When adaptation increases energy demand: A systematic map of the literature

A precise and detailed research question is a central component of a systematic review/mapping (Moher 2009). Here, our goal is to map the different potential or existing cases where 'climate change adaptation affects or may affect energy use, consumption, or demand'. The main challenge facing mapping literature on this issue comes from the choice of the definition of the word 'adaptation', which is inherently ambiguous, and therefore needs to be defined in our review (Ford et al 2015). What is of interest here is a broad definition, similar to what will be referred to as 'response to climate change' in the next IPCC report (AR6). We consider 'adaptation' being the resulting behavior of people (at the individual or collective level) when facing the effects of climate change. This point is important in the framing of the search query (see below), as a particular effort had to be devoted to the selection of words conveying this idea.

We reviewed the literature by conducting a systematic database search with a detailed search query. Two academic literature databases were used to ensure extensive literature coverage: Elsevier Scopus and Thomson Reuters Web of Science databases. We searched for publications in English, concentrating on scientific articles, book chapters and conference proceedings.

The following query was used:

((((('climat*' W/3 (chang* OR variabilit* OR impact* OR scenar* OR futur*)) OR 'global warming' OR 'futur* weather')) W/2 (adapt* OR respon* OR impact* OR effect* OR affect* OR implication* OR influence OR 'in the context of' OR considering OR incidenc* OR pressure* OR challenge* OR driv* OR under)) AND ((energ* OR electric*) W/15 (nexus OR consumpt* OR demand* OR usage* OR utilization* OR requir* OR consideration* OR expenditure* OR load* OR use)))

The query is composed of three main groups. The first reflects the concept of 'climate change'. The second—the largest group—reflects the concept of 'response [to climate change]', with the definition provided above. The third group reflects the concept of 'energy use'. This search query was built in an incremental process, based on a test list of relevant articles that the authors had identified, and on the relevant articles incrementally found with the queries previously tried. We also checked the coverage of the final search query by checking that references cited in seminal papers and in prior literature reviews related to the theme were included in the results (citation tracing).

We also searched for grey literature from different institutions that we identified as likely to produce reports on climate change shifting energy demand. We did different searches both on Google search engine and directly on institutions' websites. We searched the World Bank, the International Energy Agency (IEA), United Nations Environment Programme (UNEP), UN-Habitat, WWF, Greenpeace, Friends of the Earth, Organisation for Economic Co-operation and Development (OECD), European Union Energy Initiative (EUEI) and Climate Change Science Program (CCSP). We found possibly relevant documents from the World Bank, IEA, EUEI, CCSP, UN-Habitat and the Department of Sustainable Development Organization of American States. We also added one document from the French Observatory on the climate change effects (ONERC), and three research reports published in France (VURCA and MUSCADE and ADAMONT projects), which we knew to be relevant.

The database and grey literature search were done in Paris (France), on May 2nd 2019 and updated with Scopus database on November 11th, 2019. The search returned 10 117 papers in total, and 6684 after removing duplicates.

Not all papers were relevant, and we first excluded papers which were obviously out of scope based on their title (1266 papers selected). ⁸ We then checked for relevance based on the abstracts. This resulted in the selection of 562 papers. For robustness, more than half of the abstracts (randomly chosen) were screened by at least two different researchers (we used Abstrackr online tool (Wallace et al 2012) to carry this out). We excluded 233 publications after the full-text reading and finally found 329 publications being fully relevant at the end of the screening process. A synthesis of the whole process is shown in figure 1.

Zoom In Zoom Out Reset image size

During this screening, we aimed at selecting papers which followed the following logic: 'climate changes can lead to changes in behaviors or choices by the population', and 'these changes can lead to changes in energy demand'. We therefore used the following inclusion and exclusion criteria:

• Candidate papers must deal explicitly with energy demand, consumption need or use by the human population (e.g. papers dealing with the metabolic energy demand of plants are excluded).
• Candidate papers must deal with climate change effects (e.g. papers about the impact of temperature on AC adoption and/or energy use are excluded if they do not specifically consider climate change scenarios; likewise for papers about current trends without explicitly considering climate change).
• Candidate papers must deal with the impact of climate change on energy (and not the other way around) (e.g. papers dealing with the impacts of energy consumption on climate change magnitude are out of scope).
• Candidate papers must deal with energy demand (e.g. impacts of climate change on renewable electricity production potential or physical impact of climate change on electricity transport network are out of scope).

