climate models to produce information about regional climate. Some types of statistical models explored in this essay are: perfect prognosis, bias adjustment, and weather generators.
In general, the of statistical models to correct outputs of global models have been shown to improve upon the outputs at a single location. However, regional models perform better when considering spatial data. Nevertheless, statistical downscaling methods may be the appropriate method of generating regional climate information, given the right circumstances and goals.
Perfect Prognosis
Perfect-prognosis models are statistical models that generate a link between observed large-scale phenomenon and local-scale variables of interest, like precipitation. This relationship is then applied to global climate model outputs to generate local or regional climate predictions. This type of statistical model performs well when large-scale forcings drive regional climate. For example, precipitation and temperature are both represented well when perfect-prognosis models are used, resulting in improved data for daily time series or local weather statistics. One concern with these types of models for future climate projections is that there is limited observed data for warmer climate scenarios. Therefore, perfect prognosis models have difficulty predicting what warmer climate futures will look like.
Bias Adjustment
Bias adjustment is a post-processing technique that adjusts climate model outputs based on observational data. Using a calibration period, a bias or relative error for a specific climate feature, like precipitation, is calculated between model outputs and the observed values during the period. This bias is then applied for all modeling data to adjust the model outputs to be more realistic. Bias adjustments can be additive, multiplicative, or even variable based on quantiles. Multivariate bias adjustments also exist, where multiple climate factors are adjusted simultaneously. This type of bias adjustment has been shown to have the biggest impact, as climate features have dependency upon each other and adjustment of one can influence another. One concern with using bias adjustments is that this type of model assumes that biases are time invariant, which is not necessarily the case for all variables of interest, such as temperature. Therefore, bias adjustments may, in turn, need adjustments themselves as climate projections extend further into the future.
Weather Generators
Weather generators can be used to simulate statistical aspects of present-day daily temperature and precipitation and are calibrated for specific weather statistics, such as daily or sub-daily variability. Local weather characteristics for a singe location are realistically simulated using weather generators, but these models are less reliable when simulating multiple sites.
Fitness of Climate Models
One important aspect when evaluating climate models is whether the model is fit to simulate future regional climate. This idea of “fitness” can be evaluated by looking at how well climate models represent the present and past conditions of a specific region to build confidence in the model’s future climate projections for that region. Up until this point, the discussion of model performance has been looking at specific variables of interest, such as temperature and precipitation. However, when evaluating whether a climate model can accurately predict regional climate, it is more important to examine how the model represents processes that control these variables. Evaluating the process is key, because even if variables appear to be well represented and match expected outcomes, this may be the result of underlying process errors that happen to result in correct variable outcomes.
One method of determining model fitness is to evaluate the model performance of historical climate variability and long-term changes. Examining the model’s ability to reproduce historic trends can be an indicator to how the model will perform for future projections, but there are also limitations to this, such as observational uncertainty, as discussed in IPCC Section 10.2.
IPCC Chapter 10: Using Models for Constructing Regional Climate Information, Part 3
Julia Smith
10.5 — Combining Approaches to Constructing Regional Information
Understanding and communicating regional climate change is important because many of the specific impacts of climate change are going to occur at the regional level and decisions about how to handle these challenges are often made on the regional level. The content (i.e., what sources are selected, what information is deemed relevant, how to convey assumptions underlying the models and degrees of uncertainty) and form (i.e. what level of detail, whether it is conveyed through a storyline or more numerically) of the information needs to be tailored to the relevant stakeholders, and the best way to do this is often in partnership with the data’s intended users. The IPCC outlines the two-part process of “distilling” climate information: (1) constructing information that is backed up scientifically (this can be done with the user in mind) and (2) translating this information into the context of the specific purpose for which users need this information.
“Co-production of knowledge” refers to the collaborative involvement of users and producers (that is, non-climate-scientist stakeholders and climate scientists) in all parts of the process (research design through interpretation), and it is considered beneficial in the distillation of climate information for specific applications. An open, bi- or multi- directional flow of communication between decision-makers and climate scientists helps ensure that informed decisions are made at the regional level. Engaging users and producers in a variety of disciplines and considering a diverse set of values is also a good idea. When climate scientists understand the various goals, interests, and constraints (stressors other than climate impacts, i.e., economic) of decision-makers, involve decision-makers in their process, and build trust, their own perspectives and the interests of the stakeholders are more likely to be heard.
There is also the concept of “climate services” (10.5.4), which is closely related to the distillation of climate information, but it emphasizes setting new research challenges informed by what decision-makers will likely need to know in the future and a future where the information is proactively there for policy-makers.
Information construction
As the title of 10.5 suggests, a vital part of information construction is offering a comprehensive assessment of different sources of information, perhaps discounting some that are unfit for the context at hand, and integrating the others remaining sources into a whole. Many approaches for evaluation and integration have been explained in 10.2 and 10.3. An interesting point was that, in addition to more data or modeling focused approaches, integrating expert judgment (i.e. theory and experience found in the literature) was considered an approach. Several approaches can often be used in concert for better results.
Temporal and spatial scales of climate change and of decisions both have to be considered and mismatches between the scope of the climate processes and the scope of the decisions that need to be made can lead to poorly-informed choices. This is not to say that the scales always have to be the same: for example, speaking spatially, global climate dynamics can be relevant to more local decisions, but sometimes more resolved climate data is more relevant or a combination of global and local data is necessary. Additionally, there are decision-makers with different interests, scopes, and degrees of influence (i.e. individual farmer and agricultural ministry) and the match-up of spatial and temporal scales (as well as potentially the receptiveness to climate scientists and the approach climate scientists should take to conveying the relevant information) is likely to vary depending upon the decision-maker in question.
