Skip to main content

Advanced Model Development and Validation for Improved Analysis of Costs and Impacts of Mitigation Policies

Final Report Summary - ADVANCE (Advanced Model Development and Validation for Improved Analysis of Costs and Impacts of Mitigation Policies)

Executive Summary:
To achieve the internationally agreed goal of limiting warming to well below 2°C, or even 1.5°C, emissions reductions across all sectors and countries are required, with a full-scale decarbonisation of the industrial metabolism over the course of this century. Integrated assessment models (IAMs) of climate change provide cross-scale and cross-sector policy support for efficient and effective emission reductions, bridging geographical scales from national to global, timescales from sub-decadal to centennial, and integrate diverse sectors such as power and agriculture. These tools are essential tools for exploring consistent pathways for the achievement of long-term climate goals, and examine the implications of different courses of action and technological and socio-economic developments for energy use, land use and climate futures. As such, they are an important element of a larger discourse about our collective response to climate change.

With increasingly prominent applications and growth in complexity of IAMs, the demand for improved representations, as well as thorough validation of model behaviour, has grown significantly over recent years. The ADVANCE project responded to this demand: it facilitated the development of a new generation of advanced energy-economy and integrated assessment modelling tools and, in parallel, made a coordinated effort to improve model transparency, model validation, and data handling.

The Paris Agreement strengthened the long-term goal of international climate policy efforts, aiming to limit warming to well below 2°C and pursuing efforts to limit warming to 1.5°C. The Paris Agreement also established a bottom-up process of collecting and iteratively strengthening nationally determined contributions (NDCs). ADVANCE used the improved models to generate new and up-dated mitigation scenarios in light of these new developments. Specifically, we assessed the implications and effectiveness of the NDCs, as well as additional requirements for 1.5-2°C-consistent stabilization taking into account both technological as well as behavioural emission reduction measures. Our key findings are:
* The implementation of the Paris Agreement initiates a low-carbon transition for major emitting countries but an intensification of global effort is still required in order to limit global warming to well below 2 °C;
* The 1.5 °C temperature target requires much faster reductions in emissions from energy supply and demand than the 2°C limit, and also requires removal of CO2 from the atmosphere of at least 500 GtCO2;
* A renewable-energy based decarbonisation of power supply is technically feasible and economically affordable despite the challenges associated with the variability of wind and solar power production, and comes with considerable environmental co-benefits;
* Technological developments promoting efficiency, electrification and use of low-carbon fuels are the key to demand-side emission reductions;
* Policies influencing consumers’ attitudes will need to support the energy transformation.

Besides work on model improvement, ADVANCE developed a systematic documentation of all energy-economy and integrated assessment models participating in the project. This documentation describes the structure and assumptions of each model. In addition, a diagnostic database collects the results of harmonised model experiments and provides quantitative indicators that characterise model behaviour. The model documentation and the diagnostic indicators are crucial for enhancing transparency and enable users from both the scientific and climate policy communities to better interpret results in the light of model assumptions and characteristics.

ADVANCE made methodological improvements and data available to the broader scientific community in the form of a modelling toolbox. This toolbox includes newly developed model components, mathematical formulae, algorithmic approaches, examples of model code, and generic input datasets. In addition, a database with final scenario results produced by the improved ADVANCE models was published for further use by the scientific community, for example in the context of future assessment reports by the Intergovernmental Panel on Climate Change (IPCC).
Project Context and Objectives:
Integrated Assessment Models (IAMs) describe the environmental, social and economic factors and interactions that determine climate change. They have become central tools to inform policy makers on different climate mitigation options and their impacts. For instance, results from IAMs are heavily used by the Intergovernmental Panel on Climate Change (IPCC) in its assessment reports. However, with the increasing use and growth in complexity of the models, the demand for improved representations as well as thorough validation of model behaviour has grown significantly over the past years.

The overall objective of ADVANCE was to improve IAM tools for better informing long-term global and regional climate mitigation strategies. To this aim ADVANCE gathered leading modelling teams as well as experts on bottom-up and empirical analyses of energy issues to achieve progress in the following key areas:

(i) Improved representation of the demand side: Nearly all IAMs are more detailed on the energy supply side than on the demand side. Still, most models indicate that improved energy efficiency is a major part of climate mitigation strategies. Hence, a better representation of the factors driving energy demand in various economic sectors, including consumer price response, structural change of the economy and improvement of energy end use technologies, were a priority of the model development performed in the project.

(ii) Improved representation of heterogeneity, behavioural changes and consumer preferences: In addition to technology aspects on the demand side, it is a significant challenge for energy-economic models to account for heterogeneity of consumers. Differences in local circumstances, e.g. with respect to per capita income, access to infrastructure, and urban/rural settings, are important drivers of investment decisions (or the lack thereof) on the demand side. ADVANCE performed pioneering modelling accounting for the effects of heterogeneity, for instance to better understand vehicle choice in the transport sector, or to identify optimal policies reconciling energy access and climate change mitigation objectives.

(iii) Technical change and uncertainty: The understanding and representation of endogenous technical change and innovation in IAMs is still incomplete. At the same time, innovation is widely believed to be an essential pre-requisite for generating the transformational change necessary to achieve a low carbon world without jeopardizing economic prosperity. ADVANCE derived new and improved empirical estimates of technological learning in energy technologies, and performed pioneering research to quantify the sensitivity of crucial IAM results across a wide spectrum of input assumptions.

(iv) Integration of renewable energy sources and resource constraints: There is a number of potential bottlenecks to the transition to low carbon energy systems that thus far only poorly represented in IAMs: the integration of variable renewable energy sources for power supply, both energy consumption and use of specific raw materials of the energy sector itself, energy system demand for land and water, and the resulting interactions with agriculture and other activities and infrastructure. ADVANCE brought significant progress in all these areas. The improved representation of these energy system challenges in IAMs shall improve the validity and robustness of mitigation scenarios and allow them to be evaluated against sustainability criteria.

Model improvements should not be considered an aim in itself. Already in the project lifetime, ADVANCE was able to demonstrate that the improved models contribute to both a more complex EU policy debate and a potential enhancement of the EU Impact Assessments model toolbox. This would be accomplished by a set of policy scenarios that reflect policy debates at stake and was designed in collaboration with EU policy makers.