In order to extract relevant qualitative information from the publication, we designed the analysis grid shown in table 1. The adaptation actions or measures that we came across during full-text screening are listed in a supplementary database (note that we also included actions listed in papers excluded during full-text screening).

We imported all bibliographic data and publication metadata directly from SCOPUS and Web of Science. We then attributed a unique ID for each article and built a SQLite database with different tables:

• A table containing the ID and bibliographic data of all the 6684 publications identified in our search and retained after removing duplicates, as well as their final status after screening.
• A table containing 6106 pieces of information gathered from the 304 publications reviewed (as one publication often studies multiple places, scenarios, projections, etc.). Each observation is linked to its publication by the unique ID.
• A separate table outlining all adaptation actions or measures reviewed, with links to the associated papers, flagging those providing quantitative estimates of impacts on energy demand.

Figure 2 shows the date of publication of the papers (we stopped the selection on November 11th, 2019, so the year 2019 is incomplete). As can be seen on this figure, the number of publications dealing with our subject is sharply increasing over time, with more than 35 papers published each year between 2016 and 2019, compared to about 20 papers each year between 2010 and 2015, and fewer than 10 papers per year before 2009. Sharp increases can be observed in 2010 and 2016, just after the years of Copenhagen summit and COP 21 in Paris.

Zoom In Zoom Out Reset image size

One of the reasons behind this trend is the fact that the overall number of papers published in climate change related literature is rapidly increasing globally. In order to test whether the increase shown in figure 2 is just a reflection of this global growth, we tried to evaluate the share of the papers on 'energy' or 'climate' that these papers represent. The result can be seen in figure 3, which shows the share represented by the paper that we selected in our review, relative to the total number of papers returned by a search on 'energy" OR "climate change' on Scopus for the same year. An increasing trend is still visible, showing that a larger share of climate change-related papers is each year related to the impact of climate change on energy demand.

Zoom In Zoom Out Reset image size

As stated in the introduction, climate change could impact energy demand in many different ways. Table 2 draws a tentative list of impacts (based on Wood et al (2015), Davide et al (2019) and the authors' expectations) and assesses their coverage in the reviewed papers. ⁹

A majority of papers (80%, 277 papers) study variations in energy demand for heating and cooling in homes and office spaces. This figure may be even higher, as another 42 articles (12% of the selection) study aggregated energy demand variation (generally from top-down global approaches), which often actually mainly encompass energy demand for space heating and space cooling.

Amongst the papers focusing partly or entirely on heating and cooling, 230 study energy demand for cooling, 159 energy demand for heating, and 61 both. Other impacts of climate change on energy demand are only studied in a handful of papers. If we look at the number of case studies instead of the number of papers (one paper may contain several case studies, for instance by separately examining several cities or regions), the results are qualitatively the same (table 4 in appendix): other impacts of climate change on energy demand are only studied rarely.

Figure 19 (in appendix) shows the repartition of the papers each year, with respect to the energy use studied. No clear trend is visible: studies about heating or cooling seem to have always constituted a large part of the papers.

The papers have different coverages: they embrace different scales, from local, district level studies (see for example Giannakopoulos and Psiloglou 2006, Masson et al 2014, Waddicor et al 2016, Hooyberghs et al 2017) to global studies dealing with the whole world or large world regions (see for instance Labriet et al 2015, 2018a). The papers also use different resolutions. For instance, a paper may study a whole country by using aggregated country-wide data (for instance Parkpoom and Harrison 2008, Schweizer and Granger Morgan 2016), or by studying separately the regions inside the country (for example Zhou et al 2014, Mcfarland et al 2015). In the first case, the resolution of the paper will be the whole country, whereas in the second case the resolution will be the region. But both papers will have the same country-wide coverage. In figure 4 we present the distribution of papers as both a function of their geographical coverage, and a function of their spatial resolution.

