Proxies of climate-driven energy demand
HDD and CDD calculated with the temperature of historical climate simulations have been validated against observations17,18. We define our heating and cooling—climate-driven energy demand—proxies as the annual HDD and CDD sums calculated from daily mean, minimum and maximum temperatures following the UK Met Office methodology (Table 1, Methods section) for each of the 30 CMIP5 climate simulations (Supplementary Table 2). The advantage of HDD or CDD annual sums is that they can be compared on a global scale, regardless of the timing and length of local heating and cooling seasons. The heating and cooling proxies are presented for the MMM as averages over three 20-year periods 1941–1960, 1981–2000 and 2021–2040 (Supplementary Fig. 1).
The spatial patterns of the MMM of the heating and cooling proxies are closely linked to the MMM of temperature on a global scale (comparing Supplementary Fig. 1 and Supplementary Fig. 2). The decrease in HDD and the increase in CDD between the three studied time periods are also consistent with the underlying temperature increase. Our results show typical values of the heating proxy over land areas between 0 and 1500 HDD in inter-tropical regions (from 30°N to 30°S), between 1500 and 5000 HDD in mid-latitude regions (from 60°N to 30°N; or from 60°S to 30°S) and above 5000 HDD in polar regions (above 60°N or 60°S). Values of the cooling proxy are between 400 and 2000 CDD in inter-tropical regions, and between 0 and 400 CDD in mid-latitudes. These values change in a warming world. However the magnitude of the changes is not globally uniform (Supplementary Fig. 1).
Heating and cooling changes in the past and in the future
To quantify the magnitude of past and future changes, we use the absolute differences in the heating and cooling proxies between 1981–2000 and 1941–1960, henceforth referred to as past changes, and between 2021–2040 and 1981–2000, henceforth referred to as future changes. We estimate the heating and cooling proxies from the CMIP5 historical simulations for the past and from the projections using the Representative Concentration Pathway 8.527 (RCP8.5; unless otherwise stated) for the future. The MMM heating proxy decreased and the MMM cooling proxy increased over the course of both the past and the future time periods.
In the past, the most important changes in the heating proxy (below −200 HDD) occurred over polar regions (Fig. 1a), while, in the future, a decrease in the heating proxy of at least this magnitude occurs over the entire Northern Hemisphere (Fig. 1b). The increase in the cooling proxy was small in the past, below +100 CDD, everywhere except in some (semi-)arid parts of West Africa (Fig. 1c). The projected future increase in the cooling proxy, on the other hand, exceeds +100 CDD in most of the mid-latitude regions, exceeds +300 CDD in large parts of the tropics, and exceeds +400 CDD in Amazonia, in parts of the Sahel and in the Arabian Peninsula (Fig. 1d).
Changes in the heating and cooling proxies have similar spatial patterns in the past and the future, with an overall extension of the areas with significant changes projected for the future (Fig. 1a, c compared to Fig. 1b, d). Mid-latitude regions present significant changes in both the heating and the cooling proxies. Areas with non-significant changes in the heating proxy are projected to reduce to tropical ocean regions, including tropical islands, as well as Amazonia in the future (Fig. 1b). Conversely, areas with non-significant changes in the cooling proxy are projected to reduce to the northern (above 40°N) and southern (below 40°S) oceans, whereas there is a significant change over all continental areas (except Greenland and Antarctica) in the future (Fig. 1d).
Comparing trends in heating and cooling
Even when the absolute differences in some regions are small from one period to another, they could lead to significant changes in societal behavior, such as widespread acquisition of cooling systems21, as people feel a difference in thermal comfort relative to the past. We quantify trends in climate-driven energy demand for heating and cooling buildings by computing the relative differences in our proxies for the past and the future (cf. Methods section), which leads to important trends in the surroundings of the areas with non-significant changes (i.e., gray shaded areas in Fig. 1).
Over continental areas, the decreasing trend in the MMM heating proxy was weak, ranging from −20 to 0% in the past (Fig. 2a). This trend is projected to become clearly negative everywhere in the future, reaching at least −5% (Fig. 2b).
The increasing trend in the MMM cooling proxy was weak in the past, ranging between 0 and +20% over continental areas (Fig. 2c). This trend is also projected to be more pronounced in the future, exceeding +10% everywhere, reaching at least +20% over mid-latitude regions, and more than +60% in many northern hemisphere regions (Fig. 2d). Over mid-latitude oceans, the projected trend in the cooling proxy is to exceed +100%, which leads to strong gradients close to the coastlines, where an important part of the population lives.
Uncertainty from inter-model variability
The MMM must be interpreted with caution28, as the variability in simulated surface air temperature between individual models can be large24. We select major densely populated areas worldwide to analyze the robustness of the aforementioned MMM results across the thirty simulations. We focus on grid cells which contain (mega)-city locations to analyze the inter-model variability in the heating and cooling proxy trends in the past and in the future.
In mid-latitude cities, there is solid consensus among the model simulations (more than 20 simulations of the 30 agree on the sign of change) in the estimation of the decreasing trend in the heating proxy, as evidenced by the small inter-model variability ranging from −20 to +10% in the past (Fig. 3a). This trend becomes negative in the future for all simulations, ranging from −60 to 0% (Fig. 3b). In tropical cities, even if the negative trend is weak, the inter-model variability was large in the past (with no consensus among the simulations) and becomes smaller (overall consensus) in the future.