Translating climate information into the user context
Information can be provided to users with different levels of context-specificity, and underlying assumptions, best applications, and potential for misuse can be more or less well-explained. For example, often climate information is provided on a web portal that anyone could find and potentially misapply if they do not understand how to interpret or use it. Context-specific collaboration and co-production helps to prevent misunderstanding and ensure the relevance of the supplied information to the stakeholders. It is very important for climate scientists to understand the needs and context of the users and to be transparent about the uncertainties in their findings.
There are three main approaches to stakeholder interaction. “Top-down” frames global climate change as the driver of regional climate risk. “Bottom-up” starts with the local problem at hand according to the user and incorporates climatic and non-climatic stressors. “Interactive” is a combination of the first two approaches. The choice of approach will depend on the scale or scope of interest, and each approach has its pros and cons: bottom-up may be more tailored to a specific case, while top-down may be more widely applicable within a region.
10.6 — An illustrative example: Mediterranean Summer Warming
Figure 1. The Mediterranean Summer warming example links to concepts from other parts of the chapter. 10.2 discusses scarcity of observations and collecting observations in mountainous regions as challenges, and both apply to the Mediterranean. 10.3 describes downscaling, the process of going from a global climate model to a regional climate model, and this is one of the modeling techniques used in this example. 10.4 discusses attribution, and in the Mediterranean there is a lot of heterogeneity and the soil moisture feedback depends upon the location (North Africa does not have a feedback because it is so arid, while other regions have a positive feedback, contributing to warming).
The Mediterranean region is experiencing increasing summer temperatures, causing more common and more severe heatwaves and droughts. As climate change continues, this will lead to environmental and socio-economic stress for this region. The Mediterranean has a heterogenous climate, and climate variability is increasing over time.
There is uneven data availability for different parts of the Mediterranean region, with fewer and sparser data in the south, especially in the earlier years (pre-1970). This leads to uncertainty in different datasets, with up to 7°C of difference in some spots (especially in the mountains). This exemplifies two of the challenges for regional climate change assessment covered in 10.2.2: data scarcity (10.2.2.3) and collecting observations in mountain areas (10.2.2.5). Observational datasets agree when it comes to broader trends in the Mediterranean over time (1960-2014), finding a warming of more than 1°C since the late 19th century with an intensification of warming rate since the 1990’s, but at high spatial resolution there are some discrepancies, as previously mentioned.
There are several proposed mechanisms of enhanced Mediterranean warming, and the relative weight and interactions of these mechanisms is not fully resolved. Our chapter summary does not cover 10.4 in any great depth, but it should be noted that attribution is a challenge at regional scales mainly due to a greater contribution of internal variability than at global scales. The recent development of single-model initial-condition large ensembles (SMILEs) has helped to distinguish anthropogenic change from the “noise” of internal variability. The modeling approaches used to construct information were multi-model ensembles of global climate models (GCMs: CMIP5, CMIP6 and HighResMIP) and regional climate models (RCMs: CORDEX EUR-44 and EUR-11), and single-model initial-condition large ensembles (SMILEs: MIROC6, CSIRO-Mk3-6-0, MPI-ESM and d4PDF). The external forcings which have evidence of causing Mediterranean warming are thermodynamic and dynamic effects from the warm Atlantic Ocean, the reduction of aerosols, surface drying, and the difference in land and sea lapse rate.
It is interesting to note that these forcings occur at different scales and require different models/resolutions to capture. For example, as discussed in 10.3, soil moisture is best represented by RCMs, and indeed the soil-moisture feedback is relevant here (see Box 11.1) and accurately recapitulated by reanalysis-driven RCMs (see 10.2.1.2 for info on reanalysis). Meanwhile, the land-sea temperature contrast is supported by CMIP5 and CMIP6. There is also a discussion of biases of the various models. It is important to accurately model dynamics at different scales and understand both natural and anthropogenic forcings. For example, the summer North Atlantic oscillation (NAO) is a large-scale circulation pattern which must be modelled correctly or else we are likely to underestimate anthropogenic climate forcings (since the NAO works in opposition to them) in the western and central Mediterranean.
There were biases to each of the models. Interestingly, the different models ended up similarly underestimating the warming in certain locations — specifically part of North Africa, Italy, the Balkans, and Turkey — but for different reasons. The cold bias in GCMs was suspected to be due to systematic trends in simulated sea level pressure, soil-moisture, and cloud cover. For RCMs, potential culprits were aerosol modeling and error in sea surface temperature from the driving GCM (in the case of a coupled RCM).
Forecasts for future warming agreed broadly speaking, but at finer resolutions models had some disagreements about the degree of warming. Nevertheless, at that broad scale, there was high confidence in the warming that had already occurred and that more warming would occur (at a greater rate than the global average). Several equally plausible alternate storylines were proposed for projected summer warming in the Mediterranean. 10.6.4.6, Future climate information from global simulations, was reminiscent of the “climate services” concept. It emphasized the current knowledge gaps that will need to be filled. Climate scientists agree regional downscaling is important in order to make accurate subregional predictions given the heterogeneity of the Mediterranean.
Conclusion
While climate change is a global, existential problem which can often strike the general public as amorphous, its impacts will vary locally and are likely often going to be managed at that small scale. Thus local decision-makers will need regional information to make decisions. Climate scientists need to actively involve stakeholders in a dialogue in order to understand and meet their needs. Communication is very important when it comes to the use of regional data. On the research end, there are specific challenges when it comes to working with regional data that must be considered (and the uncertainties must be communicated to stakeholders). Some examples of these challenges and foci that are especially relevant to regional climate information are: dealing with sparse data and/or retrofitting with reanalysis, downscaling and propagating uncertainty from global models, attribution when internal variability plays a large role, and compiling multiple lines of evidence at the regional level.