Skip to main content

High-End cLimate Impacts and eXtremes

Final Report Summary - HELIX (High-End cLimate Impacts and eXtremes)

Executive Summary:
We produced a number of scenarios of 21st Century climate change using two high-resolution full-complexity Atmospheric General Circulation Models (GCMs), EC-Earth and HadGEM3 driven by sea surface temperature and sea ice from existing lower-resolution simulations in the 5th Coupled Model Intercomparison Project (CMIP5) multi-model ensemble. We used these to drive models of biophysical impacts of climate change such as fresh water availability, river flooding, crop yield and biodiversity loss, and also to calculate an index of vulnerability to food insecurity, at levels of global warming of 1.5°C, 2°C, 4°C and 6°C relative to the pre-industrial climate. We tested our impacts models against observations, performing simulations driven by observed climate data to assess the model performance in replicating observed trends and/or impacts of key climatic events. In turn, we used the impacts projections as inputs to an economic impacts assessment, and for comparison carried out additional economic impacts with alternative impacts projections and also with an Integrated Assessment Model (IAM).

We also assessed impacts at regional scales in Europe, northern hemisphere sub-Saharan Africa, using higher-resolution Regional Climate Models (RCMs) from the Coordinated Regional Downscaling Experiment (CORDEX) to provide additional regional detail for projections from some of the CMIP5 GCMs. These RCMs were again used to drive impacts models, for comparison with the coarser-resolution results from the global projections. We used an IAM to assess the implications of warming for energy demand in Africa under different socioeconomic scenarios. We have also investigated environmental factors influencing migration in a number of countries in Africa and Asia.

We also investigated potential impacts of passing tipping points in the climate system, such as a collapse of the Atlantic Meridional Overturning Circulation (AMOC), major loss of Arctic sea ice or the release of carbon from thawing permafrost.

We used simple climate models and IAMs to explore pathways to global warming levels of 1.5°C, 2°C, 4°C and 6°C, taking account of uncertainties in climate sensitivity, greenhouse gas mixes and feedback processes such as permafrost. We also developed an emulator (another simple climate model) that approximated the behaviour of complex climate and impacts models, to enable a wider range of scenarios to be examined.

Throughout the project we engaged closely with stakeholders from a range of backgrounds, including government departments, commercial companies, NGOs and local farmers. We held numerous workshops, policymaker briefings and public events in a number of countries including UK, Greece, Belgium, Germany, Senegal, Ethiopia, India and Bangladesh, and also at the European Commission, and took part in the United Nations Climate Conferences including the landmark conference in Paris in 2015. We were pro-active in contributing to a major report by the Intergovernmental Panel on Climate Change (IPCC) and the UK government’s 2nd Climate Change Risk Assessment.

Project Context and Objectives:
HELIX assesses the potential impacts of climate change at different levels of global warming in relation to those at 2°C. It compares the additional impacts at 4°C and 6°C global warming relative to those at 2°C, and also assesses the reduced level of impact that would occur if global warming is limited to 1.5°C. This is done at global scales and also in more detail in three focus regions: Europe, northern hemisphere sub-Saharan Africa, and South Asia. Pathways to reaching these warming levels are assessed, including the effect of different emissions scenarios and climate system processes. Tipping points in the climate system are also examined.

Policy context

These assessments will inform policy negotiations under the United Nations Framework Convention on Climate Change (UNFCCC) under the Paris Agreement. The Paris Agreement commits its signatories to holding global warming to well below 2°C and pursuing efforts to limit warming to 1.5°C, but the emissions pathways implied by the current Nationally Determined Contributions would still lead to warming well in excess of 2°C. HELIX results provide context for both the commitment to the original 2°C target and the increased ambition of 1.5°C. They will provide policymakers, businesses and other decision-makers with understanding of impacts at these different levels of global warming in order to inform decisions aimed at achieving the Paris targets and at planning ahead for adaptation to unavoidable change. This requires coherent information on the potential conditions which may need to be adapted to, and the consequences of different courses of adaptation action. Alongside this, ongoing international negotiations on limiting global warming also require clear information on the consequences of different levels of climate change.

While a vast array of projections, scenarios and estimates of future climate change and its impacts already exists, much of this is conflicting, unclear, of unknown levels of certainty and difficult to apply to inform decisions. The rate of future change will be a critical factor in the vulnerability or resilience of societies to the changing climate, because ongoing economic development will affect the sensitivity of societies to weather and climate, and adaptation measures will require time to be identified, planned and implemented. Interdependencies between different impacts, both biophysical and socio-economic, make the problem even more complex.

A further issue is that decisions relating to climate change often require information over a wide range of scales, from global to local. The information currently available is often inconsistent across these scales. Different methods are used for addressing different questions, and lack of consistency can lead to confusion and potentially exposes decision-makers to risks of poor decisions, either because incomplete information is available or because the available information is too varied and inconsistent to be useful.

HELIX is addressing these issues by developing a clear, coherent, internally-consistent view of a small, manageable number of “future worlds” under higher levels of global warming reached under a range of physical and socio-economic circumstances, including consideration of different adaptation scenarios, supported by advice on which aspects are more certain and which less certain. These are being delivered through ground-breaking scientific research across a range of physical, natural and social science disciplines, in close engagement with experienced users of climate change information to ensure appropriate focus, clarity and utility. Both the research and the engagement with users consider a range of scales from global to local, with internal consistency across the scales being a priority.

Scientific context

Previous state-of-the-art:
Previous attempts to quantify the global impacts of different levels of climate warming, such as the Stern Review and the IPCC 4th Assessment Report, had been limited by the fragmented, inconsistent and uncertain information upon which they relied. This made it difficult to establish a coherent picture of the consequences of climate change from global to local scales, limiting the usefulness of the scientific advice for decision-makers. A key inconsistency concerns the times at which warming levels are assumed to be reached, which vary widely according to the emissions scenario and climate model used. Moreover, lagged responses in processes such as sea level rise and large-scale ecosystem change can mean that the full impacts occur over a different timescale to the rate of warming, so there is not a one- to-one relationship between a particular level of warming and a particular magnitude of impact. Most impacts studies were local or regional and use many different scenarios and methodologies, so a consistent global picture is difficult to obtain. One critical area of interest is the implications of climate change in one region of the world for impacts in another region, acting through international linkages in trade, politics and the global economy. This is of very high interest to policymakers, but substantive information is lacking. Many studies focussed on specific sectors, neglecting between-sector interactions and giving an incoherent picture of the overall aggregate impacts, and preventing assessment of potential cascades of impacts through one process or sector to another. Some multi-sector studies provided consistency through use of common scenarios (ISI-MIP), but consideration of physical and economic interdependencies between different impacts remained lacking. Physical consistency between some impacts and the overlying climate was also recognized as a difficulty, as the need to correct biases in General Circulation Model (GCM) outputs leads to simulations of, for example, hydrological or ecological impacts that imply changes in the water cycle or energy balance which are different to those originally projected in the climate model. Neglecting these feedbacks may lead to over- or under-estimation of hydrological, ecological or agricultural impacts. While some global-scale multi-sector assessments of certain impacts such as river flows, droughts, agriculture, ecosystems and health do exist (AVOID, ISI-MIP), global socio-economic studies remained rare, especially in the context of linkages to GCM-based climate projections. Techniques from Regional Climate Model (RCM)-based bottom-up regional economic assessments (PESETA) had not yet been applied globally. Integrated Assessment Models (IAMs) provide an internally-consistent, holistic view of the climate and global economy and are used to generate emissions and climate forcing scenarios such as the Representative Concentration Pathways (RCPs); these necessarily involve substantial approximations, which can lead to inconsistencies with GCM-based studies and more detailed “bottom-up” process-based impacts assessments driven by the IAM-based scenarios. In particular, the implications of climate change for different energy futures and associated emissions / mitigation scenarios receive little detailed attention. Changing weather extremes are of critical concern, but previous GCMs are too coarse to adequately capture such events. Although RCMs use higher resolution, these alone cannot always rectify the as the underlying cause often lies upstream of the RCM domain. Due to uncertainties in climate sensitivity, the CO2 concentration at a given level of warming and hence its effects on plant (including crop) growth, water use and hydrology remains poorly-constrained. Potential tipping points in the climate and socio-economic systems have begun to be studied, but this work had yet to be integrated with mainstream climate impacts assessments. All these limitations hamper the effectiveness of adaptation decisions.

