• Abbott, K. N., , M. E. Alexander, , D. A. MacLean, , B. Leblon, , J. A. Beck, , and G. C. Staples, 2007: Predicting forest floor moisture for burned and unburned Pinus banksiana forests in the Canadian Northwest Territories. Int. J. Wildland Fire, 16 , 7180.

    • Search Google Scholar
    • Export Citation
  • Amiro, B. D., and Coauthors, 2001: Direct carbon emissions from Canadian forest fires, 1959 to 1999. Can. J. For. Res., 31 , 512525.

  • Amiro, B. D., , K. A. Logan, , B. M. Wotton, , M. D. Flannigan, , J. B. Todd, , B. J. Stocks, , and D. L. Martell, 2004: Fire weather index system components of large fires in the Canadian boreal forest. Int. J. Wildland Fire, 13 , 391400.

    • Search Google Scholar
    • Export Citation
  • Baskerville, G. L., 1971: Use of logarithmic regression in the estimation of plant biomass. Can. J. For. Res., 2 , 4953.

  • Bergeron, Y., , and S. Archambault, 1993: Decreasing frequency of forest fires in the southern boreal zone of Québec and its relation to global warming since the end of the ‘Little Ice Age’. Holocene, 3 , 255259.

    • Search Google Scholar
    • Export Citation
  • Bergeron, Y., , S. Gauthier, , M. Flannigan, , and V. Kafka, 2004: Fire regimes at the transition between mixed wood and coniferous boreal forest in northwestern Quebec. Ecology, 85 , 19161932.

    • Search Google Scholar
    • Export Citation
  • Bergeron, Y., and Coauthors, 2006: Past, current, and future fire frequencies in Quebec’s commercial forests: Implications for the cumulative effects of harvesting and fire on age-class structure and natural disturbance-based management. Can. J. For. Res., 36 , 27372744.

    • Search Google Scholar
    • Export Citation
  • Brewer, S., , S. Alleaume, , J. Guiot, , and A. Nicault, 2007: Historical droughts in Mediterranean regions during the last 500 years: A data/model approach. Climate Past, 3 , 355366.

    • Search Google Scholar
    • Export Citation
  • Briffa, K. R., , P. D. Jones, , and M. Hulme, 1994: Summer moisture variability across Europe, 1892–1991: An analysis based on the Palmer Drought Severity Index. Int. J. Climatol., 14 , 475506.

    • Search Google Scholar
    • Export Citation
  • Cook, E. R., , C. A. Woodhouse, , C. M. Eakin, , D. M. Meko, , and D. W. Stahle, 2004: Long-term aridity changes in the western United States. Science, 306 , 10151018.

    • Search Google Scholar
    • Export Citation
  • Coursolle, C., and Coauthors, 2006: Late-summer carbon fluxes from Canadian forests and peatlands along an east–west continental transect. Can. J. For. Res., 36 , 783800.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , K. E. Trenberth, , and T. Qian, 2004: A global dataset of Palmer Drought Severity Index for 1870–2002: Relationship with soil moisture and effects of surface warming. J. Hydrometeor., 5 , 11171130.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., , P. M. Bothwell, , D. H. Carlsson, , and K. A. Logan, 2003: Simulating the effects of future fire regimes on western Canadian boreal forests. J. Veg. Sci., 14 , 355364.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., , R. D. Field, , M. A. Brady, , O. Roswintiarti, , and M. Mohamad, 2007a: Development of the Indonesian and Malaysian fire danger rating systems. Mitigation Adapt. Strategies Global Change, 12 , 165180.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., and Coauthors, 2007b: Estimating direct carbon emissions from Canadian wildland fires. Int. J. Wildland Fire, 16 , 593606.

    • Search Google Scholar
    • Export Citation
  • Fire Ecology Research Group, 2005: Experimental Climate Prediction Center - Fire Weather Index Forecast. Global Fire Monitoring Center, University of Freiburg, Freiburg, Germany. [Available online at http://www.fire.uni-freiburg.de/.].

    • Search Google Scholar
    • Export Citation
  • Flannigan, M. D., , K. A. Logan, , B. D. Amiro, , W. R. Skinner, , and B. J. Stocks, 2005: Future area burned in Canada. Climatic Change, 72 , 116.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , and M. Mudelsee, 2008: Past and future changes in Canadian boreal wildfire activity. Ecol. Appl., 18 , 391406.

