Bias adjustment of numerical climate model simulations involves several technical and epistemological arguments wherein the notion of physical inconsistency is often referred to, either for rejecting the legitimacy of bias adjustment in general or for justifying the necessity of sophisticated multivariate techniques. However, this notion is often mishandled, in part because the literature generally proceeds without defining it. In this context, the central objective of this study is to clarify and illustrate the distinction between physical inconsistency and multivariate bias, by investigating the effect of bias adjustment on two different kinds of inter-variable relationships, namely a physical constraint expected to hold at every step of a time series and statistical properties that emerge with potential bias over a climatic time scale. The study involves the application of 18 alternative bias adjustment techniques on 10 climate simulations and over 12 sites across North America. Adjusted variables are temperature, pressure, relative humidity and specific humidity, linked by a thermodynamic constraint. The analysis suggests on the one hand that a clear instance of potential physical inconsistency can be avoided with either a univariate or a multivariate technique, if and only if the bias adjustment strategy explicitly considers the physical constraint to be preserved. On the other hand, it also suggests that sophisticated multivariate techniques alone aren’t complete adjustment strategies in presence of a physical constraint, as they cannot replace its explicit consideration. As a supplementary objective, this study relates common optional adjustment procedures with likely effects on diverse basic statistical properties, as an effort to guide climate information users in the determination of adequate bias adjustment strategies for their research purposes.

Timothy ( Phleum pratense L.) is expected to be more affected by climate change than other forage grasses. Therefore, alternatives to timothy, such as tall fescue [ Schedonorus arundinaceus (Shreb.) Dumort.], meadow fescue [ S. pratensis (Huds.) P. Beauv.], or meadow bromegrass ( Bromus biebersteinii Roem. & Schult.) should be explored. Our objective was to simulate and compare the yield and nutritive value of four alfalfa ( Medicago sativa L.)–grass mixtures and annual crops grown on two virtual dairy farms representative of eastern Canada under future climate conditions. The Integrated Farm System Model (IFSM) was used for these projections under the reference (1971–2000), near future (2020–2049), and distant future (2050–2079) climates for two climatically contrasting agricultural areas in eastern Canada (eastern Quebec; southwestern Quebec). In both future periods, annual forage dry matter (DM) yields of the four alfalfa–grass mixtures are projected to increase because of additional harvests, with greater DM yield increases projected in the colder area than in the warmer area. In both areas, the highest yield increase is projected for alfalfa–tall fescue mixture and the lowest for alfalfa–timothy mixture. The nutritive value of all mixtures should increase due to a greater proportion of alfalfa. In both areas, yields of silage and grain corn ( Zea mays L.), and soybean [ Glycine max (L.) Merr.] are projected to increase, but not those of wheat ( Triticum aestivum L.) and barley ( Hordeum vulgare L.). Tall fescue, meadow bromegrass, and meadow fescue are adequate alternatives to timothy grown in association with alfalfa under future climate conditions.

The objective of this study was to estimate economic losses associated with milk performance detriments under different climate scenarios. A dataset containing milk records of Holstein and daily temperature–humidity indexes compiled over 6 yr in two contrasting climatic dairy regions of Quebec [eastern (EQ) and southwestern Quebec (SWQ)] was used to develop equations for modeling milk performance. Milk performance, including milk, fat, protein, and lactose yields of dairy herds (kg·d−1), were then projected considering six plausible climate scenarios during a climatic reference period (REF: 1971–2000) and two future periods (FUT1: 2020–2049; FUT2: 2050–2079). Economic losses were assessed by comparing future to reference milk prices based on components. Results indicated that fat and protein yields could decline in the future, thus resulting in economic losses ranging from $5.34 to $7.07 CAD·hL−1 in EQ and from $5.03 to $6.99 CAD·hL−1 in SWQ, depending on the amplitude of future temperature and humidity changes and on whether it is milk quota or cow number which is adjusted to avoid under-quota production. The projected increase in occurrence and duration of heat stress episodes under climate change could result in substantial financial harm for producers, thereby reinforcing the necessity of implementing heat-abatement strategies on dairy farms.