Besides work on model development, ADVANCE aimed to improve model transparency. A harmonized documentation for all energy-economic and IAMs participating in the project would enhance comparability and interpretability of results. A web-based tool for standard validation and diagnostic tests of scenario results would allow explaining differences in response patterns.
Project Results:
1. Model transparency, validation, and common database

1.1. Comprehensive and harmonized model documentation

Process-based integrated assessment models (IAM) are the principal scientific tool used to analyse long-term global climate change mitigation. IAM-based analyses have been widely cited and adopted in IPCC assessment reports, and were crucial for informing the Paris Agreement on climate stabilisation, as well as regional (e.g. EU) and national (e.g. UK) climate policy. It is thus critical that policymakers and other model users have trust and confidence in process-based integrated assessment models and their analysis.

The development of a harmonized model documentation was one of the core project activities in support of model transparency. As a first output, the development status of the IAMs participating in the ADVANCE project was documented in a harmonized way at the beginning of the project. Technically, both the reference cards and the more comprehensive documentation were originally implemented using a Wiki platform (initially hosted at UCL, to specify and iteratively develop the models descriptions. To enhance the usefulness of this resource, a community review of the current documentation format was organized in 2015 in coordination with the Scientific Working Group (SWG) on Data Protocols and Management of the Integrated Assessment Modeling Consortium (IAMC). In the last project year, the documentation was updated reflecting new model developments that were undertaken over the past few years within the ADVANCE project and beyond. Based on review comments, the overall structure of the documentation – the so-called template – and the technical implementation were updated compared to the first release of the harmonized model documentation in 2013. With a view to the technical implementation, it was decided to migrate the harmonized model documentation from the UCL hosted Atlassian Wiki to the PBL hosted MediaWiki (

To ensure the continuation and maintenance of the harmonized model documentation beyond the end of the project lifetime, the ADVANCE consortium approached the steering committee of the Integrated Assessment Modeling Consortium (IAMC) to discuss the possibility of handing over the operation of the harmonized documentation platform under the umbrella of the IAMC. Following initial elaborations within the steering committee the proposal was presented at the 9th IAMC annual meeting which took place from 5-7 December 2016 in Beijing and was positively evaluated. In addition, several modelling teams expressed of interest to add their own models to the harmonized documentation. Therefore, a process within the IAMC’s scientific working group on data protocols and management was started to organize the transition to an IAM community resource. It is expected that these steps will be completed over the first half of 2017. Due to its significance beyond the scope of the project, the Integarted Assessment Modeling Consortium (IAMC), at its annual meeting in Beijing in December 2016, has decided that the effort will be continued under its auspices after the end of the ADVANCE project. This transition is now on its way, setting up a process and identifying responsibilities to maintain the platform, update the content as well as review new content.

1.2. New frameworks for model evaluation

ADVANCE has established a model diagnostics platform for the larger energy-economy and integrated assessment modelling community ( The ADVANCE diagnostic database collects the results from individual energy-economy and integrated assessment modelling teams in a single platform. It offers easy access to the diagnostic indicators, their comparison across participating models, and thus allows teams to assess how their model is situated in the space of available models. Testing how integrated assessment model (IAM) scenarios of the future compare to historical patterns and diagnosing model behaviour under well-defined, stylised conditions is important for obtaining a proper interpretation of the differences across different models. In addition, increasing the transparency and accessibility of the models, as well as developing and applying standardised tests for validation and diagnostics, is vital for increasing confidence in IAM results.

Moreover, a conceptual framework for the evaluation of IAMs was developed and consolidated during the final year of the project. Drawing upon experiences of the climate modelling community, where model evaluation is concerted and visible, including even a dedicated chapter in the IPCC WG1 assessment reports, evaluation of process-based IAMs to-date has been piecemeal and dispersed. The ADVANCE project therefore contributed the first comprehensive analysis of process-based IAM evaluation, drawing on a wide range of examples across eight different evaluation methods testing both structural and behavioural validity. For each evaluation method, a comparison of the application to process-based IAMs with its application to climate models is provided, noting similarities and differences, and seeking useful insights for strengthening the evaluation of process-based IAMs. Progress was made in particular with respect to identifying distinctive strengths and limitations of different evaluation methods, as well as constraints on their application. A major step toward providing a systematic evaluation framework combining multiple methods that should be embedded within the development and use of process-based IAMs was therefore made.

1.3. New methodologies and input data sets

ADVANCE developed a host of new modelling approaches for IAMs that are made available to the entire modelling community as a so-called toolbox via the ADVANCE project website ( The toolbox responds to the widespread need in the modelling community to share methodologies and input data for the further improvement of integrated assessment and energy-economy modelling. The toolbox collects detailed descriptions of the methodologies developed in the project, including new model components, mathematical formulae, algorithmic approaches, examples of model code, and generic input datasets. Each methodology is accompanied by a manual providing instruction for implementation. Tools were selected based on criteria of relevance for the wider modelling community, successful adoption by ADVANCE modelling teams in the course of the project as well as easy applicability by external modelling teams. All tools were subjected to a peer-review to ensure high standards of data quality and adaptability to other models. Overall 13 tools are made available online as open access resource.

1.4. New climate change mitigation scenarios

As part of the ADVANCE project a database of new mitigation scenarios based on the improved models was established. Following the request by the European Commission to explore the possibility of projects to generate input relevant for the forthcoming IPCC Special Report on “global warming of 1.5 °C above pre-industrial levels and related global greenhouse gas emission pathways”, the ADVANCE consortium explored post-Paris climate change mitigation pathways with a focus on the aggregate effect of the intended nationally determined contributions (INDCs), and the efforts required to achieve climate stabilization below the 2°C and 1.5°C levels. To this end, it considered the set of scenarios along the two axis of (i) near-term policies and (ii) long-term stabilization targets. A total of eight integrated energy-economy-climate modelling systems participated in this exercise. We find that the INDCs result in substantial emission reductions compared to those inferred by pre-existing climate policy commitments, but fall short of reaching reduction levels consistent with cost-optimal 1.5-2°C mitigation. Limiting global mean warming to well below 2°C or even 1.5 °C demands a very tight budget for future greenhouse gas emissions. Based on the ADVANCE model results, we were able to explore the potential and limitations for deep emissions reductions in individual sectors and relate them to the emissions reduction requirements of the 1.5°C and 2°C targets. Key conclusions of the scenario analysis show that mitigation strategies limiting end-of-century warming to 1.5°C have to combine the following crucial elements: (i) diversion of power supply-side investments so as to avoid further lock-ins into fossil capacities and to achieve rapid up-scaling of low-carbon power generation, (ii) achievement of accelerated demand-side energy efficiency improvements and electrification of energy end-uses in the industry, transportation and buildings sectors, (iii) development and upscale of carbon dioxide removal technologies to offset residual carbon emissions, which are likely to substantially exceed the CO2 budget consistent with the 1.5°C limit. The ADVANCE synthesis scenarios also form an important basis and starting point for further explorations of 1.5°C-2°C- scenarios consistent with the recent climate policy developments.