Advancements delivered by HELIX:
HELIX assists decision-makers in making the climate adaptation problem more tractable, by providing a manageable set of credible, coherent, global and regional views of different worlds at 4ºC, 6ºC and 2ºC, capturing levels of certainty in a clear and transparent manner. We used the existing ISI-MIP projections based on the 5th Coupled Model Intercomparison Project (CMIP5) and the RCPs to quantify ranges of global impacts on hydrology, ecosystems, agriculture, health and coastal infrastructure at these levels of global warming. We used these as inputs to socio-economic impacts models and indices using socio-economic scenarios from the new Shared Socio-economic Pathways (SSPs) in order to quantify the uncertainty in social and economic impacts arising from the uncertainties in climate projections, biophysical impacts responses and changes in exposure and vulnerability. Alongside this, we generated new climate change simulations at 1.5 ºC , 2ºC, 4ºC and 6ºC with high resolution GCMs which have been shown to produce improved representations of weather extremes, and used these to drive specialist hydrological models to assess the impacts on drought and flooding. We also used integrated land surface models to provide internally-consistent simulations of terrestrial impacts, additionally including the effects of glacier melt on river flows, and compared these with specialist sector- based models to assess the trade-offs between specialist process representation and integration. We examined the extent to which direct effects of CO2 may modify the climate impacts on crops, ecosystems and hydrology, and also compare with the simulations of these quantities directly within the GCMs to assess the implications of neglecting feedbacks compared to the need for bias correction. The new simulations of climate extremes and biophysical impacts were used for new global socio-economic assessments, examining economic-wide implications of climate change and identifying time- and path- dependencies of different adaptation options. These were compared with new IAM projections also updated to account for extremes. Our parallel RCM-based assessments in 3 focus regions (Europe, northern Sub-Saharan Africa, South Asia) provided regional detail, model validation and active stakeholder engagement. Potential tipping points in any of the physical- biogeochemical Earth System, natural ecosystem responses and/or socio-economic responses were assessed. Accompanied by a bespoke visualisation tool, the final product is a set of quantitative scenarios supported by a confidence assessment, which can be used by decision-makers to inform scenario planning of different adaptation and mitigation pathways.


The objectives of HELIX are:

1. Develop coherent, internally-consistent global scenarios of the combined natural and human world at 2, 4 and 6°C global warming, including reaching the 4°C level earlier (2060s) or later (after 2100) and with and without pro-active adaptation planning by society. The focus is on land and coastal impacts and their socio-economic consequences: food, water and energy security; coastal and river flooding; infrastructure; ecosystems and biodiversity; health; migration; and risk of conflict

2. Provide additional detail for focus regions of Europe, northern sub-Saharan Africa and South Asia, with active contribution from stakeholder groups and decision-makers in the regions.

3. Provide a reliable assessment of confidence in the different components of these scenarios, based on a comprehensive assessment of uncertainties throughout the different component of the projections

4. Carry out a further assessment of reduced impacts at 1.5°C global warming relative to 2°C, using methods consistent with our work on impacts of higher levels of warming.

5. To ensure the research addresses the needs of decision-makers, through both its implementation and
communication.

Project Results:

Introduction
The majority of previous climate change impacts assessments have tended to be framed in terms of future time horizons, eg. impacts by the middle or end of the 21st Century. However, with international climate policy now largely focussed on limiting warming to specific levels of global mean temperature such as 2°C or 1.5°C, policy-relevant climate impacts assessments increasingly need to be framed in terms of such warming levels.
Following the Paris Agreement, there are two major policy-relevant research questions concerning the impacts of climate change, which are relevant to both mitigation and adaptation policy areas:
What are the impacts at different levels of global warming? What impacts are avoided if global warming is limited to 2°C? What further impacts would be avoided by meeting the more ambitious target of limiting warming to 1.5°C? This is the primary question arising from the Paris Agreement and is relevant to mitigation policy, informing judgements and actions on holding the global temperature rise to “well below 2°C” and “pursuing efforts to limit the temperature increase to 1.5°C”
What is the range of regional climate conditions and related hydrological and ecological conditions could occur at a particular level of global warming, such as 2°C? This is relevant to adaptation policy and planning – exploring the possible outcomes for these levels of warming will help facilitate adaptation and improved resilience to account for a 1.5°C or 2°C world. It is recognised that many adaptation decisions require information on timing of specific impacts or risks, but nevertheless, framing regional impacts assessments in terms of associated global warming levels may help provide context of the levels of climate change that may be avoidable or unavoidable (and hence require adaptation).
A combination of the above questions is also relevant – how does the range of outcomes at, for example, 4°C compare to that at 2°C and 1.5C? This is also relevant to adaptation policy, as it can inform assessment on whether to adapt to potential impacts at 2°C or just 1.5°C. Putting in place adaptation measures to deal with potential impacts at 1.5°C and then increasing these to deal with 2°C later may be more expensive and difficult than adapting to potential risks at 2°C at the outset. On the other hand, since adaptation actions may themselves have consequences, unnecessary over-adaptation may have undesirable effects which it may be preferable to avoid or at least delay until absolutely necessary.
Both questions require an appropriate assessment of uncertainty. There are considerable uncertainties in projections of regional climate change, with different climate models projecting regional climate changes that can differ in magnitude or even, in the case of precipitation and impacts quantities strongly related to this, differ in sign. This may have important implications for regional impacts at specific levels of global warming.
A common approach to exploring and presenting such uncertainties is to examine the ensemble mean and the level of consensus amongst the ensemble members on the sign of the change. While this can often be useful in informing an assessment of the level of confidence in future projections, it may not always be sufficient to fully inform decisions. Risk assessment approaches require consideration of a range of possible risks, not just the most likely. This paper explores a range of regional climate states and related impacts that occur at global warming levels of 1.5°C and 2°C, and a range of climate sensitivities resulting in a range of timings and CO2 concentrations for these warming levels.
The pathway to reaching a particular level of global warming can be important both for the impacts of climate change and the ability for human society or biodiversity to adapt. The rate of warming depend on the emissions scenario, but also different climate models project different rates of warming for any given scenario of radiative forcing. This leads to uncertainties in the timing of reaching specific levels of global warming, and also uncertainties in the concentration of CO2 reached at a specific warming level. The uncertainty in CO2 concentration accompanying a specific level of global warming may be important for impacts involving direct responses to CO2, such as CO2 fertilization of plant productivity and ocean acidification. In a climate model with high climate sensitivity, a warming level such as 2°C is reached relatively early at a relatively low CO2 concentration. Conversely, in a model with low climate sensitivity, 2°C is reached relatively late and at a relatively high CO2 concentration. Moreover, since CO2 is not the only anthropogenic greenhouse gas (GHG) of concern, and different policy approaches may target the various GHGs differently, the relative contribution of CO2 and other greenhouse gases to the total radiative forcing is another factor that may need consideration. A given level of radiative forcing could be accompanied by stronger or weaker CO2 fertilization and ocean acidification effects, depending on whether CO2 is a larger or smaller component of the total GHG mix.

Methods and models
Defining global warming levels of 1.5°C, 2°C, 4°C and 6°C
There are a number of ways in which a global warming level such as 1.5°C or 2°C world could be defined – one could be the long-term climate state following a stabilisation of warming at that level, another could be the state over a shorter period around the time of first reaching that level. Here we choose the second definition, which is arguably more meaningful to human experience of climate and will be what is seen first and hence needs to be adapted to. We produced a new set of higher-resolutions transient climate and impacts simulations to assess global scale impacts at these resolutions for the first time. The advantage of using transient simulations is that additional global warming levels can be examined within the range. Indeed, the original scope of HELIX was to examine impacts of “high-end” climate change, defined as global warming above 2°C. Following the Paris Agreement, we were requested to consider whether we could carry out additional research focussing on 1.5°C. Our decision to use transient climate projections meant that this could be done without further climate simulations and therefore was feasible within the lifetime of the project.