  • Girardin, M. P., , J. Tardif, , M. D. Flannigan, , and Y. Bergeron, 2004a: Multicentury reconstruction of the Canadian Drought Code from eastern Canada and its relationship with paleoclimatic indices of atmospheric circulation. Climate Dyn., 23 , 99115.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , J. Tardif, , M. D. Flannigan, , B. M. Wotton, , and Y. Bergeron, 2004b: Trends and periodicities in the Canadian Drought Code and their relationships with atmospheric circulation for the southern Canadian boreal forest. Can. J. For. Res., 34 , 103119.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , Y. Bergeron, , J. C. Tardif, , S. Gauthier, , M. D. Flannigan, , and M. Mudelsee, 2006a: A 229-year dendroclimatic-inferred record of forest fire activity for the boreal shield of Canada. Int. J. Wildland Fire, 15 , 375388.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , J. C. Tardif, , M. D. Flannigan, , and Y. Bergeron, 2006b: Synoptic-scale atmospheric circulation and boreal Canada summer drought variability of the past three centuries. J. Climate, 19 , 19221947.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Ya, and Coauthors, 2007: Potential forest fire danger over Northern Eurasia: Changes during the 20th century. Global Planet. Change, 56 , 371386.

    • Search Google Scholar
    • Export Citation
  • Hogg, E. H., , J. P. Brandt, , and B. Kochtubajda, 2005: Factors affecting interannual variation in growth of western Canadian aspen forests during 1951-2000. Can. J. For. Res., 35 , 610622.

    • Search Google Scholar
    • Export Citation
  • Kasischke, E. S., , and M. R. Turetsky, 2006: Recent changes in the fire regime across the North American boreal region - Spatial and temporal patterns of burning across Canada and Alaska. Geophys. Res. Lett., 33 , L09703. doi:10.1029/2006GL025677.

    • Search Google Scholar
    • Export Citation
  • Keetch, J. J., , and G. M. Byram, 1968: A drought index for forest fire control. U.S. Dept. of Agriculture, Forest Service Research Paper SE-38, Southeastern Forest Experiment Station, Asheville, NC, 32 pp.

    • Search Google Scholar
    • Export Citation
  • Kurz, W. A., , G. Stinson, , G. J. Rampley, , C. C. Dymond, , and E. T. Neilson, 2008: Risk of natural disturbances makes future contribution of Canada’s forests to the global carbon cycle highly uncertain. Proc. Natl. Acad. Sci. USA, 105 , 15511555.

    • Search Google Scholar
    • Export Citation
  • Lawson, B. D., , and G. N. Dalrymple, 1996: Ground-truthing the Drought Code: Field verification of overwinter recharge of forest floor moisture. Canada-British Columbia Partnership Agreement on Forest Resources Development, FRDA Rep. 268, Pacific Forestry Centre and Research Branch, Victoria, BC, Canada, 21 pp.

    • Search Google Scholar
    • Export Citation
  • Lawson, B. D., , O. B. Armitage, , and W. D. Hoskins, 1996: Diurnal variation in the Fine Fuel Moisture Code: Tables and computer source code. Natural Resources Canada, Canadian Forest Service and British Columbia Ministry of Forests, FRDA Rep. 245, Pacific Forestry Centre and Research Branch, Victoria, BC, Canada, 20 pp.

    • Search Google Scholar
    • Export Citation
  • Lee, T. C. K., , F. W. Zwiers, , and M. Tsao, 2008: Evaluation of proxy-based millennial reconstruction methods. Climate Dyn., 31 , 263281.

    • Search Google Scholar
    • Export Citation
  • Li, J., , F. Chen, , E. R. Cook, , X. Gou, , and Y. Zhang, 2007: Drought reconstruction for north central China from tree rings: The value of the Palmer drought severity index. Int. J. Climatol., 27 , 903909.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , S. Rutherford, , E. Wahl, , and C. Ammann, 2005: Testing the fidelity of methods used in proxy-based reconstructions of past climate. J. Climate, 18 , 40974107.

    • Search Google Scholar
    • Export Citation
  • Masters, A. M., 1990: Changes in forest fire frequency in Kootenay National Park, Canadian Rockies. Can. J. Bot., 68 , 17631767.