Zoom In Zoom Out Reset image size

As can be seen in figure 4, both the geographical coverages and the geographical resolutions are widely different across the papers. Almost half of the papers (48%) have a geographical coverage of a whole country or a larger zone (the world or a world region). Conversely, approximately two thirds of the papers (65%) have a geographical resolution smaller than a country (e.g. states, provinces or cities) and only a quarter (26%) rely on data at country scale.

Figure 5 and figure 6 are maps of the locations studied in the papers, showing the number of papers studying a given country or a sub-national location inside it. Energy demand for heating and cooling has been studied in almost every country, the exceptions being mainly Central Asia and Northern Latin America. However, the number of articles per country varies widely. Most African countries have been studied by only one paper (Kahil et al 2018a), whereas 67 papers study the USA and 35 study China. The same pattern appears for the papers studying other subjects than heating or cooling, with, relatively, an even more extreme emphasis on the USA (26 papers, to be compared to 7 on China, the second most studied country) (see figure 6).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As most papers use data with a sub-national resolution, it is also interesting to look at the precise locations of the sub-national places studied. Figure 7 and figure 8 show maps of these locations. Figure 7 is for papers about heating or cooling, and figure 8 for papers on the other subjects. Studies on heating or cooling concentrate on USA, Europe, Eastern China and South-Western Australia. Almost no localized study examines Africa, India, or South America beyond Chile, Argentina and Mexico.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The pattern is different for studies focusing on usage other than heating and cooling: California has been by far the most studied area, with a strong focus on water issues. A few studies focus on Europe (for instance Flechsig et al 2000 studies wastewater in Spain; Gi et al 2018a, Sabunas and Kanapickas 2017 study water heating, among other questions; Hoffmann and Rath 2012 studies greenhouses heating; Spandre et al 2019 studies snowmaking), one on China (Yan et al 2018 about irrigation) and two on Australia (Mushtaq et al 2013 about irrigation; Pickering and Buckley 2010 about snowmaking).

It should be noted that, as it is common with systematic reviews, there may be a potential language bias in these results (see for instance Neimann Rasmussen and Montgomery 2018). Indeed, putting aside four reports in French that the authors knew to be relevant, we only reviewed here the academic and grey literatures in English.

32% of the studies examine the medium term (around 2050), 19% the near term and 26% the long term (figure 9) (a paper can examine different time horizons, in which case we consider that it includes several studies). A few papers consider time horizons that were in the future at the time of publication but which are now in the past (see for instance Rosenthal et al 1995).

Zoom In Zoom Out Reset image size

Several climate scenarios are studied, and we grouped them in four categories: Low (climate scenarios corresponding to +2 °C globally), Medium (scenarios leading to between about +2 °C and +4 °C globally), High (+4 °C or more globally), and Other (no information, or impossible to connect with a global scenario). The distribution of scenarios studied is rather mixed (figure 10). Whereas only a minority of papers (15%) study low climate change scenarios, about one third study medium and one third high scenarios. We provide more details in the appendix. 1

Zoom In Zoom Out Reset image size

The variation in demand due to climate change may a priori not be homogenous over time, and this may be of importance for energy providers, especially if one wants to also take into account the impact of climate change on renewable energy production.

Fewer than 20% of the papers reviewed explicitly consider, beyond average, peak energy demand ¹⁰ (see for example Baxter and Calandri 1992; Colombo et al 1999; Hamlet et al 2010; Damm et al 2017; Maximilian Auffhammer et al 2017). Moreover, the fraction of these papers in the literature does not show any increasing or decreasing trend over time (figure 11). Almost all papers studying peak energy demand focus on Europe or the USA (figure 20 in the appendix).

Zoom In Zoom Out Reset image size

The papers of our sample study different time units, from hourly analyses (such as Conry et al 2015) to yearly averages (such as Gaterell and Mcevoy 2005; H. Wang and Chen 2014). The majority of papers—more than two thirds—study annual averages, and only a few deal with hourly, daily or monthly data (figure 12).