New global climate simulations at 1.5°C, 2°C, 4°C and 6°C global warming
An important aspect of HELIX was provision of an improved representation of atmospheric and land surface processes including extremes by using new global climate models at higher spatial resolution than previous state-of-the-art in CMIP5. To explore uncertainties in the regional impacts of climate change, we carried out 13 simulations with higher-resolution global atmosphere models EC-Earth and HadGEM3 driven by sea surface temperature and sea ice projections from a subset of projections from the 5th Coupled Model Intercomparison Project (CMIP5) with the RCP8.5 scenario.
The EC-EARTH3 model is developed jointly by the EC-EARTH consortium that comprises a number of European Met Services and research institutes (http://www.ec-earth.org). The EC-EARTH3 model v3.1 is an updated version of EC-EARTH v2.3 that was used for CMIP5, and lies in the transition from the CMIP5 model towards the CMIP6 model.
The atmosphere module in EC-EARTH3 is based on IFS cy36r4 that is also used in the seasonal prediction system S4 at ECMWF. The main differences to v2.3 are the introduction of the McRad scheme for a better representation of the interaction between radiation and subgrid clouds, and an entirely new microphysics with 5 prognostic classes of hydrometeors. These two changes, in combination with an updated convection scheme have helped to greatly improve the representation of tropical precipitation. In addition to these updates a new humidity conservation was added to the EC-EARTH3 model. The humidity conservation has been backported from a later cycle of the IFS model to address the spurious sources and sinks of humidity that have been identified.
The standard resolution of the atmosphere in EC-EARTH3 increased from T159 in CMIP5 to T255 (planned for CMIP6). For the simulations for the HELIX project the high-resolution version of the model (EC-EARTH3-HR) was used with T511 which corresponds to 0.35 deg nominal horizontal resolution (~40 km). The vertical resolution increased from 62 to 91 levels with a higher model top and most of the new levels added in the stratosphere.
HadGEM3 (Hadley Centre Global Environment Model version 3) is a configuration of the UK Met Office Unified Model (MetUM) which has been developed for use for both climate research and weather prediction applications. It is the result of converging the development of the Met Office’s weather and climate global atmospheric model components so that, where possible, atmospheric processes are modelled or parameterised seamlessly across spatial resolutions and timescales.
The HadGEM3 family of climate models represents the third generation of HadGEM configurations, leading on from the HadGEM2 family of climate model configurations which was used for CMIP5. The HadGEM3 family comprises a range of specific model configurations incorporating different levels of complexity but with a common physical framework. It includes a coupled atmosphere-ocean configuration, with or without a vertical extension in the atmosphere to include a well-resolved stratosphere, and an Earth-System configuration which includes dynamic vegetation, ocean biology and atmospheric chemistry. A range of atmospheric resolutions is available. There is a choice of vertical resolutions between 38 levels extending to ~40km height, and 85 levels extending to ~85km in height (of which 50 are below 18km), the latter allowing improved representation of stratospheric processes. Horizontal resolutions vary between 2.5 degrees of latitude by 3.75 degrees of longitude and 0.556 degrees of latitude by 0.833 degrees of longitude, depending on the application.
The high-resolution simulations for HELIX have been performed using the HadGEM3-A Global Atmosphere (GA) 3.0 model at a resolution of N216 (~60km). This is the atmospheric component of the HadGEM3 GC2 coupled climate model. Key improvements over HadGEM2 include increased vertical levels in the atmosphere (85 compared to 38) and substantial changes to the model dynamics (ENDGame). As with EC-EARTH, this version of the HadGEM3 model lies in the transition from CMIP5 to CMIP6 versions. The Met Office is operationally running the coupled HadGEM3-GC2 model at N216 resolution for seasonal and decadal forecasting and clear benefits are emerging from this use at higher resolution.
HadGEM3 was run at N216 resolution (with gridboxes of approximately 60km length), and EC-Earth at T255 (approximately 40km resolution). These models used only their atmosphere and land components, with time-varying sea surface temperatures (SSTs) and sea ice concentrations (SICs) prescribed as input quantities. The SSTs and SICs were taken from a subset of the CMIP5 projections performed with the RCP8.5 scenario – the CMIP5 members were selected as representative of a range of outcomes for future climate change, including high and low climate sensitivity, different biases in baseline precipitation climatology, and different global patterns of precipitation change.
Our climate projections with EC-Earth and HadGEM3 were transient simulations, ie: they were driven with ongoing increases in GHG concentrations following a time-dependent scenario. Within each transient projection, we identified the times at which global mean temperature was at specific levels of warming relative to pre-industrial: 1.5°C, 2°C, 4°C and 6°C. Our simulations followed the RCP8.5 GHG concentration scenario, which gives a relatively rapid rate of warming, in order to allow the warming levels of 4°C and 6°C to be reached before the end of the simulations at 2100, which was the limit of the available driving SST data in some cases. Only a few projections reached 6°C by 2100, but most had passed 4°C.
Regional climate models – CORDEX
To allow more detailed focus on our particular regions of interest, and also provide a semi-independent source of climate projections, we used ensembles of regional climate models (RCMs) from the Coordinated Regional Downscaling Experiment (CORDEX) for Europe, Africa and Asia, at spatial resolutions ranging from 50km to 0.11km.
Integrated Assessment Models (IAMs)
IAMs aim to represent the entire climate system including impacts and anthropogenic emissions, using simplified models of the components of the system that interact with each other. We used the TIAM-UCL and FUND IAMs in HELIX.
Temperature and precipitation extremes and other weather-related metrics
Using climate model output, we calculated indices related to a number of characteristics of extreme weather and climate events considered to be of relevance of impacts. Specifically, these were:

Temperature of the hottest day of the year (TXx)
Changes in the percentage of days above the 90th percentile of daily maximum temperature in the 1981-2010 average (Tx90p)
Cumulative dry days (CDD)
Number of days with precipitation above 95th percentile in the 1981-2010 average
Number of days with precipitation below 20th percentile in the 1981-2010 average
Annual maximum 5-day rainfall (Rx5day)

We calculated these from the six HadGEM3 simulations.

Freshwater resources (global)
Impacts on fresh water were assessed with a version of the JULES land surface model, a coupled ecosystem-hydrology-surface exchange model which simulates land-atmosphere fluxes of water, energy and carbon in an internally-consistent way, typically applied at global scales. Variants of JULES form the land surface scheme of Met Office Hadley Centre Earth System Models and have been used to assess impacts of climate change on global terrestrial ecosystems and hydrology within such models. JULES can also be used outside of the ESM, driven by meteorological outputs of other ESMs to assess impacts of a wider range of climate projections. Here we use a new, higher-resolution configuration of JULES on a global grid of 0.5° resolution.
It has been noted that hydrological impacts models driven by climate change projections from climate models tend to give more severe drying than simulated in the climate models themselves. This is largely attributed to the inclusion of plant stomatal closure in response to elevated CO2 in the climate model land surface schemes, which generally reduces evapotranspiration relative to climate projections without this process and hence further increases runoff / streamflow or ameliorates decreases [36]. This process is often omitted from standard hydrological models. Plant physiological responses to CO2 are included in the JULES model, so our projections of changes in runoff here do account for this process.
We used each HadGEM3 and EC-Earth simulation to drive JULES to simulate changes in runoff due to the effects of climate change and CO2 rise on precipitation, evaporation and transpiration. We analysed thirty-year periods centred around the year of crossing global warming levels of 1.5°C and 2°C relative to pre-industrial. We examined changes in both mean flows and low flows (defined as the flows for the lowest 10% of time).
We performed additional analysis using a second hydrological model, Waterworld. This is another process-based global hydrological model.
River flooding: global
We use the LISFLOOD model to assess the impacts of climate change on river flooding risk. This model uses locally-specific damage functions to relate projected streamflow to socioeconomic impacts such as the population affected and economic damages. Flooding risks are quantified here in terms of population affected and economic damages, using present-day socioeconomic conditions. This therefore does not consider either ongoing population or economic changes or adaptation, so is not a prediction of actual impacts but rather an indication of exposure of current populations and infrastructure. Moreover, the simulations do not account for possible changes in direct human intervention, such as water extraction and dams.
Hydrological modelling: catchment scale
The Soil and Water Assessment Tool (SWAT) model (Arnold et al., 1998, Neitsch et al., 2002) is a distributed parameter and continuous time simulation model. The SWAT model has been developed to predict the hydrological response of un-gauged catchments to natural inputs as well as the manmade interventions. Water and sediment yields can be assessed as well as water quality. The model (a) is physically based; (b) uses readily available inputs; (c) is computationally efficient to operate and (d) is continuous time and capable of simulating long periods for computing the effects of management changes. In SWAT, a watershed is divided into multiple sub-watersheds, which are then further subdivided into unique soil/land-use characteristics called hydrologic response units (HRUs). The water balance of each HRU in SWAT is represented by four storage volumes: snow, soil profile (0-2m), shallow aquifer (typically 2-20m), and deep aquifer (>20m). Flow generation, sediment yield, and non-point-source loadings from each HRU in a sub-watershed are summed, and the resulting loads are routed through channels, ponds, and/or reservoirs to the watershed outlet. Hydrologic processes are based on a water balance equation.
Coastal flooding
Delft3D-FLOW is a multi-dimensional (2D or 3D) hydrodynamic (and transport) simulation program which calculates unsteady flow and transport phenomena that result from tidal and meteorological forcing on a rectilinear or a curvilinear, boundary fitted grid. In 3D simulations, the vertical grid is defined following the σ co-ordinate approach. The Delft3D-FLOW model solves the Navier Stokes equations for an incompressible fluid, under the shallow water and the Boussinesq assumptions. Delft3D-FLOW solves the Navier Stokes equations for an incompressible fluid, under the shallow water and the Boussinesq assumptions. In the vertical momentum equation the vertical accelerations are neglected, which leads to the hydrostatic pressure equation. In 3D models the vertical velocities are computed from the continuity equation. The set of partial differential equations in combination with an appropriate set of initial and boundary conditions is solved on a finite difference grid. The time step for the hydrodynamic computation is selected equal to 5 min. Turbulence effects are computed by means of the K-epsilon model. Horizontal background eddy viscosity and diffusivity are set equal to 1 m^2/s.
An important aspect of the coastal flooding work was that the model experimental design was different. It was not appropriate to frame results in term of levels of global warming, because of the very long lag between warming and sea level rise – a particular global warming level could be associated with a range of sea level rises depending on the time horizon. Instead, coastal impacts were examined at specific values of sea level rise; 0.5m 1m and 1.5m. The former is approximately the upper bound of the rise projected to occur under a low emissions scenario limited to 2°C warming by the end of the century, but also near the lower bound of a high-emissions scenario. High emissions could lead to 1m or 1.5m by 2100.