  • McAlpine, R. S., 1990: Seasonal trends in the Drought Code component of the Canadian Forest Fire Weather Index System. Forestry Canada, Petawawa National Forestry Institute, Chalk River, ON, Canada, Information Rep. PI-X-97 E/F, 10 pp.

    • Search Google Scholar
    • Export Citation
  • McKenney, D. W., , J. H. Pedlar, , P. Papadopol, , and M. F. Hutchinson, 2006: The development of 1901–2000 historical monthly climate models for Canada and the United States. Agric. For. Meteor., 138 , 6981.

    • Search Google Scholar
    • Export Citation
  • Mitchell, T. D., , and P. D. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25 , 693712.

    • Search Google Scholar
    • Export Citation
  • Mudelsee, M., 2000: Ramp function regression: A tool for quantifying climate transitions. Comput. Geosci., 26 , 293307.

  • Mudelsee, M., , and M. Alkio, 2007: Quantifying effects in two-sample environmental experiments using bootstrap confidence intervals. Environ. Model. Software, 22 , 8496.

    • Search Google Scholar
    • Export Citation
  • Muraro, S. J., , and B. D. Lawson, 1970: Prediction of duff moisture distribution for prescribed burning. Canada Department of Fisheries and Forestry, Canadian Forestry Service, Information Rep. BC-X-46, Pacific Forest Research Centre, Victoria, BC, Canada, 13 pp.

    • Search Google Scholar
    • Export Citation
  • Otway, S. G., , E. W. Bork, , K. R. Anderson, , and M. E. Alexander, 2007: Predicting sustained smouldering combustion in trembling aspen duff in Elk Island National Park, Canada. Int. J. Wildland Fire, 16 , 690701.

    • Search Google Scholar
    • Export Citation
  • Palmer, W. C., 1965: Meteorological drought. U.S. Department of Commerce, Weather Bureau Research Paper 45, 58 pp.

  • Régnière, J., 1996: A generalized approach to landscape-wide seasonal forecasting with temperature-driven simulation models. Environ. Entomol., 25 , 869881.

    • Search Google Scholar
    • Export Citation
  • Régnière, J., , and P. Bolstad, 1994: Statistical simulation of daily air temperature patterns in eastern North America to forecast seasonal events in insect pest management. Environ. Entomol., 23 , 13681380.

    • Search Google Scholar
    • Export Citation
  • Rodionov, S. N., 2006: Use of prewhitening in climate regime shift detection. Geophys. Res. Lett., 33 , L12707. doi:10.1029/2006GL025904.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., , and W. Skinner, 2004: Summer drought patterns in Canada and the relationship to global sea surface temperatures. J. Climate, 17 , 28662880.

    • Search Google Scholar
    • Export Citation
  • Stahle, D. W., , M. K. Cleaveland, , and J. G. Hehr, 1985: A 450-year drought reconstruction for Arkansas, United States. Nature, 316 , 530532.

    • Search Google Scholar
    • Export Citation
  • Stocks, B. J., and Coauthors, 2003: Large forest fires in Canada, 1959–1997. J. Geophys. Res., 108 , D18149. doi:10.1029/2001JD000484.

  • Taylor, S. W., , A. L. Carroll, , R. I. Alfaro, , and L. Safranyik, 2006: Forest, climate and mountain pine beetle outbreak dynamics in western Canada. The Mountain Pine Beetle: A Synthesis of Biology, Management, and Impacts in Lodgepole Pine, L. Safranyik and W. R. Wilson, Eds., Natural Resources Canada, 67–94.

    • Search Google Scholar
    • Export Citation
  • Thornthwaite, C. W., , and J. R. Mather, 1955: The water balance. Publ. Climatol., 8 , 186.

  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

    • Search Google Scholar
    • Export Citation
  • Turner, J. A., 1972: The drought code component of the Canadian Forest Fire Behaviour System. Environment Canada, Canadian Forest Service Publication 1316, Ottawa, ON, Canada, 14 pp.

    • Search Google Scholar
    • Export Citation
  • van der Schrier, G., , K. R. Briffa, , P. D. Jones, , and T. J. Osborn, 2006: Summer moisture variability across Europe. J. Climate, 19 , 28182834.