Zoom In Zoom Out Reset image size

We classified the methodologies used in the papers into four groups: Building energy models, Econometric/statistical models, Integrated models (coupled techno-economic models), and Others (qualitative studies, for instance). Building energy models make only sense for studies dealing with air or water heating and/or cooling, whereas the other methodologies may be used for all the subjects.

The methods primarily used in heating and/or cooling studies are building energy models (more than half of the studies) and econometric models (about a third). Integrated modeling represents about 10% of the publications (See figure 13). This, as would be expected, is different on other subjects, as the integrated models represent for instance a large part of studies on water related energy demand. Econometric models are used in a significant share of studies, whatever the subject.

Zoom In Zoom Out Reset image size

In the studies about heating and/or cooling, there is a clear link between the scale of the studies and the method used (figure 14). Building energy models are used primarily for city-scale studies, integrated models for global studies, and econometric studies for intermediate (in particular, national) scales. This pattern is stable over time (figure 21 in appendix). In the studies which are not about heating and/or cooling, building energy models are not necessarily meaningful, but a similar pattern can be seen between studies using econometric modelling and integrated modelling (figure 22, figure 23 and figure 24 in appendix).

Zoom In Zoom Out Reset image size

Some papers consider a direct, mechanistic link between climate change and energy demand variation, whereas other papers explicitly consider, for the same climate change scenario, that different response pathways may lead to different energy demands. This is what we assess in figure 15.

Zoom In Zoom Out Reset image size

More than half (52%) of the publications only consider one response to climate change (planned or spontaneous adaptation). To take a few examples, Isaac and van Vuuren (2009) consider only one scenario for AC adoption by the population (the paper states that the scenario 'does not include any policy efforts towards either mitigation or adaptation to CC'), and studies the resulting variations in AC-related energy use. Many papers such as Sailor (2001), Giannakopoulos et al (2009) or Damm et al (2017) examine the current or past relationships between temperature and energy use. They then use this relationship to project potential future energy demand, thereby effectively ignoring any specific adaptation measure.

16% of the papers explicitly acknowledge alternative adaptation measures without quantifying their impact on energy demand. For instance, Solecki et al (2005) study different strategies to mitigate the heat island effect, but do not quantify their impact on energy demand. Amato et al (2005) or Auffhammer et al (2017) mention that different adaptation measures could impact the relationship between temperature and energy use, but do not provide quantitative estimates. Finally, only the remaining 32% of the papers assess quantitatively different response scenarios (using the methods reviewed in section 3.6). As can be seen in figure 25 in the appendix, there is no visible increase in this share over time, and this share is comparable for the different climate change impacts.

Whether different adaptation pathways are studied in the papers strongly depends on the scale of the studies and the methodology used. Quantified alternatives are more prevalent among smaller-scale studies (figure 16). Only four articles quantify alternatives at the regional or global scale (Quéfélec and Allal 2015, IEA 2018, Kahil et al 2018a, Nematchoua et al 2019). Quantified alternatives are rare among econometrics- or statistics-based papers (figure 17). In contrast, they are studied in almost half of the papers using integrated or building energy models. This is not illogical, as it is more difficult to study the consequences of potential changes in technology adoption or behaviors with econometrics models, which are by nature designed to analyze the consequences of existing technologies and behaviors.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The number and type of actions studied is very heterogeneous across the studies. Table 3 lists the main adaptation solutions studied in the papers dealing with heating or cooling (see table 5 in the appendix for the complete list, and table 6 for other energy usages). In some papers, generic measures are studied (e.g. 'better insulation', (see for instance X. Wang et al 2010, Ren et al 2011, Hosseini et al 2018, Lapisa et al 2018) whereas in other papers specific actions are documented (e.g. 'mixed mode ventilation', (see for instance Daaboul et al 2018, Sánchez-García et al 2019). A direct comparison across the papers is therefore difficult.

Among the many adaptation alternatives considered, 'better insulation' and 'use of shading devices/blinds' stand out as most frequently cited. Both appear and are quantified in about a third of the papers. In contrast, at least 20 adaptation actions—e.g. 'reduced HVAC operation hours'—are only studied or mentioned in as few as two papers (see table 5).