2. Improved representation of the demand side

More than 50% of total energy-related emissions come from end-use sectors and therefore reducing energy demand can heavily contribute to emission savings. A better representation of the demand sectors within the integrated system increases the ability to translate energy reduction potentials into policy measures.

2.1. The transport sector

A large share of the model development has taken place in the transport modules, where capturing technology development, mode choice and shift, infrastructure capacity and costs, consumer heterogeneity and behaviour have been identified as main areas where improvements can be made. By including these details the aim is to improve the models representation of the transport sector and its ability to transition to low carbon intensive options with potential pitfalls and opportunities. Moreover, including more detail in these processes offers the ability to address sector specific policies. Examples can be energy efficiency standards or technology regulations.

Our analysis on the transport sector comparison shows that in all models energy efficiency and fuel mix change play a major role in decreasing emissions in the carbon policy scenario, and in some models activity decrease also plays a role. Most models represent different passenger modes; however mode shift throughout the century is limited in the baseline scenario, and plays a limited role in decreasing emissions in the climate policy scenario. An important factor in transport scenarios is that it is unclear at what point activity demand saturates, which can be seen in the range of activity projections across models. Varying projections of technology development of alternative drive train vehicles in terms of cost and efficiency lead to different fuel mix and efficiency potentials.

We also explored the price- and income-responsiveness of energy demand. The analysis has focused on transport sector fuel prices (namely oil products, natural gas, biofuels, electricity, and coal), but it could be applied to other fuels in other demand sectors as well. Results were compared across models. Results show that energy demand response depends on the strength of the price signal as well as on timing. Interestingly, the models show very comparable oil price elasticity values for the first 10 to 20 years post-shock that are also close to the range described in the empirical literature. When looking at the very long term (30 to 40 years), demand elasticity values widely vary between models. Where in some models transport energy demand is highly responsive to price signals, others show hardly any response. Moreover, our results show that income has a bigger impact on service demand compared to fuel prices. Based on these results we conclude that other policy instruments, in addition to pure financial instruments (such as fuel taxes or carbon pricing), would be needed to significantly reduce future levels of carbon emissions.

2.2. The industry sector

An in depth comparison of the industrial sector representation was performed to understand by what means industrial emissions are reduced, and where uncertainties in model projections lie. In the baseline scenario, the projected behaviour across the models is comparable in the coming decades: the industry sector is relatively energy intensive and remains reliant on fossil fuel (>50%). In the second half of the century however the models project either continuous growth or saturation. Comparable to the transport sector findings saturation of demand is uncertain. This leads to more than a factor 2-difference between the highest and the lowest industrial energy demand projection in 2100. Saturation of industrial energy demand strongly depends on whether regional differences remain, and Non OECD countries will reach similar energy intensity levels as achieved in OECD countries. Models show different measures to mitigate CO2 emissions in the industry sector, similar to the findings in the transport or buildings sector, uncertainties lie in the potential of fuel switching or energy intensity improvements. The models switch from coal to electricity use to reduce industrial emissions. Modelling industrial technologies can constrain the flexibility to use different fuel types and this is recognized in the mitigation scenario results, where technology rich models show less fuel switching as a measure to mitigate GHG emissions.

The cement and iron and steel sector have been selected as key industrial sectors for improved modelling, as both sectors have a large contribution to industrial GHG emissions, and are very policy relevant. For both sectors the modelling challenges have been described. For the cement sector a modelling guide is developed with elaborate description of the cement demand drivers, and how they relate to cement production, the production process and key opportunities to reduce material use, energy use and GHG emissions.

2.3. The buildings sector

5. Buildings model development has focused on explicit representation of energy functions in residential or services sector buildings, such as space heating, space cooling, lighting and appliances. Modelling this detail has as advantage that end use demand can be related to climate conditions and thus captures regional differences. Moreover end use specific efficiency and fuel switching opportunities, and structural change can be accounted for. To improve our understanding of how this affects buildings scenario results, the model comparison has focused on the breakdown of functions in the residential models.

2.4. Integration of bottom-up elements in macro-economic models

ADVANCE developed an energy sector calibration routine with the aim to: i) facilitate the process of splitting the electricity sector of GTAP to individual power generation technologies, ii) improve the representation of energy transactions in Input-Output (IO) tables and iii) develop a methodological approach that will take advantage of the energy bottom-up data and allow to introduce bottom-up modules in CGE (top-down) models. This routine has been developed so as to help reducing the gap between top-down macroeconomic CGE models and bottom-up energy system models. The final output of this is a proposed methodology for introducing a detailed representation of the power generation system in a top-down modelling framework like CGE models. A detailed documentation and technical guide on the energy split routine and enhanced energy transactions calibration has been developed. The documentation also provides an overview of the data used and the methodology developed so as to take full advantage of the extended database compiled. The hybrid modelling methodology is also documented in detail.

3. Improved representation of heterogeneity, behaviour and consumer choices

Drawing upon empirical evidence ADVANCE improved existing integrated assessment modelling frameworks so that they better capture spatial, social, and policy heterogeneities, particularly behavioural changes and consumer preferences.