Quantile mapping (QM) is a technique often used for statistical post-processing (SPP) of climate model simulations, in order to adjust their biases relative to a selected reference product and/or to downscale their resolution. However, when QM is applied in univariate mode, there is a risk of generating other problems, like intervariable physical inconsistency (PI). Here, such a risk is investigated with daily temperature minimum (T-min) and maximum (T-max), for which the relationship T-min > T-max would be inconsistent with the definition of the variables. QM is applied to an ensemble of 78 daily CMIP5 simulations over Hudson Bay for the application period 1979-2100, with Climate Forecast System Reanalysis (CFSR) selected as the reference product during the calibration period 1979-2010. This study's specific objectives are as follows: to investigate the conditions under which PI situations are generated; to test whether PI may be prevented simply by tuning some of the QM technique's numerical choices; and to compare the suitability of alternative approaches that hinder PI by design. Primary results suggest that PI situations appear preferentially for small values of the initial (simulated) diurnal temperature range (DTR), but the differential between the respective biases of T-min and T-max also plays an important role; one cannot completely prevent the generation of PI simply by adjusting QM parameters and options, but forcing preservation of the simulated long-term trends generates fewer PI situations; for avoiding PI between T-min and T-max, the present study supports a previous recommendation to directly post-process T-max and DTR before deducing T-min.

A spatial analog is a type of climate scenario that associates a projected future climate at a target location with similar recent-past climates at remote locations. Identifying spatial analogs is usually a step preceding the analysis of local climate impacts and adaptation actions in regions already experiencing a plausible future climate for the target. Such methodology involves several important technical choices. For instance, climates must be synthesized, for example as distributions of a few study-specific annual indices calculated from daily temperature and precipitation time series, before a metric is used to rank candidate analogs in terms of similarity with the target. Moreover, simulations used as target plausible future climates must often be bias-corrected, raising the issue of the order of operations between indices calculation and bias correction (BC). This study investigates the impact of the order of operations on the subsequent identification of analogs, by considering three approaches: BC after (the calculation of the indices); BC before; and BC both before and after. The study, part of a broader project related to agricultural pest management, considers five farms as the targets, which are located in southern Quebec, Canada. Target climate scenarios are built from an ensemble of twenty CMIP5 simulations. The reference product, used for both synthesizing recent-past analog climates and bias-correcting simulations, is the MERRA reanalysis. Results led to the formulation of recommendations for climate services centres providing users with spatial analogs, including the recourse to the self-analog test, proposed as a verification procedure to potentially reveal methodological deficiencies.

Floods account for a large part of global economic losses from natural disasters. As a result, the private insurance sector is increasingly participating in the financial risk sharing, thus expanding the role of actuaries to flood risk management. In this article, we investigate pricing and spatial segmentation of flood risk in the context of private insurance, meaning that individual risk assessment should minimize adverse selection. As such, we design a hierarchical flood risk model that allows an assessment at the individual level. Our model relies on a chain of physics-based climate, hydrological, and hydraulics modules combined with civil engineering methods to map the distribution of individual flood losses at high resolution. Building on such approach, we design pricing and segmentation methods tailored for flood risk management. We then apply the methods to study flood risk in a small city in the province of Quebec. We calculate premiums, analyze the impacts of risk sharing, set pricing territories consistent with the spatial flood risk, and finally, quantify the impact of greenhouse gas emission scenarios on individual and aggregate losses, premiums, and tail risk measures.

Climate change can cause negative impacts to the agricultural sector by increasing pest damage to crops. The European corn borer (ECB) is a major insect pest of corn in North America. Its speed of development could potentially accelerate under a warmer climate, leading to an earlier development of the first generation and an increase in the number of generations per year. The main objective of this study was to assess the potential impacts of climate change on ECB management for the future period 2041–2070 in Quebec, Canada, using bioclimatic modelling and climate analogues. First flight of ECB moths could occur about 15 days earlier in the season in 2041–2070 compared to the reference period 1970–1999. The window for insecticide interventions may be reduced under climate change by 15.6% to 27.8% for univoltine ECB and by 13.8% to 52.7% for bivoltine ECB. Climate change could promote the development of an additional generation in the southern region for both races, considering temperature increases and factors inducing the overwintering diapause. ECB management could become more costly both economically and environmentally under the future climate, and it should be revised according to the results of this study.