Terrestrial biosphere impacts: gross primary productivity and forest cover
ORCHIDEE is a process-driven Dynamic Global Vegetation Model (DGVM) designed to simulate C and water cycle from site-level to global scale. It is composed of two main modules. SECHIBA computes the energy and hydrology budget on a half-hourly basis, together with photosynthesis based on enzyme kinetics [41]. These results are fed to a module called STOMATE, which simulates C dynamics on a daily basis: gross primary production (GPP) is allocated to different organs, and then respired by the plant or by soil microorganisms when parts of the plant die. These processes determine several ecosystem state variables such as leaf area index (LAI) and canopy roughness, which are fed back to SECHIBA because they control the energy and water budgets. The equations of ORCHIDEE can be found at http://orchidee.ipsl.jussieu.fr/. As most DGVMs, the vegetation is described into a discrete number (13) of PFTs over the globe. For grassland, C3 and C4 grass are included, and treated like unmanaged natural systems, where C / water fluxes are only subject to atmospheric CO2 and climate change. Recently, a more complex 11-layer soil-water diffusion scheme; and the mechanistic intermediate-complexity snow scheme (ISBA-ES; have been implemented in ORCHIDEE Trunk.rev3623 which improves the representation of water and snow-related processes. This version of ORCHIDEE includes dynamic vegetation scheme, which allows the vegetation distribution to respond to climate change and rising CO2 concentrations.
Hunger and climate vulnerability index
The Hunger and Climate Vulnerability Index (HCVI; Krishnamurthy et al., 2014) was developed to provide a country-level assessment of vulnerability to food insecurity as a result of climate-related events. The latest iteration of the HCVI makes use of gridded climate model projections to understand the impact of climate change on vulnerability to food insecurity, and also the benefits adaptation can bring via scenarios of adaptation investment (Richardson et al., 2017). This iteration of the HCVI only considers in-country production of food and does not account for food trade. For this reason, the HCVI is only calculated for developing and least-developed countries (defined here as countries not in the OECD or EU).

The index provides quantification at the national level across the globe of the scale and direction of impact of climate change on food insecurity. As such, it aims to provide the following: 1) information to help policymakers understand the level of challenge to global food security that climate change presents; 2) information on the geography of the impacts and help to evaluate the relative benefits of mitigation and adaptation responses.
The index is not intended to be a detailed planning tool, but aims to help planners evaluate the nature of the top-level threat to food insecurity that climate change presents, thereby supporting prioritisation of effort.

The HCVI consists of three equally-weighted components: exposure to climate-related hazards, sensitivity of national agricultural production to climate-related hazards, and adaptive capacity – a measure of a country’s ability to cope with climate-related food shocks. The exposure component is comprised of proxies for the average length of flood and drought events calculated with daily precipitation data. The impacts of projected climate change therefore act through changes in these quantities. In the current version of the HCVI, climate change impacts on other quantities such as crop yield are not considered. Socioeconomic factors affecting sensitivity and adaptive capacity are fixed at present-day conditions.

Crop model
A statistical model was developed linking crop yield with climatic conditions, for major global crops. These models (ClimaCrop) are a variation of the models developed by Schlenker and Lobell (2010) which have successfully been used to analyse the effects of climate change on crop yield in Africa. Here the models were extended to the global scale and a linear mixed model (LMM) approach was used instead of the linear fixed model approach employed by Schlenker and Lobell.
The models assume that the natural logarithm of crop yield is a function of growing season temperature, precipitation and time where the relationships are considered to be both linear and quadratic. It is thus assumed that there exist optimal values for both temperature and precipitation and that deviations from this optimum will reduce yield. By including a quadratic dependency on time the model accounts for technological advance (Schlenker & Lobell, 2010) and the mixed model approach accounts of the inter-dependency of entries for the same country.

Biodiversity model
The impact model used for biodiversity is that used in the Wallace Initiative and previously published (Warren et al 2013). For this study we used newly developed models of ~80,000 plants, birds, mammals, reptiles and amphibians for multiple dispersal scenarios at a spatial resolution of 20km x 20km. Primary biodiversity data were obtained from the Global Biodiversity Information Facility (GBIF). Bioclimatic models were then developed using the program MaxEnt using a limited set of eight climate variables in order to minimize potential issues with autocorrelation and to prevent ‘overfitting’ of the MaxEnt modelled species distributions.

Economic model
The economic simulations have been performed with the GEM-E3-CAGE Computable General Equilibrium (CGE) model. CGE models (e.g. Shoven and Whalley, 1992) combine a high resolution dataset of the economy (the Social Account Matrix, SAM) with standard microeconomic theory.
CGE models are multi-agent, multi-sector and multi-country, features which make them ideal for considering how the overall economy (all agents and markets) would adjust to an external shock like climate change, considering the second and higher order round effects of the primary climate shock, on top of the direct damage or primary climate shock. A CGE model is in equilibrium when all agents are at their optimum and all factor, goods and services markets are simultaneously cleared (i.e. Walras' Law is holding). The market adjustments captured by a CGE model can be interpreted as market or price-driven adaptation (e.g. OECD, 2015).
The two main agents of the economy are households and firms, whose endogenous behaviour is simulated, assuming they optimize their objective function (utility or welfare for households and profits for firms) subject to a set of constraints (e.g. technology, costs, prices). The modelling setting also considers the public sector (usually exogenous) and the external sector (modelling trade as a function of relative prices).
The model employed for this analysis, GEM-E3-CAGE, is a static multi-country, multi-sector computable general equilibrium model of the world economy linking the economies through endogenous bilateral trade. The CGE analysis of climate impacts follows a static comparative approach (as in e.g. Aaheim et al., 2012; Hertel et al. 2010; and Ciscar et al. 2012), estimating the counterfactual of future climate change (reaching SWLs) occurring under the current socioeconomic conditions. Therefore, the climate shock-induced changes would occur in the economy as of today.

There are three main channels through which the direct damages as computed by the biophysical impacts affect the economy. Two of the channels would affect the supply side of the economy and a third one the demand side.

Regarding the supply side, firstly, climate change can affect the productivity of the economy. The productivity is defined as the unit of output per unit of input. The clearest case is that of agriculture: climate change can lead to reduced yields (output), while all other factors of production (inputs) are the same. Secondly, climate change can alter the capital stock of the economy, for instance when floods damage infrastructure. These supply-side effects would trigger a series of adjustments in the economy also indirectly affecting sectors and regions different to that where climate change is impacting directly.

Regarding the demand side of the economy, climate change can also influence consumption decisions. For instance, damage to residential buildings due to a flood leads to a change in the consumption behaviour of households as they would repair the damage and consequently reduce other consumption expenditures. There would be a substitution of consumption: additional consumption to repair the dwellings damage (e.g. buy a new fridge) and an equivalent reduction in consumption (e.g. less travelling), keeping the overall consumption constant. As the reparation of the flood damage is part of obliged or compulsory consumption (which would not occur in the absence of climate change), the economic model interprets that there is a welfare loss associated to the damage to residential buildings.

Results
We present our results framed in two ways, each addressing one of the questions discussed in the introduction. (i) For a world at 2°C global warming, we present a range of outcomes to provide insight into the level of agreement between models for a particular projected change, and hence an indication of potential robustness of the projected changes for informing adaptation.
(ii) For levels of global warming below and above 2°C, specifically 1.5°C and 4°C (and in some cases 6°C), we quantify the difference from 2°C in order to investigate the level of impact that would be avoided by limiting global warming to different levels. Bearing in mind the uncertainty in regional climate outcomes, we address this in a number of ways. For individual realisations, we compare the impacts at different warming levels to see if they are systematically smaller at 1.5°C and larger at 4°C in relation to 2°C, even if the sign of the change is uncertain. We also compare the range of outcomes at different global warming levels, to see if the regional-scale uncertainty itself increases with global warming.
Climate change impacts and extremes at 2°C global warming
Temperature extremes and heat stress:
For 2°C global warming, the ensemble mean increase in annual daily maximum temperature was above 2°C for most of the land surface, with the exception of the Indian sub-continent, most of Australia and Antarctica. The increase was higher still in many regions; most North America, much of China and North Asia, north-western South America and all of Europe. In the northern and eastern USA and much of northern and western Europe, the annual daily maximum temperature increased by over 4°C for 2°C global warming.

These results for TXx can be compared with those from the Inter-Sectoral Impacts Model Intercomparison Project (ISIMIP) [9] which used a subset of CMIP5 models and hence are at lower spatial resolution than the models used here. It is notable that our results show more geographical variation than those from the ISIMIP projections
The different ensemble members give somewhat different results at regional scales, although there is a strong consensus on the temperature extremes examined here becoming warmer. In the simulations driven by SSTs and SICs from the two IPSL CMIP5 models, most of the global land surface sees an increase in annual daily maximum temperature which is similar to the global annual mean temperature increase. This is illustrated in Figure 3 for 1.5°C global warming – similar results are seen for 2°C. In the IPSL-driven simulations, increases in TXx substantially larger than the global warming level are confined to the eastern USA, Europe and part of north-east Asia. In contrast, the GFDL-driven simulation shows much of the global land surface seeing increases in annual daily maximum temperature larger than the global mean warming. The very largest increases are seen in central North America, Europe and north-western Asia.
Wet bulb globe temperature (WBGT), an indicator of heat stress that includes effects of temperature, humidity and solar radiation, is also projected to increase. A WBGT of 32°C is a widely-used industry standard threshold for high risk of heat stress, and on average is reached between 5 and 20% of summer days in some parts of North Africa and the Indian subcontinent in a simulation of present-day climate. At 2°C global warming, this threshold is projected to be reached over a slightly wider area for 5 to 20% of summer days, and up to 50% of the time in a few locations.
Precipitation extremes – heavy rainfall and drought:
Indices based upon daily precipitation show more spatially variability in changes than the temperature-based indices. The length of dry spells is simulated to increase over some regions and decrease in others. Southern Africa, the Mediterranean, Australia and and NE South America are projected to have increased dry spell lengths, while central and eastern Asia are projected to have shorter dry spells. This ensemble mean result is broadly consistent with the median of the ISIMIP-Fast Track ensemble, although there are some differences. Our ensemble mean suggests shorter dry spells in the central Amazon, whereas ISIMIP indicated longer dry spells. Also, as with the temperature indices, our results show greater geographical differentiation in the intensity of changes.