3.1. Improved representation of heterogeneity, behaviour in transport

To improve the behavioural realism of global IAMs in the realm of consumers’ vehicle choices, ADVANCE developed a novel modelling approach which disaggregates light-duty vehicle demands into a heterogeneous mix of consumer groups and then assigns additional cost terms (“disutility costs”) to the vehicle technologies within each of these groups, in order to capture consumer anxieties regarding risk, vehicle range, and refuelling and recharging station availability, etc. This modelling approach also drew on new empirical analysis in ADVANCE on the strength of social influence effects on vehicle choice taking into account cultural differences between countries.

Based on the new approach, ADVANCE produced the first global model comparison exercise to date dedicated exclusively to realistically representing consumer behaviour in long-term energy transitions. The key finding from our work on modelling consumer behaviour in IAMs is that (i) strategies and policies explicitly targeting consumer attitudes toward alternative fuel vehicles are necessary for driving widespread adoption of these advanced technologies (fuel taxation alone is not sufficient); and (ii) carbon pricing is needed to ensure that the electricity and hydrogen used to power these vehicles are derived from low-carbon sources. In other words, the two classes of policies are found to be complementary in accelerating the transition to a low-carbon vehicle fleet. Policies that may be effective at influencing consumer attitudes include:

• Targets for cumulative vehicle sales, sales quotas, vehicle mandates
• Vehicle efficiency or emission standards
• Vehicle sales incentives (purchase subsidies, tax credits, fee-bates, reduced registration fees)
• Vehicle manufacturer support (RD&D, production subsidies)
• High transport fuel taxes (also carbon taxes or pricing)
• Government and company vehicle procurement policies, other demonstration & test fleets
• Trialling in car clubs or car-sharing networks
• Recharging and refuelling public infrastructure investments
• Workplace or home charging incentives
• Preferential parking or roadway access; reduced congestion charges or tolls
• Promotions, social marketing, outreach, information campaigns

Based on recent experience with these policies in different jurisdictions around the world, two things we have learned are that (i) multi-pronged efforts to promote advanced vehicle adoption are more effective than a single sectoral or economy-wide policy (such as a carbon tax), and (ii) whatever the mix of policies, strong coordination across different levels of government (national, state/provincial, and local) is very important.

3.2. Improved representation of energy subsidies and taxes

In order to fill a major gap with regards to the representation of energy subsidies and taxes in IAMs, IIASA developed a database on energy prices, taxes and subsidies. The IIASA team also pioneered implementation of this dataset in the MESSAGE model with and without energy subsidies in both a baseline and a climate stabilization case.

Fossil fuel subsidies are estimated to be a half a trillion dollars and have been described as a “negative carbon price” and “public enemy number one for the growth of renewable energy”. The IPCC claims that subsidy reforms “can achieve significant emission reductions”4. ADVANCE designed an experiment to investigate energy and emission impacts of subsidy phase-out using different assumptions about the future oil price. The key findings from our fossil fuel subsidy removal experiments are that subsidy removal would have a small impact on energy demand, global CO2 emissions and would not significantly increase renewable energy use. Under a modest climate change stabilization target (550 ppm), subsidy removal would only reduce the price of carbon by 2-9%. Furthermore, removing subsidies in most regions would deliver smaller emission reductions than existing climate pledges and in developing regions subsidy removal may actually increase emissions. We thus conclude that subsidy removal efforts should focus on oil and gas exporting countries where they result in the largest CO2 emission reductions without compromising poverty alleviation efforts.

Moreover, ADVANCE developed a modelling architecture to analyse the emission impacts of continued low oil prices which became an incredibly policy-relevant question during 2015 and 2016 in the face of low oil prices. In other words, as the world marched towards the Paris climate agreement, international think tanks and national delegations alike began to worry, will low oil prices prevent us from reaching our climate goals? Another important question which arose in the face of these low oil prices is how would low oil prices impact investment choices into renewables, coal, nuclear and energy efficiency? And ultimately how would all the uncertainties interact with the uncertainty over the oil price? We find that sustained low or high oil prices could have a major impact on the global energy system over the next several decades; and depending on how the fuel substitution dynamics play out, the carbon dioxide consequences could be significant (for example, between 5 and 20% of the budget for staying below the internationally agreed 2°C target). As expected, low oil prices lead to considerably greater (cumulative) use of oil in both the baseline and climate policy scenarios. Similarly, if natural gas prices remain coupled to oil prices across all regional markets, then the future might also see a similar expansion of gas use in a low-oil- price world. Whether or not oil and gas prices decouple going forward is found to be the biggest uncertainty. If prices do manage to decouple globally, then natural gas deployment stands to gain considerably from sustained high oil prices (because gas would remain moderately priced, midway between the low- and high-case levels), whereas the opposite would be true in a low-oil-price world. Although these dynamics may be expected in the directional sense, the magnitude of the swing is arguably pronounced.

3.3. Capturing distributional effects of climate policy, with a focus on the poor

In order to understand the distributional impact of climate mitigation policy on the energy poor, and the interaction of climate and energy access policies for determining ways to shield the poor, ADVANCE developed the standalone residential sector cooking fuel choice and demand model “Access” designed to run iteratively with the MESSAGE model.

The MESSAGE-Access model was used to pioneer the implementation of an increasingly stringent set of climate mitigation policy scenarios in combination with a range of energy access policy scenarios for the region of South Asia. Analysis of the climate policy impacts on the energy poor and for achieving energy access goals show that stringent climate policies would indeed increase the cost of fossil fuels, including those used in cleaner burning stoves that the poor aspire to transition to. Thus without simultaneous targeted efforts to increase funding for energy access, many who would otherwise have been able to switch from traditional solid fuels to modern cooking fuels would no longer be able to afford the switch. However, the additional costs associated with supporting those unable to afford cleaner burning stoves are within the range of financial transfers to South Asia estimated in efforts-sharing scenarios of international climate agreements. Efficient design of access policies can also effectively shield the poor at reduced policy cost.

4. Capturing technical change and uncertainty

Technical change and uncertainty are key factors for producing more robust estimates of climate policies. Technical progress is one of the fundamental drivers of the long term transformation towards a low carbon society. Uncertainty permeates climate change; without accounting for its sources and impacts, any assessment would not be sufficiently robust.