Univariate quantile mapping (QM), a technique often used to statistically postprocess climate simulations, may generate physical inconsistency. This issue is investigated here by classifying physical inconsistency into two types. Type I refers to the attribution of an impossible value to a single variable, and type II refers to the breaking of a fixed intervariable relationship. Here QM is applied to relative humidity (RH) and its parent variables, namely, temperature, pressure, and specific humidity. Twelve sites representing various climate types across North America are investigated. Time series from an ensemble of ten 3-hourly simulations are postprocessed, with the CFSR reanalysis used as the reference product. For type I, results indicate that direct postprocessing of RH generates supersaturation values (>100%) at relatively small frequencies of occurrence. Generated supersaturation amplitudes exceed observed values in fog and clouds. Supersaturation values are generally more frequent and higher when RH is deduced from postprocessed parent variables. For type II, results show that univariate QM practically always breaks the intervariable thermodynamic relationship. Heuristic proxies are designed for comparing the initial bias with physical inconsistency of type II, and results suggest that QM generates a problem that is arguably lesser than the one it is intended to solve. When physical inconsistency is avoided by capping one humidity variable at its saturation level and deducing the other, statistical equivalence with the reference product remains much improved relative to the initial situation. A recommendation for climate services is to postprocess RH and deduce specific humidity rather than the opposite.

Climate conditions for Québec’s viticultural potential (VP) during upcoming decades are estimated through high-resolution probabilistic climate scenarios (PCS) based on a large ensemble of simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5). VP is investigated through four temperature-related indices identified as current limiting factors for cold, northern latitudes: length of frost-free season (CNFD), growing degree-days (DDB10), annual winter minimum temperature (AWMT), and annual number of very cold days (ANVCD). Results show that by 2040–2050, most of southern Québec can reasonably expect favorable climatic conditions, with enough consecutive frost-free days and growing degree-days for growing current hybrid-grape varieties, as well as some Vitis vinifera grape varieties. Regions with new VP are identified, for example southern Outaouais and along the St-Lawrence River. Cold winter temperatures remain problematic, but technical solutions to this limiting factor exist.

The science of complex systems is increasingly asked to forecast the consequences of climate change. As a result, scientists are now engaged in making predictions about an uncertain future, which entails the efficient communication of this uncertainty. Here we show the benefits of hierarchically decomposing the uncertainty in predicted changes in animal population size into its components due to structural uncertainty in climate scenarios (greenhouse gas emissions and global circulation models), structural uncertainty in the demographic model, climatic stochasticity, environmental stochasticity unexplained by climate–demographic trait relationships, and sampling variance in demographic parameter estimates. We quantify components of uncertainty surrounding the future abundance of a migratory bird, the greater snow goose ( Chen caeruslescens atlantica ), using a process-based demographic model covering their full annual cycle. Our model predicts a slow population increase but with a large prediction uncertainty. As expected from theoretical variance decomposition rules, the contribution of sampling variance to prediction uncertainty rapidly overcomes that of process variance and dominates. Among the sources of process variance, uncertainty in the climate scenarios contributed less than 3% of the total prediction variance over a 40-year period, much less than environmental stochasticity. Our study exemplifies opportunities to improve the forecasting of complex systems using long-term studies and the challenges inherent to predicting the future of stochastic systems.