In the ensemble mean, maximum 5-day rainfall (Rx5day) is projected to increase over regions including parts of southeast Asia, southern Brazil, northern Australia and the east coast of the US (Figure 6). However, some regions (particularly the central Amazon and the northern coast of South America, are projected to see a decrease in Rx5day. The decrease in Rx5day in some regions here contrasts with the ISIMIP projections [9] which suggested an increase in Rx5day almost everywhere where at least 66% of the model ensemble agreed on the sign of the change, including all of northern South America.
The inter-model variation in regional outcomes is greater for precipitation-related indices than those related to temperature, although there is still some consistency between simulations in some regions. Large increases in Rx5day are simulated in South and South-East Asia in all models, but with local details varying. South-eastern South America (broadly southern Brazil and northern Argentina) also see large increases in Rx5day in all models. All models show only small changes over central and north Africa, Europe and most of Asia. In northern South America, however, some models show increases in Rx5day but others show decreases. This suggests that the ensemble mean result of a decrease in Rx5day in this area may be subject to large uncertainty. Inter-model variations in the sign of changes are seen in a few other local localised regions.
Length of flood and drought events:
The length of average flood events (defined as precipitation above the 95th percentile of the baseline) generally increased over most of the land surface, although this increase was mostly by a day or less. However, some areas were projected to see an increase in flood event lengths of 4 days or more, particularly India and Bangladesh, for which such increases were projected in all ensemble members to some extent. Increases of 2 to 4 days were also projected in parts of Brazil by all ensemble members, although the magnitude and location within the country varied between members. Similar increases were projected in the region of the Horn of Africa and souther Arabian Peninsula in several members.
Projected changes in the length of average drought events (defined as precipitation below the 10th percentile of the baseline) were more variable, both geographically and between ensemble members. All members projected increased drought length in southern Africa but decreased drought length in the Sahel region. Longer drought was projected in varying parts of Brazil in most members, while shorter drought was projected in most or all of India in all members. Most of China was projected to see longer drought, except for in the south, and most of Russia was projected to see longer drought, except for in the centre or west in some ensemble members. In most of the Middle East, projections ranged between increases and decreases of several days, with a similar range of results for Australia.
Fresh water:
The JULES simulations driven by the EC-Earth and HadGEM3 ensemble show a general tendency towards increased streamflows over the majority of the global land surface. The ensemble mean change in mean streamflow shows an increase of between 5 and 25% over most of the northern hemisphere land surface, with some regions seeing an increase of over 50% at 2°C global waming. Notable exceptions to this are western Europe and south-central USA, which see less than a 5% change in streamflow.

Ensemble mean projected changes in low streamflow are generally larger, with the regions seeing an increase in mean streamflow seeing a larger percentage increase in low runoff – over 75% increases over much of North America, Eastern Europe and Asia. Note that this does not necessarily imply a larger increase in absolute low flow compared to absolute mean flow, since the baseline is (by definition) smaller for low flows. In western Europe, where the changes in mean flows were less than 5%, the ensemble mean low flow decreases by between 5 and 75%, especially in the Iberian Peninsula. Southern Africa also sees a decrease in low flows where changes in mean flows were small. Changes in high runoff show similar patterns and magnitudes to those in mean runoff.

The simulated changes in both mean and low runoff flows show substantial differences among the 13 simulations. In most basins examined here, the range of outcomes include both increases and decreases in mean and low flows for any particular basin, but generally with the largest proportion simulating increases in both mean and low flows. In a few cases, notably the Lena in north-east Asia and Ganges in south-east Asia, the ensemble agreed entirely or almost entirely on increased flows.
Exceptions to the general picture of consensus on increasing flows are seen in the Amazon, Orange and Murray basins where the range of model outcomes includes a greater proportion of decreased flows. The signal of decreased flows was stronger for low flows than mean flows, and indeed in the Niger, the range of mean flow changes extended more towards increases whereas the range of low flow changes extended more towards decreases.
River flooding:
We used the EC-Earth-driven LISFLOOD simulations to assess projected impacts at the scale of major river basins. We examined the 20 most significant global rivers in terms of their flooding impact on population and economic damage respectively. In most cases, the ensemble means of both population affected and economic damages increase with the level of global warming – they are higher at 2°C compared to the baseline. Exceptions to this include the Congo, Niger and Zambezi in Africa and Dniepr in Asia, which show slight reductions in ensemble mean population affected at 2°C.
However, the picture is more nuanced when considering the ensemble range, ie: the full set of possibilities as represented by the set of models used here. Generally the range is larger at 2°C compared to the baseline (NB the baseline is still an ensemble of simulations, not observations). For more than half of the top 20 rivers for population impacts, the ranges of future populations affected include both decreases and increases in impacts relative to the baseline. This illustrates the large uncertainty in future climate change impacts relating to quantities such as precipitation change, which can change in different directions in different models.
However, in many other basins the range of future impacts does not span the baseline range, and in these cases the minimum of the future impact range is above either the minimum or ensemble mean of the baseline range. There are no cases where the future range maximum for 2°C is below the baseline maximum for population affected, and only one (the Niger) for economic damages. The range of flood risks therefore tends to shift towards higher risks under these global warming levels.
Hunger and climate vulnerability index:
The ensemble mean HCVI increased in nearly all assessed countries. The greatest increase was in Oman, followed by India, Bangladesh and Saudi Arabia, then Brazil and a number of its neighbouring countries. Smaller increases in HCVI were seen across Africa. South-eastern Africa showed larger increases than Central Africa. The HCVI decreased in three countries: Mali, Burkino Faso and Sudan.
The ensemble members showed broadly consistent changes in HCVI at 2°C global warming, with increases in most assessed countries and generally similar sets of countries experiencing the largest and smallest changes. South-eastern Africa consistently showed larger increases in HCVI than Central Africa, due to increased length of drought events projected in all ensemble members. Length of flood events was not projected to increase in this region. The Sahel region consistently showed one or more countries with a small decrease HCVI, although the precise country or countries varied between ensemble members. The decrease in HCVI here was due to projected decreases in length of drought, with length of flood events projected to change little.
India is projected to see increased HCVI by all ensemble members, due to a consistent increase in length of flood events projected in all members, outweighing the beneficial impact of decreased length of drought which is again projected in all members.
Brazil is projected to see increased HCVI, but for reasons which vary between ensemble members. Although the location of projected longer flood events varies across the country in different members, the aggregation of the HCVI to the country level renders this geographical variability irrelevant for such a large country. Some ensemble member project longer drought for Brazil, which again contributed to increased HCVI.
Using the HCVI and current population data, approximately 5.73 billion people could become more vulnerable to food insecurity in a 2°C warmer world. Of these, an estimated 189 million people are projected to become more vulnerable to food insecurity than is currently experienced anywhere under the current climate.
Land ecosystems and gross primary productivity:
Three of the above EC-Earth and HadGEM3 simulations were also used to drive a global vegetation model, ORCHIDEE, to simulate changes in vegetation productivity due to climate change and the associated CO2 rise. Gross primary productivity was projected to decrease in some marginal lands such as the Sahel and the Middle East, but increase in many other areas, especially at 2°C. Forest cover was projected to increase in many tundra regions and some drylands, again especially at 2°C, but also small areas of decreased forest cover are also projected in some regions.
A key difference between individual ensemble members was that the simulation using SSTs from a GCM with relatively low climate sensitivity (GISS-E2-H) resulted in larger increases in global gross primary productivity than with SSTs from a high climate sensitivity model (IPSL-CM5A-LR). This seems likely to be due to the higher concentrations of CO2 at any particular SWL (such as 2°C) under GISS climate than IPSL. This GISS-driven model climate is also wetter than that in IPSL-CM5A-LR so the vegetation may also be less water-limited and hence better able to respond to warming temperatures and rising CO2. The longer timescale of reaching the SWLs may also be a factor, as the dynamic vegetation is allowed more time to respond.
Coastal flooding in Bangladesh:
If a 0.5m sea level rise were to occur, this could lead to an inundation of 2000km2 or 1.6% of the country. This would affect 2.5 million people at present-day population levels and distribution. If a tropical cyclone similar to cyclone Sidr of 2007 were to occur in addition to this 0.5m sea level rise, the inundation would be 3380 km2 affecting 4.1 million people (more than double the number affected in 2007).