4.1. Advanced frameworks for modelling technological innovation

Technological change is recognized as a one of the key elements of the carbon strategy. Improvement in the efficiency of clean energy production and increased efficiency in energy use are perhaps the most straightforward paths towards a wealthy and green future for Europe. Integrated Assessment Models must reflect this pivotal role of technological change. When forecasting for the long term, such models needs to correctly identify the drivers of technological change, accurately describe the process of green innovation capturing its main features – such as spill over effects and delayed adoption and predict its consequences for the economy and the environment. We improved the representation of technological change in Integrated Assessment Models (IAMs) by conducting research in three areas.

First, we worked on development of the learning curve – the most popular tool used so far to endogenize technological change in IAMs. The learning curve is a simple log-linear relation between cumulative installed capacity of a green technology (e.g. capacity of wind farms) and its cost (price of wind turbines). It offers modellers a simple way to forecast the drop in technology price after an increase in demand. The simplicity of the learning curve has been the reason for its success: a learning-by-doing process modelled along its lines is included in almost any endogenous technological change IAM. Given the relevance of this concept, we intentionally chose to work on its developments and improvements rather than proposing some new framework to endogenize technological change. Specifically, we provide new estimates of the learning rate (i.e. the slope of the curve), more robust and more consistent with the economic theory. We also endogenize the learning rate by analysing the historical relation between learning rate, public and private R&D spending and changes in material prices.

Comparing our estimates of learning rate with previous available results confirmed our initial theoretical prediction that past estimates are biased upward. Using the data on wind turbines technology, we find that the unbiased two stage estimate of the learning rate is 6.7-6.9% - more than one percent lower when compared to the OLS estimate. The lower learning rate implies smaller response of wind turbines prices to increase in demand and higher mitigation costs. The descriptive statistics of learning rates for wind and solar technologies using the TECHPOL dataset suggest that the two technologies show the same profile with a period during which learning is lower than average, followed by an sharper decrease in cost, then after 2000 a stabilization or even an increase in costs. The average learning rate is about half for Wind, as for PV (LR wind = 9% LR PV = 23%). The novelty in the analysis is to examine if the variation in the learning rate over time can be explained with the dynamics of R&D spending and material costs. We find that learning rate and R&D spending may move together, though we also find that there are periods which display a gap between R&D knowledge stock and learning rate. However those periods are consistent with the periods of high or low level in material prices. Econometric or by analytical methods, will allow further defining satisfactory specifications of a dynamic learning curve with an endogenous learning rate, explained both by growth in the knowledge stock and by the materials price level.

The second line of research focused on the efficiency of energy use. An increase in efficiency is crucial for reduction in energy consumption while maintaining healthy economic growth. Yet, until now, IAM did not model the way in which technological change shapes the demand for energy, with a few exceptions. We designed a new tool which can utilize forecasts of energy expenditures to predict future increase in efficiency of energy use. The tool consists of the system of two equations that are first derived from the macroeconomic theory and then calibrated using a rigorous econometric model.

The model which examines the drivers and consequences of energy saving technological change predicts that the R&D effort for energy saving technologies in any sector is determined by energy expenditure in this sector and by the parameters governing the innovation process. The econometric estimation suggests that an increase in energy expenditure by 10% results in 9% increase in number of energy related patents. An increase in a flow of patents by 1% results then leads to increase in growth of energy efficiency by 0.3%. The empirical model also shows that both intertemporal and international spill overs play a significant role in innovation process.

The third activity performed was the collection of data that can be relevant for calibration and development of large energy sector simulation models. We constructed two databases. The aim of the TECHPOL database was to provide reliable data on the costs and performance of representative supply and demand energy technologies. In the second database we computed the global public energy R&D budgets for the key energy technologies.

In the TECHPOL database, almost 30 different generic technologies are considered which belong to three broad categories: large scale power generation, renewable power generation and transport technologies. The large scale power generation includes pulverized coal, integrated gasification, gas turbines, conventional oil power plants, 2nd and 3rd generation nuclear technologies as well as CO2 capture technologies. The renewable category includes hydraulic power plants, small and large PV systems, concentrating solar power and biomass. Some examples of variables included in the database are overnight investment costs, construction time, technical lifetime, load factor, operation and maintenance costs and electrical efficiency. In the dataset of energy technologies, public R&D we construct tables of year by year and cumulative public energy R&D according to the IEA main technological categories.

4.2. New methods to represent uncertainties

ADVANCE developed three new numerical methods for dealing for uncertainty in IAMs. Our analysis shows that incorporating new methods to represent uncertainties is indeed feasible in large scale IAMs. Methods allowed to assess the robustness of the IAMs results, and to attribute the sources of uncertainty to different fundamental policy drivers.

The first one has used a recently developed decomposition methods to carry out ‘Global Sensitivity Analysis’, which allows to compute both individual and interaction effects. We have applied this novel methodology to assess the role of key factors in determining research and development, and emissions in the newly developed Shared Socio Economic scenarios (SSPs). In a multi-model comparison exercise on global sensitivity we specifically looked at the effect of a set of assumptions (population, income per capita, energy intensity, fossil fuels availability, and low-carbon technologies development) on emission changes when a) moving from the SSP2 scenario, the "middle of the road" case, to the SSP1 scenario, the more sustainable scenario and b) moving from the SSP2 to the more challenging world of SSP3. Overall, this analysis shows that the assumptions about energy demand and per capita income underlying the SSPs appear to be the most influential factors in explaining the projected change in future cumulative CO2 emissions.

The technique of modelling to generate alternatives (MGA) allows relaxing one key assumption of IAMs, meaning that of cost optimality. In reality, due to uncertainties in political and behavioural elements, economic optimization can be hardly achieved in reality. MGA allows mapping the diversity of different energy systems that lie within its near cost minimum solution space, allowing assessing the stability of the solution outside its optimality. This technique has been applied to one IAM from the ADVANCE consortium, assessing a business as usual (BAU) case and a global CO2 reduction pathway scenario applied to SSP2.

Another technique to incorporate uncertainty into IAMs is sampling to assess how to best allocate Research and Development investments in low carbon technologies. ADVANCE used real data from three large expert elicitation studies and incorporates them into an IAM. The results show that it is important to account for both the prospects for technological advancement and the interactions of the technologies in and with the economy.