    • Search Google Scholar
    • Export Citation
  • Van Wagner, C. E., 1974: Effect of duff weight on drying rate. Bi-Mon. Res. Notes Can. For. Serv., 30 , 1112.

  • Van Wagner, C. E., 1985: Drought, timelag and fire danger rating. Proc. Eighth Conf. on Fire and Forest Meteorology, Publication 85-04, Detroit, MI, Society of American Foresters, 132–149.

    • Search Google Scholar
    • Export Citation
  • Van Wagner, C. E., 1987: Development and structure of the Canadian Forest Fire Weather Index System. Forestry Tech. Rep. 35, Canadian Forest Service, Ottawa, ON, Canada, 37 pp.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., , and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Weber, M. G., , and M. D. Flannigan, 1997: Canadian boreal forest ecosystem structure and function in a changing climate: Impact on fire regimes. Environ. Rev., 5 , 145166.

    • Search Google Scholar
    • Export Citation
  • Wilmore, B., 2001: Duff moisture dynamics in black spruce–feather moss stands and their relation to the Canadian Forest Fire Danger Rating System. M.S. thesis, University of Alaska, Fairbanks, 84 pp.

  • Xiao, J., , and Q. Zhuang, 2007: Drought effects on large fire activity in Canadian and Alaskan forests. Environ. Res. Lett., 2 , 044003. doi:10.1088/1748-9326/2/4/044003.

    • Search Google Scholar
    • Export Citation
  • Zar, J. H., 1999: Biostatistical Analysis. 4th ed. Prentice Hall, 929 pp.

  • View in gallery

    MDC [Eq. (11)] vs monthly means of the daily DC mean [Eq. (3)] for four locations across Canada over the 1900–2006 period. Months under analysis are May–October (n = 642 data points). A least squares linear regression (gray line) is shown along with the model R2. Refer to Table 1 for information on these locations.

  • View in gallery

    Spearman rank correlation coefficients squared (R2) computed between MDC and monthly means of the daily DC over 1900–2006 for each calendar month [May (M)–October (O)]. Refer to Table 1 for information on these locations.

  • View in gallery

    Climatological averages of MDC and of monthly means of the daily DC, per month, for the 1900–2006 period. Months under analysis are May (M)–October (O).

  • View in gallery

    Mean July MDC over Canada, 1901–2002 period. Scale (unitless) ranges from moist (blue, MDC < 139) to dry (yellow, MDC > 245).

  • View in gallery

    (a) Mean Canadian July MDC. The values are a spatial average of all grid cells, weighted with the cosine of the lat. The solid line shows a 10-yr low-pass filter (order 4); the dashed line shows long-term changes in mean as detected using the sequential algorithm method {Rodionov 2006; 30-yr window with a red noise [i.e., AR(1)] parameter set to 0.07}. The three driest and wettest years on record are 1961, 1955, 1958, and 1957, 1978, 1993, respectively (omitting the portion earlier than 1920). (b) Std dev.

  • View in gallery

    Percentage of area experiencing moderate (MDC > 120) to extreme (MDC > 280) July droughts in Canada. The solid line shows a 10-yr low-pass filter (order 4). The dashed line shows long-term changes in mean as detected using the sequential algorithm method [Rodionov 2006; 30-yr window with AR(1) parameters from top to bottom set to 0.24, 0.15, 0.00, and 0.00].

  • View in gallery

    Linear trends of July MDC for (a) 1901–2002 and (b) 1951–2002 as detected using the Spearman rank correlation coefficient (for skewness bias). Correlation scale ranges from decreasing drought severity (blue) to increasing dryness (yellow). Trends significant at the 5% level are indicated by black dots.

  • View in gallery

    Regional July MDC records. Coordinates refer to upper-left and lower-right corners of the gridcell domains. The solid line shows a 10-yr low-pass filter (order 4). The dashed line shows long-term changes in mean as detected using the sequential algorithm method [Rodionov 2006; 30-yr window with AR(1) parameters set to (a) 0.00, (b) 0.13, and (c) 0.00]. The dotted line shows linear trends of MDC for 1901–2002 (P value is indicated; all three regression models passed the Durbin–Watson, normality, and constant variance tests): the linear trends are, respectively, (a) −0.37, (b) +0.20, and (c) −0.26 MDC (units yr−1).