Comparison of 2C and 4C: the benefits of avoiding high-end climate change
A major result from our projections at 4°C global warming concerns human heat stress. All of the tropics and some of the lower mid-latitudes (southern USA and southern Europe) experience WBGT above the critical threshold of 32°C for at least 5% of summer days. The threshold is projected to be exceeded for at least 50% of summer days over widespread areas of the tropics, and at least 80% of summer days in parts of North Africa, the Indian subcontinent and northern Australia.
On river flooding, our projections suggest that at 4°C global warming, countries representing 73% of the world population and 79% of the global GDP would
experience increasing flood risk at an average 580% increase in population affected and 500% increase in damage, as compared to the impact simulated over the baseline period 1976-2005. These figures reduce to a 170% increase in population affected and 120% 170% increase in damage for a warming level of 2°C.

At 4°C global warming, approximately 5.76 billion people are projected to be more vulnerable to food insecurity than at present day, compared to the additional 5.73 billion calculated for 2°C. An estimated 1.8 billion people would potentially be in unprecedented levels of vulnerability at 4°C global warming, compared to 189 million at 2°C.
Biodiversity impacts are projected to be considerably more substantial at 4°C global warming than 2°C. Even if dispersal is considered as a potential adaptation, the percentage of species remaining at 4°C global warming are projected as 49% (plants), 66% (reptiles), 66% (amphibians) and 93% (mammals). 100% of bird species are projected to remain, due to the high ability to disperse. At 2°C global warming, these percentages increase to 71% (plants), 81% (reptiles), 81% (amphibians) and 100% (mammals). The percentages become smaller at 6°C global warming for all groups, including birds for which 86% of species are projected to remain.
The global increase in vegetation carbon and productivity were larger at 4°C compared to 2°C. The differences were more marked in climate simulations with low climate sensitivity (a slower rate of warming for a given CO2 rise), because the relative impact of CO2 on vegetation growth was larger.
If a .5m sea level rise were to occur, this could lead to an inundation of 5300km2 or 5.1% of the country. This would affect 8 million people at present-day population levels and distribution – more than three times the number affected by a 0.5m sea level rise. If a tropical cyclone similar to cyclone Sidr of 2007 were to occur in addition to this 1.5m sea level rise, the inundation would be 7588 km2 affecting 9.1 million people.

Comparison of 1.5C and 2C: the impact of more ambitious mitigation under the Paris Agreement
The aim of comparing impacts at 1.5°C and 2°C is to assess whether the impacts are smaller when global warming is limited to 1.5°C. In many respects this question can be answered by comparing ensemble mean changes at the two levels of global warming relative to the baseline – in locations where there is strong consensus between the models on the sign of impacts, then comparing the ensemble mean impacts will represent the difference in impacts for most or all individual members. However, in locations where these is disagreement between the sign of the change, such as river flow in many river basins, a comparison of ensemble means may not give the full picture. In this cast it is worthwhile making the comparison of impacts at the two global warming levels for individual ensemble members.
At 1.5°C global warming, the ensemble mean TXx increases with similar geographical patterns as for 2°C, with larger changes in continental interiors especially in the mid-latitudes, but with local changes consistently smaller in magnitude than at 2°C. Over parts of Europe, where annual maximum daily temperature was projected to increase by over 5°C for a 2°C global warming, the local increase is limited to 3-4°C for 1.5°C global warming.

The percentage of days exceeding the 90th percentile of daily temperature (Tx90p) also increases less at 1.5°C global warming than at 2°C, again retaining a similar geographical pattern of greater increases in the tropics than mid-latitudes. In many tropical regions, the ensemble mean Tx90p increase was limited to around 20 – 30% at 1.5°C global warming compared to 50% or more at 2°C (Figure XX).

For both the above temperature extremes metrics, the geographical patterns of warming for individual ensemble members remained similar at 1.5°C and 2°C global warming. (Supplementary information).
Precip...Generally similar changes are seen at 1.5°C global warming as at 2°C, but smaller in magnitude, suggesting that most of these changes are a response to radiatively-forced climate change as opposed to internal climate variability. However, some localised changes do vary in sign between the global warming levels, such as in South Australia, suggesting a possible dominance of internal variability over the global warming signal in these places.
Where Rx5day increases, the increases are projected to be larger – in some cases approximately double - at 2°C global warming than 1.5°C. Where Rx5day decreases, again the decreases are projected to be larger at 2°C global warming than 1.5°C
For most countries assessed, the ensemble mean HCVI calculated at 1.5°C global warming is smaller than that at 2°C, indicating that the projected increase in vulnerability to food insecurity would be reduced by limiting global warming to 1.5°C. At 1.5°C, fewer countries move to HCVI states that are above any seen in the baseline, so the level of vulnerability to food insecurity remains more within current experience at 1.5°C than at 2.5°C.
For river flows, in the ensemble mean, changes in mean, low and high flows are generally larger at 2°C global warming compared to 1.5°C. (Figure xx) This is the case for both increases and decreases in flows – increasing the level of global warming magnifies the pattern of river flow changes. The range of projected changes is also generally larger for 2°C than 1.5°C, indicating greater uncertainty in changes in flows under greater climate warming. However, this was not always the case.
For river flooding, the ensemble means of both population affected and economic damages increase with the level of global warming – they are higher at 1.5°C compared to the baseline, and higher still at 2°C. Exceptions to this include the Congo, Niger, Nile and Zambezi in Africa – the Congo and Zambezi show slight reductions in ensemble mean population affected at 1.5°C global warming and again at 2°C, while the Niger and Nile show reduced population affected at 1.5°C and an increase when moving to 2°C. In Asia, the Dniepr and Amur see an increase in ensemble mean impact at 1.5°C and then a smaller decline on moving to 2°C.
For the ranges of projected outcomes, he differences between 2°C and 1.5°C are less clear, with the ranges of impacts often similar at the two global warming levels, and in some cases larger at 2°C and in other cases smaller at 2°C. This may be a consequence of internal variability contributing to the range maxima and minima.
Some example results for selected major rivers:
The Ganges-Brahmaputra has the largest population affected in the baseline – between approximately 1.5 and 2.5 million people per year. This rises to between 2 and 11 million at 1.5°C global warming and between 5 and 17 million per year at 2°C global warming.
The Yangtze sees the greatest expected damages in the baseline – between approximately 4 and 8 billion Euros per year. This increases to between 5 and 16 billion Euros at 1.5°C global warming and 7 to 20 billion Euros for 2°C.
In the Tigris & Euphrates, the range of expected damages expands from 0.2 to 1.2 billion Euros in the baseline to between 0 and 3 billion Euros at 2°C, with only a slightly smaller range at 1.5°C.
The above calculations use the Euro at 2007 values.

In the terrestrial biosphere, the global increase in vegetation carbon and productivity was smaller at 1.5°C compared to 2°C. Biodiversity loss was smaller: with dispersal, the percentages of species remaining at 1.5°C were: 78% (plants), 85% (reptiles) and 85% (amphibians), all smaller than at 2°C.

Economic impacts
The global GDP reduction is relatively similar for the 1.5oC and 2oC warming (0.18% and 0.22% loss), but significantly larger at 4.0oC warming: 0.65% loss. In money-metric the losses account to 114, 137 and 396 bn€, respectively.

The welfare losses are larger by factor of 2 to 3. The reason of larger welfare losses results from the fact that some damages (eg damage to residential buildings or changes in energy demand) affect level of households' obliged consumption, hence welfare, while GDP levels are only indirectly affected.

The overall welfare loss is estimated at 0.5%, o.6% and 1.8% reductions for the three warming levels or, in absolute terms, at 160, 170 and 491 bn€.

Up to 90% of the economic losses account to inland flooding and coastal damage, agricultural damage makes up about 10% of the total damage, while the impact of changes in energy demand is barely noticeable at the aggregate, global level which, however, only masks the regional heterogeneity of results as can be seen from regional results.

In relative terms, the largest GDP reductions are estimated for Asian and Oceanian regions (up to over 3% reduction in China at SWL4.0) South America (about 1% at SWL4.0) and Russia and FSU regions (more than 2% at 4.0oC). Some regions (mainly Europe and the USA) benefit from reduction in net demand for energy (increase in demand for cooling is smaller than reduction in demand from heating) which, at the global level, compensates the increase for residential energy demand in other regions resulting in minor net global effect.

In absolute terms the largest GDP losses occur in China (50bn€ to 150bn€), about a third of the global effect. Large economic implications are also noticed in the USA (about 30bn€) and Central Europe North (45bn€).
When compared to the GDP losses welfare impacts are larger. Also, contributions of different impacts to the total are different. For example, changes in residential energy demand lead to change in households' disposable income (and welfare) rather than to changes in overall regional production activity (GDP). Also, river floods are more likely to affect residential households and alter households' budgets (increase in budget's share spent on damage repair leading to welfare reduction) rather than changes in GDP.
There is a series of limitations that should be considered when interpreting the results. Firstly, this is not an assessment of the benefits of climate mitigation policy because fundamental impact areas, like the effects on human health or ecosystems are not accounted for in the suggested methodological framework. Furthermore, possible catastrophic climate impacts are not taken into account either in the analysis.