4.3. Decision making under uncertainty

Uncertainty is one of the defining characteristics of the climate change problem. It affects the science, as well as the policy space. First, there is a great deal of uncertainty in projecting emissions forward, more in translating these emissions into temperature increase, and even more in assessing the final impacts on the economy and society. Furthermore, there is uncertainty in the strategies aimed at mitigating and adapting to climate change.

In this context ADVANCE assessed different decision making criteria to assess optimal climate policy under uncertainty. We have used the data coming from the latest IPCC assessment report –from all the three working groups- to quantify uncertainty about climate dynamics, mitigation (as produced by IAMs) as well as impacts. We have shown that the choice of the criteria is as important as other fundamental parameters such as time discounting.

We have also focused on the specific case of tipping points in the climate system, and their repercussions on climate policy. This is an important issue, given the growing awareness of potential catastrophic events related to climate change. We show how the distribution of agreement in experts or models impacts optimal climate policy. In particular, when mitigating GHG emissions lowers experts or models disagreement, and when the decision maker is averse to model uncertainty (contrary to the expected utility case), then a higher level of precaution is warranted.

5. System integration, path dependencies, and resource constraints

ADVANCE contributed to a better understanding of the potential energy supply-side bottlenecks and challenges for a low-carbon energy system transformation and improved their representation in the models. Ultimately this improved the validity and robustness of mitigation scenarios, and also allowed them to be evaluated against additional sustainability criteria.

5.1. System integration of Variable Renewable Energy

ADVANCE developed a variety of different approaches to represent the integration challenges of variable renewable energies in IAMs, which manage to capture many relevant non-linear feedbacks of VRE on the rest of the power sector. The improved representation of VRE integration will allow policy-makers to better understand the potential role of VRE for decarbonisation.

First, ADVANCE developed and implemented new wind and solar data sets, i.e. i) country-level resource data with global coverage, differentiated by achievable capacity factor and distance to grid, for the solar technologies Photovoltaics and Concentrating Solar Power and ii) country-level resource data set with global coverage, differentiated by achievable capacity factor and distance to grid/distance to shore, for the wind technologies wind onshore and wind offshore.

Second, it used results from the highly resolved power sector model REMix ((DLR) to create residual load duration curves. The use of the REMix model to create residual load duration curves showed that international transmission grid expansion leads to strongly reduced variability of wind, solar and demand time series: when assuming a well-connected Europe, a combined share of wind and solar of 55% can be integrated at less than 4% curtailment, 40GW of storage and grid extensions of 150 TW km.

Besides the implementation of the approaches by each individual model, ADVANCE performed a multi-criteria evaluation across models. It looked into how the different IAMs react to variations of VRE characteristics to determine the main drivers and uncertainties concerning VRE deployment and it developed a framework to analyse and evaluate power sector modelling and VRE representations in IAMs, and applied this framework to evaluate the newly-developed IAM modelling approaches.

The multi-criteria evaluation of the various modelling approaches led to the following main conclusions: the development of technology costs and integration challenges are the most important determinants of VRE deployment, while assumptions on resource availability are less important; under the assumption of international cooperation on expanding transmission grids, scenario results produced with the new model versions (updated VRE resource potentials, updated VRE investment costs, improved power sector modelling) lead to a more robust view across models on VRE deployment in climate policy scenarios, and project higher contributions from wind and solar; while global net VRE shares averaged over the second half of the century in a 2°C climate policy scenario formerly ranged from 18-64%, they increased to 46-75% with the new model versions – thus the model average net VRE share increased by 20 percentage points.

To conclude, one of the major findings of our analysis is that the energy supply sector could be almost fully decarbonised through wind and solar power alone, without the use of nuclear and carbon capture and storage (CCS). This would require, however, considerable additional investments into grid infrastructure and storage systems. Based on the improved methodologies, the models arrive at higher estimates of the role of wind and solar for the decarbonisation of power supply, compared to previous model versions with a less sophisticated representation of the variability. This is largely due to overly conservative assumptions on technology costs and on the integration challenges related to coping with a variable renewable electricity supply in earlier model versions.

5.2. Accounting for indirect emissions in the energy transformation

ADVANCE combined, for the first time, life cycle assessment methods with integrated energy scenarios in a forward-looking framework to explore the role of embodied energy demand and indirect emissions in the energy transformation sector in the coming decades. Thus far, no systematic analysis of the direct and indirect greenhouse gas emissions of alternative climate mitigation scenarios has been performed; however low-carbon power technologies differ in their embodied life cycle energy requirements for construction and operation, and related life cycle greenhouse gas emissions. From our analysis we find high upstream energy requirements for electricity production from coal, gas and biomass, while nuclear, wind and solar technologies have a more favourable life-cycle energy and greenhouse gas balance. We also find that wind- and solar-based power-sector decarbonisation strategies have considerably smaller non-climate environmental impacts than strategies based on nuclear and CCS as main technology options.

As part of its analysis ADVANCE developed a new dataset on embodied energy use of electricity supply, breaking down energy demand by energy carrier, energy service/material demand and electricity technology across regions and future time steps. Also, ADVANCE makes the new methodology on calculating embodied energy use and lifecycle GHG emissions available open access via the ADVANCE toolbox for use of other modelling teams.

5.3. Accounting for the land use-water-energy nexus

ADVANCE enhanced the representation of water demand for power generation and land use systems and subsequently investigating how system transformations under reference and decarbonisation scenarios affect water resources. ADVANCE improved the available data and methods for the representation of water use for electricity and used the advanced methodological frameworks to assess numerous reference and decarbonisation scenarios via a multi-model inter-comparison exercise.

The central finding of our work is that the decarbonisation of the electricity sector has clear co-benefits for water resources primarily due to the phase out of water-intensive coal-based thermoelectric power generation. However, these co-benefits can vary substantially. Wind and solar photovoltaic power represent a win-win option for both climate and water resources, but nuclear expansion may result in increased pressures on the water environment. Also, model-induced uncertainty provided useful insights with regards to the most crucial factors of real-world uncertainty. These were identified to be future water-efficiency improvements of power generation technologies, future composition of the electricity mix, and potential energy efficiency improvements.