Climate change can impact energy demand via multiple types of usage. However, one specific usage, namely heating and/or cooling, received the majority of research effort (more than 80%, and almost 92% if we include the papers about 'aggregated energy demand', which often actually study heating and cooling demand). It is not obvious, though, whether this mechanism is the main channel through which climate change could impact energy demand. Desalination for water demand, or water pumping to prevent floods or submersion may also, for instance, represent a large consumption of energy, either globally or locally (See chapter 4 in IPCC 1.5 report).

A few reasons may explain why they have not been studied as much. First, present or past use of such technologies to cope with a changing climate is not commonplace, so it is difficult to find data for econometric studies, or with which models could be calibrated. Second, the impact of climate on heating and cooling demand primarily rests on behavioral adjustments, given a certain technology endowment (heating system, air conditioner). ¹¹ In contrast, most other usages considered here—chiefly among them, water desalination (Jones et al 2019)—require new investments, involving more significant costs. The question whether this makes such energy-intensive technologies less likely to be used in the future is, however, open.

Another, third, reason may also explain our result. For heating and cooling, there is an already established and long-studied connection between weather and energy demand (through heating and cooling degree days). This serves as a convenient basis for estimating climate impacts, and has no equivalent with other usages. In almost all the studies about these other usages, energy is not the main variable studied (for example, it would be water in the case of irrigation) and figures about energy demand variation are often a by-product of another analysis. By construction, we have only analyzed in this review the papers which analyze the full chain between climate change and energy demand variation. Beyond papers about heating and cooling, however, this chain may likely be splitted between articles studying responses to climate change and articles studying potentially induced energy demand variations. Focused research and reviews on each of these sectors should be conducted to better investigate into this and assess the actual potential energy demand variations.

Adaptation possibilities are highly case-specific and it is therefore not surprising that a large fraction of the papers study cities or smaller scales. It is a well-known fact that research case studies tend to concentrate on places where research funding is more abundant (Nagendra et al 2018, Lamb et al 2019). Yet it is striking to see that large fractions of the world are not studied in the literature. It should be noted that the same conclusion is true for the literature on climate change impacts on energy production, as highlighted by Cronin et al 2018).

The disconnect between the research effort within our corpus and the associated stakes becomes strikingly apparent when one compares our figure 7 to Ruijven et al (2019)'s distribution of projected impacts of climate change on energy use for heating and cooling (see figure 18). Almost all the literature concentrates on places where heating demand is expected to decrease, and only a few studies are on places where cooling demand is expected to increase (e.g. sub-Saharan Africa). These last studies moreover take a global or national perspective and therefore lack the degree of detail that can be found in less highly-disaggregated studies. This raises the question of whether the current literature, as a whole, presents a systematic bias leading to an under-estimation of climate change impacts on cooling demand.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

High-frequency variations in energy demand, and in particular the magnitude of demand peaks, are important issues for the optimal management and planning of electricity infrastructure, especially those associated with renewable energy production. Some papers study either specifically peak energy demand (20% of the papers) or time scales shorter than a year (less than a third of the papers), but they are a minority, and almost none of them deals with countries beyond Europe or the USA.

The impacts of adaptation to climate change on energy demand depends on the adaptation choices made: promoting air conditioning or adding vegetation in cities are two possible ways to fight against heat waves consequences, for instance, but they lead widely different energy consumption increases (Viguié et al 2020). This is a key idea if one wants to lower the impact of adaptation on energy demand and prevent maladaptation. However, fewer than half of the articles that we reviewed take into account the fact that different adaptation possibilities exist, and only a third actually assess quantitatively the consequences of such different possibilities. Also, most articles comparing different adaptation choices have a local geographical coverage, with virtually none taking a regional or global perspective. This is intrinsically linked to the methods used by the papers. Most large scale assessments indeed rely on statistical/econometric methods, in which it is by nature difficult to take into account of potential adaptation pathways which have not been implemented yet. As a consequence, globally, there is only a weak understanding of the extent to which various adaptation pathways could impact global energy demand. Moreover, we also found that most articles tend to focus on the same adaptation solutions. The literature does not seem representative of the actual panel of adaptation options available today.