Migration
It is today widely acknowledged that climate change will induce significant population movements of different nature - forced and voluntary, temporary and permanent, internal and international. Logically, different studies have attempted to forecast the magnitude of future migration flows induced by climate change, but these have not proved conclusive so far. We provided a more realistic approach to forecast migration futures under different climate scenarios. For this, we have built an approach based on case-studies representative of different contexts and different climate impacts, in Africa and South Asia. By studying how populations have reacted to climate shocks in the past, we were able to identify a series of key patterns that will determine migration flows under different levels of warming. Our key argument is that warmer temperatures will not just affect the amplitude of migration, but more importantly its patterns.

We found that there is no correlation between the magnitude of climate and social impacts: an accumulation of small climate impacts can lead to major transformations, while some migratory systems will be able to absorb important climate shocks. Furthermore, aggravated climate change will not necessarily result in an increase of migration, as more people will also find themselves unable to move and relocate to a safer place. This is the reason why the scale of migration resulting from climate change cannot only be understood in terms of alteration of volume, but rather of patterns and structures.

Mobility dynamics associated with extreme climate change will often be non-linear, and question our acception of the expected migration patterns. At the end of the day, such will come down to the issue of inhabitability, which is by itself a very subjective concept. These characteristics should help better anticipate the patterns of future migration flows associated with climate change.


Pathways to warming levels
Recent global emissions, and those projected for the next few years on the basis of Gross Domestic Product, are tracking the Intergovernmental Panel on Climate Change (IPCC) high emissions scenario.
By the 2060s, the high emissions scenario leads to global warming exceeding 2°C in all IPCC climate models. The fastest-warming models simulate 4°C warming by this time, and at least one reaches 6°C by this time. There are, however, a number of different emissions scenarios that could lead to any particular levels of global warming being reached.
The uncertainty in forcing from different greenhouse gases was estimated from emissions in the new Shared Socio-economic Pathways (SSP). This scenario set samples across socio-economic pathways, across different Integrated Assessment Models (IAM) used to create them (which are taken as a proxy for uncertainty in technological assumptions and costs) and across different levels of climate ambition. This supplies over a hundred scenarios of future emissions providing a much better sampling of uncertainties than the Representative Concentration Pathways (RCPs) used in the IPCC’s fifth assessment, which contained just four scenarios.
Given recent and ongoing progress toward successful a global climate agreement, analysis was made of the extent to which any uncertainty in forcing would affect the timing of reaching 2°C under the most ambitious climate targets in the SSP, that of a 2100 radiative forcing of 2.6 Wm-2. It was shown that uncertainty in emissions (and hence forcing) arising from socio-economic assumptions and emissions estimates from different IAM can have a significant impact on the timing of reaching 2°C. A key result from this deliverable was that the median timing of reaching 2°C global warming could range from the mid-2050s to the 2090s under ambitious mitigation depending on plausible perturbations to methane emissions.
TIAM-UCL was used along with results from state of the art climate models to examine the energy-economic impacts of two different climate tipping points, the release of carbon from permafrost, and the collapse of the Atlantic Meridional Overturning Circulation (AMOC). For HELIX a methodology was developed in TIAM where, unlike most other IAM, regional climate differences feedback on the regional energy demand for heating and cooling. This propagates into the all other aspects of the model allowing an assessment of the impacts on the energy sector development and mitigation costs over the century.
A key finding from the inclusion of permafrost is a disinvestment of coal which may vary by up to M$ 350,000 over 15 years between the start of mitigation policy (2020) and 2035 as the permafrost carbon release reduced the allowable anthropogenic emissions. Gas generation reduces by 2035 with the inclusion of permafrost however no large difference in investment is found as gas technology is instead used as backup for the large development of intermittent renewables rather than as base load generation. This suggests that the impact of permafrost thawing on carbon budgets, and by inference the impact other Earth system feedbacks also, should to be included in integrated assessment of even ambitious climate targets.
The changes to energy-economics when inducing an AMOC collapse where examined by introducing a feedback of regional climate change on to the regional demands for heating and cooling, which currently represent about 20% of total energy demands.
The impact of climate change on cooling and heating combined represent around 10% of total global energy total demand. Moreover, under a warmer climate the changes can compensate themselves with an increase in cooling and associated decrease in warming required in buildings. These two demand modifications will under the socio-technological development of SSP5 (fossil fuel intensive) mostly impact the electricity demand as both cooling and heating with time will be mostly supplied by efficient heat-pump technologies.

Conclusions
To provide more detailed representations of climate processes and impacts, we used a new, higher-resolution (approximately 60km resolution) global atmospheric General Circulation Model driven by patterns of sea surface temperatures and sea ice from selected members of the 5th Coupled Model Intercomparison Project (CMIP5) ensemble, forced with the RCP8.5 concentration scenario. We used impacts-relevant indices and impacts models (both statistical and process-based) to examine the projected changes in weather extremes and their implications for fresh water availability, river and coastal flooding, biodiversity, vegetation productivity, vulnerability to food insecurity, and economic impacts. We also assessed relationship between climate and migration. Uncertainties in regional climate responses are assessed, examining ranges of outcomes in impacts to inform risk assessments. The projections for weather extremes indices and biophysical impacts quantities support expectations that the magnitude of change is generally larger for 2°C global warming than 1.5°C. Impacts are substantially larger at 4°C global warming.
Temperature-related warm extremes become even warmer, with increases being more intense than seen in standard CMIP5 projections. Precipitation-related extremes show more geographical variation with some increases and some decreases in both heavy precipitation and drought. There are substantial regional uncertainties in hydrological impacts at local scales due to different climate models producing different outcomes. Nevertheless, hydrological impacts generally point towards wetter conditions on average, with increased mean river flows, longer heavy rainfall events, and increased river flooding, a particularly in South and East Asia where impacts are projected to be very large, affecting tens of millions of people. Similar numbers are at risk from increased coastal flooding. Some areas are projected to experience shorter meteorological drought events and less severe low flows, although longer droughts and/or decreases in low flows are projected in many other areas, particularly southern Africa. The changes imply increased vulnerability to food insecurity and negative economic impacts.

Potential Impact:

The potential impact of these results will be a contribution to improved decision-making on mitigation and long-term adaptation to climate change, through improved awareness and understanding of the risks associated with potential changes in regional climate.

A major potential impact of these results will be a contribution to international negotiations on emissions reductions in the United Nations Framework Convention on Climate Change (UNFCCC) under the terms of the Paris Agreement. Our comparison of impacts of climate change at different levels of global warming will inform judgements on the targets for limiting levels of global warming, especially in the light of the increased ambition of the Paris Agreement. Our work on pathways to these warming levels will inform judgements on the level of emissions reductions necessary to meet these targets.

The results will also inform national policies on achieving these reductions, and decision-making on long-term adaptation to climate change, through improved awareness and understanding of the risks associated with potential changes in regional climate.

Other economic impacts, in conjunction with advice from other ongoing scientific research, could be to contribute to better spending decisions in relation to long-term adaptation to climate change by the Commission, governments of EU member states, other governments, the private sector, non-governmental organisations and civil society. This could include either or both: (i) avoidance of unnecessary expenditure on adaptation measures for threats which may actually be low risk, and/or (ii) improved confidence or urgency in expenditure on adaptation measures for threats which are high risk. HELIX should not, and will not, be the sole source of information on which such decisions will be based; nevertheless, our research will make a direct contribution to the wider knowledge base and also make indirect contributions by demonstrating improved methodologies which can be taken up by future research.

Specifically, the results of our research on the plausible range of changes in freshwater availability and crop yields could contribute to an understanding of the risks or opportunities that climate implies for water and food security, which could help inform policymaking for international development and long-term infrastructure planning, alongside other considerations. Similarly, our research on the range of implications for human health and the two-way relationships between climate change vulnerability and migration could also contribute to policymaking in support of the long-term development of healthcare support and settlement planning. Our research on the range of potential impacts on ecosystems and biodiversity may contribute to policy and long-term planning on conservation.

Wider societal implications, again through HELIX contributing as part of a wider knowledge base, could include improved resilience of society to long-term climate change as a result of improved adaptation or avoided maladaptation, and/or indirect benefits from reduced opportunity costs of unnecessary adaptation.

The results will also contribute to public understanding of the potential impacts of climate change, which is important for the successful development and implementation of policies and plans on emissions reductions and adaptation. This could help improve resilience at community and even household levels to long-term climate change.

These impacts of our research are most likely to be seen in relation countries included in our Focus Regions; Europe, northern Hemisphere sub-saharan Africa (Senegal, Kenya and neighbouring countries), and south Asia (particularly India and Bangladesh). This could include national governments, regional and local authorities within these countries, wider organisations above the National level (eg: the European Commission), multi-national organisations and other large corporations, NGOs, and the public. However, since HELIX also has a global scope as well as our focus regions, the research could also lead to impact in other countries outside of our focus regions and at the global level via the United Nations. Follow-on work with partners in Brazil in support of their next National Communication to the UNFCCC has already begun.