5.4. Representing infrastructure and network externalities

ADVANCE investigated the representation of network infrastructures in IAMs, such as pipelines, electricity grids, roads, railways, etc. as they have considerable network effects on the economy, energy security and climate change mitigation. Transport infrastructure modelling was further developed in the IMACLIM-R model. More specifically, the model now includes (i) deployment costs of transport infrastructure for automobiles, public transport and air transport and (ii) detailed modelling of how road infrastructure for automobiles is deployed.

Baseline calibration was improved by inclusion of infrastructure costs and constraints on the deployment of roads. The new baseline tracks more precisely than before historically observed data for the period 2001-2013 for transport energy demand and passenger kilometres travelled. This suggests that the improved modelling approach better reflects the dynamics of aggregate transport demand. From a mitigation perspective the analysis shows that a carbon budget scenario together with a restriction on infrastructure deployment lowers the cost of meeting CO2 goals. There are three reasons for this. First, restricting the infrastructure expansion of road and airports leads to an increased use of public transport (to satisfy transport needs) and to an increase of other economic activities that are less carbon intensive and imply less use of oil and coal (natural gas use increases). Second, investments in infrastructure increase the activity of the construction sector and thus slow structural change of the economy towards more productive and less carbon intensive sectors. This results in lowered GDP and higher energy intensity. Thus when infrastructure expansion is restricted the fall in GDP is reversed. Thirdly, early restrictions of infrastructure for roads and air travel promote an expansion of infrastructure for public transport adequate to meet low-carbon transport service demand. Thus, the needed public transport infrastructure is already in place when energy prices increase later in the century (as a result of diminished resources and higher CO2 taxes). Without restrictions in infrastructure, i.e. when less public transport infrastructure is available, more costly mitigation investments are necessary in other parts of the economy.

6. Impacts of mitigation policies in the EU and beyond

Over four years ADVANCE has improved Integrated Assessment Models (IAM) to better quantify the requirements for climate stabilization and the implications of international climate agreements, including the implications of the Paris Climate Agreement. The Paris Agreement reinforced the objective of keeping global temperature rise well below 2°C, and of pursuing efforts to limit the temperature increase even further to 1.5°C above pre-industrial levels. Such low stabilization requires swift action and an almost full-scale decarbonisation of energy systems worldwide.

Based on the methodologically enhanced models, ADVANCE conducted a multi-model analysis of the global implications of the Intended Nationally Determined Contributions (INDC) in 2030, including energy system and economic impacts as well as a gap analysis of INDC levels from the optimal 1.5°C - 2°C scenarios. One of the main findings was that the INDCs result in substantial emission reductions compared to those inferred by pre-existing climate policy commitments, but fall short of reaching reduction levels consistent with cost-optimal 1.5-2 °C mitigation.

The consortium also analysed the role of sectors for achieving the global long-term target set by the Paris Agreement. It found that the power sector accounts for more than half of CO2 emission reductions in the INDC scenario, but also holds the greatest potential for further reductions to put the world on track for the 1.5-2°C limits. Energy demand has a lower effect on near-term abatement but its share increases with increasing abatement efforts. Within the demand sector, industry achieves almost half of the reductions, followed by the transport sector while buildings contribute less, depicting also their lowest share in total demand side CO2 emissions. Emission reductions in the LULUCF sector are important in all scenarios. As its share in total reductions decreases with increasing abatement efforts, the abatement potential of the sector is utilized already in the INDC scenario due to cost-efficient marginal abatement costs.

As a result of the outcome at the COP21 in Paris, there is a renewed momentum to look further towards the mid-century at a (sub-) regional level. This is why ADVANCE also looked into emissions, energy demand and supply, as well as GDP impacts in 2030 and 2050 in reference, 1.5°C and 2°C scenarios. Findings are a first step of a long process, forming the basis of further analysis to be conducted in the CD-Links project that has a more (sub-) regional focus, including an analysis of the EU.
Potential Impact:
1. Science-based policy support

The entire project was devoted to the coordinated improvement of state of the art integrated assessment models, MESSAGE (IIASA), IMAGE (PBL), WITCH (FEEM), REMIND (PIK), the CGE models IMACLIM (CIRED), GEM E3 (ICCS) and the energy system models TIAM (UCL) and POLES (IPTS, UPMF EDDEN & ENERDATA), with the objective of overcoming relevant limitations. This included the extensive development of new model components related to technological innovation, non linearities, path dependencies and irreversibilities, e.g. due to infrastructure and system level effects and energy (service) demand at the sectoral and consumer level.

Model improvements help quantifying the requirements for climate stabilization and the implications of international climate agreements, including the implications of the Paris Climate Agreement. For instance, the project report “Deep Decarbonisation Towards 1.5 °C – 2 °C Stabilisation” syntheses policy-relevant project results with a focus on the decarbonisation challenge of the Paris Agreement. It addresses the implications of the Intended Nationally Determined Contributions (INDCs) and the required low-carbon transformation of the energy system, including emissions reduction measures in electricity supply and energy end-use sectors.

ADVANCE results also come timely to inform the more formal follow-up process of the Paris Agreement. In fact, the consortium followed the request by the European Commission to ongoing projects to generate input relevant for the IPCC special report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways. ADVANCE provides the first multi-model analysis of post-Paris 1.5-2°C-consistent mitigation pathways. Our scenarios fully reflect the effect of the intended nationally determined contributions (INDCs) allowing a first analysis of (a) the effectiveness of the Intended Nationally Determined Contributions (INDCs), and (b) the requirements for limiting warming to below 1.5°C. Results will be published in peer-reviewed journals in time to be assessed by the IPCC for the 1.5°C Special Report. They are also expected to contribute to the review in 2018 under the global stocktake – a process that will enable countries to assess progress towards meeting the long-term goals set out under the Paris Agreement – as well as to the identification of research needs for the Sixth Assessment Report of the IPCC.

2. Transfer of knowledge to the wider scientific community

To ensure transferability of project results and exploitation by the international modelling community, ADVANCE developed several online resources. The ADVANCE toolbox collects detailed descriptions of the methodologies developed in the project, including new model components, mathematical formulae, algorithmic approaches, examples of model code, and generic input datasets. Each methodology is accompanied by a manual providing instruction for implementation. In addition, a database of new mitigation scenarios based on the improved models was published for further use by the scientific community. The scenarios form an important basis and starting point for further explorations of 1.5°C-2°C- scenarios consistent with the recent climate policy developments, for example in the context of future assessment reports by the Intergovernmental Panel on Climate Change (IPCC).