The inter-model variability in the increasing trend of the cooling proxy is smaller in tropical cities than in mid-latitude cities in the past and in the future. In tropical cities, the increasing trend in the cooling proxy was weak in the past (about + 5% for the MMM) and associated with a small inter-model variability ranging from about −5% to +40% (Fig. 3c). For most of the tropical cities studied, there is a consensus not to simulate any trend (close to 0%) in the past. In the future, the projected trend in the cooling proxy is stronger (about + 30% for the MMM) and also associated with a small variability between simulations of about +10% to +60% (Fig. 3d). Consequently, the projected increasing trend in cooling over tropical regions is robust.
In mid-latitude cities, where the cooling is generally low, the increasing trend in the cooling proxy was weak in the past (about + 15% in the MMM) and associated with a large inter-model variability (Fig. 3c), ranging from moderate negative trends (about −20%) to strong positive trends (about + 60%). In the future, the increasing trend in the cooling proxy becomes stronger, with the MMM near +70% for most cities (Fig. 3d). There is a consensus among the simulations to predict an increase exceeding +10%. Nevertheless, the inter-model variability is large, with a trend in the cooling proxy reaching up to +400% in some cities. Consequently, a robust increase in the need for cooling over mid-latitude cities is predicted, but the quantification is highly uncertain.
Uncertainty from future emission pathways
The uncertainty of anthropogenic emission pathways and of climate projections both contribute to the wide range of projections of future climate-driven energy demand for heating and cooling buildings. We study two pathway scenarios16, (i) business-as-usual and (ii) moderately mitigated, which are referred to as (i) RCP8.5 and (ii) RCP4.5. To investigate the impact of future greenhouse-gas emissions, we compare the trends in the heating and cooling proxies for the two scenarios in the near-future (as in the previous sections), and by the end of the century, the centennial trend (using the period 2081–2100 instead of 2021–2040 to compare with 1981–2000, cf. Supplementary Material Section II).
The difference between temperature projections based on RCP8.5 and RCP4.5 is small in the near-future. As a result, the magnitude of the trends in the MMM heating and cooling proxies, as well as the inter-model variability, are similar between the two scenarios (Supplementary Fig. 1 and comparing Supplementary Fig. 3 against Supplementary Fig. 4). As temperature projections based on RCP8.5 and RCP4.5 diverge during the twenty-first century, the question arises how scenario-dependent trends in heating and cooling relate to the uncertainty coming from the inter-model variability.
The projected centennial trends in the heating proxy for mid-latitude cities are similar for both RCP scenarios by the end of the century, ranging from (i) −20 to −80% for RCP8.5 compared to (ii) −10 to −60% for RCP4.5 (comparing Supplementary Fig. 5 against Supplementary Fig. 6). The decreasing trend in the heating proxy is therefore robust and comparable for both scenarios.
The projected trends in the cooling proxy for tropical cities are also robust and comparable between the scenarios when considering the near-future, but there is an important increase in the model variability by the end of the century, ranging from (i) + 40% to +200% for RCP8.5 compared to (ii) + 20% to +100% for RCP4.5. In mid-latitude cities, the quantification of the centennial trends in the cooling proxy becomes highly uncertain, but there is a consensus among the simulations to project a centennial trend greater than (i) + 200% for RCP8.5 and (ii) + 100% for RCP4.5. However, by the end of the century, the trends projected by individual simulations between RCP8.5 and RCP4.5 overlap although the MMM trends are different.
Uncertainty from the methodology
Several tests were performed to study the influence of alternatives in the methodology on the simulated trends in the heating and cooling proxies: (a) Changing the calculation method of Degree-Days (UK vs. US) (cf. Supplementary Information Section IIIa, Supplementary Figs 7 and 8); (b) Changing the spatial resolution of the multi-model grid used to calculate HDD and CDD from 1° × 1° to 2° × 2° (cf. Supplementary Information Section IIIb, Supplementary Fig. 9); (c) Changing the temporal averaging periods from 20-year to 30-year averages (cf. Supplementary Information Section IIIc, Supplementary Fig. 10); (d) Correcting model biases based on the difference in monthly mean temperature of historic simulations and observations29 for the reference period 1981–2000 before calculating HDD and CDD (cf. Supplementary Information Section IIId, Supplementary Figs. 11, 12 and 13); (e) Calculating HDD and CDD from MMM daily temperatures (cf. Supplementary Information Section IIIe, Supplementary Fig. 14).
For the five tests, similar results are obtained in terms of the spatial patterns and of the magnitudes of the trends for the MMM. The inter-model variability is also comparable, even when the biases in temperature simulation are corrected. Furthermore, we demonstrate that our results are not sensitive to the choice of base temperature in the Degree-Days calculation (cf. Supplementary Information Section IV). The difference of HDD (resp. CDD) calculated with different base temperatures are constant in time, which means that the trends are the same when comparing the three time periods (Supplementary Fig. 15). We conclude that the quantification of future trends in the heating and cooling proxies is uncertain due primarily to the large inter-model variability and not due to details of the methodology.
Credit: Source link