Summary of dissemination activities during the project

We have engaged with a very wide range of stakeholders throughout the project, from the very outset, and disseminated our findings widely to policymakers, scientific peers, the media and the public.

We held numerous workshops, dialogue and briefings in Bangladesh, India, Italy, Kenya, the UK and at the European Commission, including presenting results at the end of the project. We published three booklets summarising our results in Bangladesh, India and Europe for policymakers. All three were published in English and the first two were also published in Bengali and Hindi respectively, to increase accessibility.

We had a very strong presence at all Conferences of the Parties to the Framework Convention on Climate Change (COPs) that were held during the lifetime of the project. We provided a mid-project briefing to policymakers at the European Commission in advance of the historic COP21 meeting in Paris, which resulted in the Paris Agreement. We held a number of events at the COP23 in Bonn in November 2017. We led two side events presenting our findings, one in collaboration with our sister projects IMPRESSIONS and RISES-AM. We also took part in a high-level roundtable discussion alongside high-ranking national government representatives including the Prime Minister of Saint Lucia.

We have been highly visible at a number of major scientific conferences, and played a major role in the organisation of the 3rd European Climate Change Adaptation conference (ECCA2017) in Glasgow, UK.

We made substantial contributions to the IPCC 1.5°C Special Report. The HELIX coordinator Prof Richard Betts took part in the process of establishing whether the proposed IPCC report could be delivered, and co-hosted one of the Lead Author meetings. Several HELIX papers were produced addressing the 1.5°C issue and were made available for assessment in the report. One HELIX member served as a Lead Author on the report, and several HELIX members also took part in the Expert Review.

A major impact of HELIX WP1 has been to make a difference by improving skills and know-how of the delivery of climate change research to its audiences. We have developed and disseminated new thinking to better present climate change information with our strand of work on the Challenge of Communication Unwelcome Climate Messages and our Science of Climate Change Communication training workshops. This work continues and is influential beyond HELIX. The training workshops are now an online package ensuring that the legacy lives beyond the project’s funding.

A further legacy beyond the life of the project is the online impacts atlas HELIXscope, delivered after the end of the project so that stakeholders and technical public can see and interrogate HELIX results for the places in the world that they are interested in, and their sectors of interest. These range from socio-economic such as biodiversity or crops, or impacts through changing rainfall or temperature.

We engaged with artists as routes to public audiences. These include the major exhibition for families and guided visits by school pupils Respond|Climate Change and the paintings of Erica Nockalls, first for an event at Glastonbury Music Festival in the UK and for delegates at COP23 in Bonn. Each of the annual meetings also hosted public events for staff and students and locally interested people.

Several HELIX members were prominent in the media throughout the project. Project Coordinator Prof Richard Betts was interviewed in the media on numerous occasions, including ITV News at Ten prior to the landmark Paris Climate Conference, and the BBC Radio 4 flagship news and current affair programme, ‘Today’ and on BBC World TV. Prof Betts also wrote a number of popular articles on climate science. Prof. Saiful Islam of BUET was a public-profiled expert in climate change writing for Bangladeshi newspapers, tweeting, blogging and interviewed for TV news. Prof Anders Leverman also made numerous media appearances. Our online twitter activity was also good, with 1000 followers as a subset of the Richard Betts’ active twitter with nearly 14,000 followers by the end of the project. The project website www.helixclimate.eu was used to host news items on HELIX research and other activities.


Further impact and dissemination activities

There are substantial opportunities to use existing and new HELIX results to generate further societal benefit, through effectively informing national, EU and international climate policy and raising public awareness.

While over 70 papers had been published by the end of the project, these only used a fraction of the model output since much of that became available only in the closing weeks of the project. Further papers are currently in preparation, both by the coordinator at the University of Exeter and by other partners, and many more can be expected – it is estimated that HELIX papers will continue to be published for at least the next two years. Importantly, these papers will be the ones which present the final results of the project and represent the most complete view of the conclusions – they therefore have the potential to have the greatest impact. It is important that the societal benefit of this research is fully realised, so opportunities for further dissemination will continue to be identified and exploited.

Examples of opportunities for impact, influence and facilitating societal benefit

1. Ongoing exposure of existing and new HELIX results. Our results can be communicated through our existing channels (project website, social media, press office), including supporting HELIX partner in communication of their work.

2 Input to United Nations Framework Convention on Climate Change (UNFCCC).
i) The HELIX coordinating institute, the University of Exeter, now has Observer status in the United Nations Framework Convention on Climate Change (UNFCCC), and had a very successful presence at the 23rd Conference of the Parties to the UNFCCC (COP23) in Bonn centred around the HELIX project.
ii) Civil servants and/or government scientists in 3 countries have expressed interest in using HELIX results in the production of their next National Communication to the UNFCCC. In particular, work is beginning on providing HELIX data to government scientists in Brazil for this purpose.
iii) Forthcoming reports by the Intergovernmental Panel on Climate Change (IPCC) provide further opportunities for exposure of HELIX and post-HELIX research to the UNFCCC and wider community.


3 Input to national and EU climate policy
i) The European Union remains a pioneer in climate policy, both for mitigation and adaptation. The Adaptation Strategy is currently under review. HELIX coordinator, Prof R Betts, remains in close contact with Policy Officers and others at the European Commission in order to continue dissemination of post-HELIX findinds.

ii) The Met Office Hadley Centre (MOHC) is the UK government’s official climate science advisory organisation. HELIX director Prof R Betts is Head of Climate Impacts Research in MOHC, alongside his role as Chair in Climate Impacts at the University of Exeter. He is responsible for key areas of the MOHC Climate Programme, which addresses key government question on climate change and directly informs UK climate policy. Government customers in the departments of Business, Energy and Industrial Strategy (BEIS) and the Department for International Development (DfID) have expressed interest in HELIX results. A further development of Prof Betts’s joint position across MOHC and UoE to maximise collaboration and pull-through of UoE research to MOHC delivery to government is in discussion.

iii) Under the UK Climate Change Act (2010), a national Climate Change Risk Assessment (CCRA) is carried out every 5 years. UK government advisors involved in the CCRA process have expressed interest in the results of HELIX for the next assessment.



Examples of specific activities

1. Ongoing communication of existing and forthcoming HELIX results via website, press and social media
(i) Finalising HELIXscope visualisation tool, and ongoing updates and maintenance
(ii) Systematic presentation of existing results, presented in an accessible way (in conjunction with policy booklet – see below).
(iii) Keeping website up to date with new papers from HELIX partners as they get published, and facilitating blog posts on these. www.helixclimatewill be kept active to March 2019.
(iv) Social media coverage of results, with bespoke graphics

2. Summary booklet on global results. This will present global maps and statistics on HELIX projections of impacts of climate change at 1.5C 2C and 4C, with accessible explanations of what they mean and how they were produced. These were not included in the Policy Booklet that was produced within the life of the project (published at the ECCA2017 conference in June 2017 and therefore only including results up until January 2017). Moreover, that policy booklet focussed only on Europe, since it was in collaboration with two other projects. Two other booklets were produced with a specific focus on India and Bangladesh. There have, however, been specific enquiries from several stakeholders about whether a similar booklet will be produced including global results from the end of the project. The audience will be policymakers with an interest in projected climate change risks in regions across the world, particularly United Nations Framework Convention on Climate Change, European Commission and relevant departments in national governments.

3. Organise HELIX-related presence at UNFCCC COP24. Organise events and a stand, and also briefings to policymakers in advance of the conference (most likely in September) to allow them to make use of new information arising from post-HELIX work in the negotiations. Presence at the conference will also allow HELIX members to join other events at the conference – for example, a very high-profile event will be discussion of the results of the IPCC Special Report on 1.5C global warming, which HELIX played important roles in (contributing papers, expert review and hosting one of the Lead Author meetings).

4. Produce information sheets on HELIX results for specific countries / regions
Processing of spreadsheet data, presentation of graphics and writing explanatory text

5. Facilitate HELIX input to Brazilian national communication to UNFCCC, and national communications by other countries. Liaising with partners, providing data and background information, facilitating teleconferences & meetings, assisting with editing of reports

6. Pro-active approach to input to IPCC process. Communicate papers to Editors and Lead Authors of Special Reports currently in preparation, and the forthcoming 6th Assessment Report. Facilitate contributions to Expert Review of the reports.

List of Websites:
www.helixclimate.eu

Professor Richard Betts. Project Coordinator
Chair in Climate Impacts at the University of Exeter
Head of Climate Impacts at the Met Office Hadley Centre
E-mail: R.A.Betts@exeter.ac.uk
Phone: +44(0)1392 725343

Alissa Haward. Project Administrator
University of Exeter
E-mail: A.B.Haward@exeter.ac.uk
Phone: +44 (0) 1392 726757

Maria Pearce. Project Administrator
University of Exeter
E-mail: M.Pearce4@exeter.ac.uk
Phone: +44 (0) 1392 724112
final1-helix-painting-iceberg.jpg
final1-helix-painting-irma.jpg
final1-helix-painting-seascape.jpg
final1-wet-glasto.jpg
final1-hot-glasto-by-erica-nockalls.jpg