The broad methodological improvements achieved in ADVANCE make it possible to address new research questions. They resulted in more than thirty scientific journal articles in peer-reviewed journals as listed at and most notably, the Special Issues on Renewable Energy Integration in the journal Energy Economics and the Special Issue on the decarbonisation of transport in the journal Transportation Research Part D: Transport and Environment. Furthermore, project findings were published in high impact journals, such as Science, Nature Climate Change, Nature Energy and Global Environmental Change.

3. Increased the transparency of model-based analysis of climate policy strategies

ADVANCE prepared open access tools to increase the reliability and robustness of model-based analyses of climate policy strategies, and ensure increased confidence – and thereby, increased acceptance – in model results. It developed comprehensive and harmonized model documentations for all participating IAMs elucidating structure, assumptions, limitations and input data of the models. The steering committee of the Integrated Assessment Modeling Consortium (IAMC) offered to continue operating the harmonized documentation platform under its auspices after the end of the ADVANCE project and started a process of identifying responsibilities to maintain the platform, update the content as well as review new content. Also, the consortium developed a set of well-defined diagnostic experiments that elucidate model responses to price signals, regulation and socio-economic drivers and allow differences in model behaviour to be better understood.

4. Interaction with relevant stakeholders and the policy community

To validate the chosen approaches and share project achievements, ADVANCE organized encounters with experts, policy and business representatives.

The workshop “System Integration of Variable Renewable Energy (VRE)" was held at PIK in appendage to the kick off-Meeting. It brought together bottom-up experts and IAM modellers and made use of synergies between the ADVANCE project and the Renewable Initiative initiated by NREL. Important topics covered include the state-of-the-art of VRE modelling in IAMs, insights from data-focussed and technology rich bottom-up approaches, and requirements from the perspectives of stakeholders and decision makers.

The workshop “Enhancing the state of transport modelling in IAMs” was organized jointly by PBL and IIASA and held in November 2013 in Laxenburg. It looked at the representation of energy demand in integrated assessment models (IAMs) by focusing specifically on transport, which is an important demand sector within the models, given that energy use and carbon emissions are constantly increasing and their mitigation hard to achieve. The purpose of the workshop was to bring together transport experts from various areas in order to share knowledge on the sector and ultimately provide guidance for model improvement. Topics covered included data, behaviour and infrastructure.

The workshop “Uncertainty in climate change modelling and policy” was organized by FEEM and took place in May 2014 in Milan. The workshop brought together researchers and experts to discuss latest developments in uncertainty and risk analysis in climate change, and their potential applications to IAMs. It focused on decision theory, dynamic and stochastic programming as well as data availability, which all contribute to an improved representation of uncertainty. However, it also discussed challenges in terms of applicability of these methods to large scale integrated assessment models (IAMs) which are routinely used for assessing climate change policies. The agenda covered theoretical, numerical and applications aspects as well as general recommendations for modelling climate change policies under uncertainty.

On 20-21 January 2015, ADVANCE held the expert workshop “Buildings energy demand” in Utrecht (NL). The workshop brought together external experts and stakeholders to discuss technological and behavioural options to increase energy efficiency in buildings as well as demand management options to support grid integration of VRE. Experts agreed on the high potential for energy savings in the sector and on the need to better represent such potential in global long-term models. This however is a big challenge as models do not have the necessary level of detail yet.

On 15 September 2015 the ADVANCE and wholeSEM projects organized a public panel discussion on the role of IAMs in long-term public and private decision making in London (UK). The panel was composed of modellers and representative from Shell and the UK government. While both policy and business representatives acknowledged the support provided by the models to decision making processes in their organizations, they also expressed concerns about increasing model complexity and related difficulties in understanding and interpreting model results. They would welcome easier-to-use tools as well as greater transparency.

The results of the project and implications for climate and energy policies were discussed at the ADVANCE final conference which took place on 24 October 2016 in Brussels and gathered more than 100 stakeholders, climate policy experts and scientists. The conference focused on policy findings relevant for the implementation of the Paris Agreement. The EU Commission (DG for Research and Innovation, DG for Climate Action, DG for Energy), the International Energy Agency as well as NGOs (World Business Council for Sustainable Development, Greenpeace International) and research organisations (ETH Zürich) joined the conference as speakers and panellists.

Besides actively organising encounters, ADVANCE also participated in several conferences to present project results and impact on policy development. For instance, (i) ADVANCE participated in the International Scientific Conference Our Common Future Under Climate Change held on 7-10 July 2015 in Paris. Building on the results of IPCC 5th Assessment Report (AR5), the conference addressed key issues concerning climate change and offered the opportunity to deliver a clear message to the Paris COP in December 2015. ADVANCE contributed presentations to the parallel session “Low carbon pathways for staying below 2°C: Global requirements” as well as to a poster session; (ii) ADVANCE participated in the COP22 UNFCCC Climate Conference and contributed to the side-event “NDCs and beyond: Global and national challenges in implementation and raising of ambition” on 8.11.2016 in Marrakech. It presented the latest findings on the potential of NDC pledges to close the global emissions gap to well below 2°C or 1.5°C; (iii) ADVANCE was called to participate in the first session of the High Level Panel on Decarbonisation Pathways on 21.10.2016 in Brussels which was established by the European Commission to mobilise the means of science and innovation for implementing the Paris Agreement and supporting EU climate action.

5. The project website

The project website promotes project achievements and makes them accessible to and exploitable by external user groups. The website is structured along three central components: 1. Project publications provides an almost comprehensive overview of the scientific results achieved in the project; 2. Outreach promotes the main policy findings of ADVANCE and documents the continuous dialogue with stakeholders and experts along the entire project lifetime; 3. The open access platform Modelling Resources hosts the model documentation, the diagnostic database and methodological toolbox. The platform facilitates the wide-spread use of the project’s model improvements and ensures maximum transparency.
List of Websites:
ADVANCE public website address:
Potsdam Institute for Climate Impact Research
Coordination: Dr. Elmar Kriegler, Dr. Gunnar Luderer
Communication: Laura Delsa, e-mail: Phone: +49 331 288 25 28