The interdependence between climatic variables should be taken into account when developing climate scenarios. For example, temperature-precipitation interdependence in the Arctic is strong and impacts on other physical characteristics, such as the extent and duration of snow cover. However, this interdependence is often misrepresented in climate simulations. Here we use two two-dimensional (2-D) methods for statistically adjusting climate model simulations to develop plausible local daily temperature (T-mean) and precipitation (Pr) scenarios. The first 2-D method is based on empirical quantile mapping (2Dqm) and the second on parametric copula models (2Dcopula). Both methods are improved here by forcing the preservation of the modeled long-term warming trend and by using moving windows to obtain an adjustment specific to each day of the year. These methods were applied to a representative ensemble of 13 global climate model simulations at 26 Canadian Arctic coastal sites and tested using an innovative cross-validation approach. Intervariable dependence was evaluated using correlation coefficients and empirical copula density plots. Results show that these 2-D methods, especially 2Dqm, adjust individual distributions of climatic time series as adequately as one common one-dimensional method (1Dqm) does. Furthermore, although 2Dqm outperforms the other methods in reproducing the observed temperature-precipitation interdependence over the calibration period, both 2Dqm and 2Dcopula perform similarly over the validation periods. For cases where temperature-precipitation interdependence is important (e.g., characterizing extreme events and the extent and duration of snow cover), both 2-D methods are good options for producing plausible local climate scenarios in Canadian Arctic coastal zones.

The path toward a warmer global climate is not smooth, but, rather, is made up of a succession of positive and negative temperature trends, with cooling having more chance to occur the shorter the time scale considered. In this paper, estimates of the probabilities of short-term cooling (P-cool) during the period 2006-35 are performed for 5146 locations across Canada. Probabilities of cooling over durations from 5 to 25 yr come from an ensemble of 60 climate scenarios, based on three different methods using a gridded observational product and CMIP5 climate simulations. These methods treat interannual variability differently, and an analysis in hindcast mode suggests they are relatively reliable. Unsurprisingly, longer durations imply smaller Pcool values; in the case of annual temperatures, the interdecile range of P-cool values across Canada is, for example, similar to 2%-18% for 25 yr and similar to 40%-46% for 5 yr. Results vary slightly with the scenario design method, with similar geographical patterns emerging. With regards to seasonal influence, spring and winter are generally associated with higher P-cool values. Geographical P-cool patterns and their seasonality are explained in terms of the interannual variability over background trend ratio. This study emphasizes the importance of natural variability superimposed on anthropogenically forced long-term trends and the fact that regional and local short-term cooling trends are to be expected with nonnegligible probabilities.

Objectives: We propose a novel approach to examine vulnerability in the relationship between heat and years of life lost and apply to neighborhood social disparities in Montreal and Paris. Methods: We used historical data from the summers of 1990 through 2007 for Montreal and from 2004 through 2009 for Paris to estimate daily years of life lost social disparities (DYLLD), summarizing social inequalities across groups. We used Generalized Linear Models to separately estimate relative risks (RR) for DYLLD in association with daily mean temperatures in both cities. We used 30 climate scenarios of daily mean temperature to estimate future temperature distributions (2021–2050). We performed random effect meta-analyses to assess the impact of climate change by climate scenario for each city and compared the impact of climate change for the two cities using a meta-regression analysis. Results: We show that an increase in ambient temperature leads to an increase in social disparities in daily years of life lost. The impact of climate change on DYLLD attributable to temperature was of 2.06 (95% CI: 1.90, 2.25) in Montreal and 1.77 (95% CI: 1.61, 1.94) in Paris. The city explained a difference of 0.31 (95% CI: 0.14, 0.49) on the impact of climate change. Conclusion: We propose a new analytical approach for estimating vulnerability in the relationship between heat and health. Our results suggest that in Paris and Montreal, health disparities related to heat impacts exist today and will increase in the future.

Ouranos is a nonprofit consortium launched in 2002 with the mandate to provide climate services to its governmental, academic, and private partners. These services have focused on the impacts of climate change in the province of Québec, the identification of vulnerabilities and opportunities associated with the future climate, and the assessment of adaptation options. This paper discusses the experience and insights acquired at Ouranos over the last 10 years in building climate scenarios in support of these impact and adaptation studies. Most of this work is aimed at making climate science intelligible and useful for end users, and the paper describes approaches to developing climate scenarios that are tailored to the needs and level of climate expertise of different user categories. The experience has shown that a group of professionals dedicated to scenario construction and user support is a key element in the delivery of effective climate services.