  • View in gallery

    (a) Correlation pattern between the total annual area burned in Canada by large forest fires (size > 200 ha) and July MDC (1959–99). Correlations significant at the 5% level are indicated by black dots. (b) Total annual area burned in Canada predicted from July MDC using the composite-plus-scale analysis, plotted against the observations.

  • View in gallery

    July MDC records for (a) Fort McMurray and (b) Thunder Bay (Table 1) before and after applying an overwintering adjustment using program SimMDC [Eq. (4)]. Overwintering empirical constants a and b were subjectively set to 0.75 (area subject to bare ground in winter) and 0.50 (well-drained soils), respectively (see Lawson and Dalrymple 1996). The Spearman R2 between MDC records over the 1901–2002 period is shown. Linear trends of MDC are also shown (P < 0.01 for the Thunder Bay location).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 128 128 31
PDF Downloads 79 79 17

Summer Moisture and Wildfire Risks across Canada

View More View Less
  • 1 Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, Québec, Quebec, Canada
  • | 2 Natural Resources Canada, Canadian Forest Service, Great Lakes Forestry Centre, Sault Ste. Marie, Ontario, Canada
© Get Permissions
Full access

Abstract

The Fire Weather Index System has been in use across Canada for the past 30 years in the daily operations of fire management agencies. As part of this system, the Drought Code (DC) was developed to act as a daily index of water stored in the soil. A major obstacle to the completion of climate risk analyses on the DC is that lengthy series of daily temperature and precipitation are not available for large portions of the circumboreal forest. Here the authors present a methodological modification to the daily DC to allow its approximation using monthly data. This new Monthly Drought Code (MDC) still retains its ability to capture moisture trends in deep organic layers. On the basis of high-resolution temperature and precipitation data, an analysis of summer moisture availability across Canada over 1901–2002 is presented. The driest periods on record were from the 1920s to the early 1960s, with the driest years being 1955, 1958, and 1961. The wettest period was from the mid-1960s to the 1980s. For the century-long period, drying was statistically significant in northern Canada. Locations south of the Hudson Bay, in the eastern Maritimes, and in western Canada recorded a trend toward decreasing dryness. When analyzed over 1951–2002, trends could hardly be distinguished from the (multi) decadal variability. Annual values of a spatial average of all July MDC grid cells showed an excellent fit against fire statistics: 63% of the variance in the Canada-wide annual area burned from 1959 to 1999 was explained by summer moisture availability.

Corresponding author address: Martin P. Girardin, Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, 1055 du P.E.P.S., P.O. Box 10380, Stn. Sainte-Foy, Québec, QC G1V 4C7, Canada. Email: martin.girardin@rncan.gc.ca

Abstract

The Fire Weather Index System has been in use across Canada for the past 30 years in the daily operations of fire management agencies. As part of this system, the Drought Code (DC) was developed to act as a daily index of water stored in the soil. A major obstacle to the completion of climate risk analyses on the DC is that lengthy series of daily temperature and precipitation are not available for large portions of the circumboreal forest. Here the authors present a methodological modification to the daily DC to allow its approximation using monthly data. This new Monthly Drought Code (MDC) still retains its ability to capture moisture trends in deep organic layers. On the basis of high-resolution temperature and precipitation data, an analysis of summer moisture availability across Canada over 1901–2002 is presented. The driest periods on record were from the 1920s to the early 1960s, with the driest years being 1955, 1958, and 1961. The wettest period was from the mid-1960s to the 1980s. For the century-long period, drying was statistically significant in northern Canada. Locations south of the Hudson Bay, in the eastern Maritimes, and in western Canada recorded a trend toward decreasing dryness. When analyzed over 1951–2002, trends could hardly be distinguished from the (multi) decadal variability. Annual values of a spatial average of all July MDC grid cells showed an excellent fit against fire statistics: 63% of the variance in the Canada-wide annual area burned from 1959 to 1999 was explained by summer moisture availability.

Corresponding author address: Martin P. Girardin, Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, 1055 du P.E.P.S., P.O. Box 10380, Stn. Sainte-Foy, Québec, QC G1V 4C7, Canada. Email: martin.girardin@rncan.gc.ca

1. Introduction

The Canadian Forest Fire Weather Index (FWI) System (Van Wagner 1987) has been in use across Canada for the past 30 years in the daily operations of fire management agencies (http://cwfis.cfs.nrcan.gc.ca/). The FWI System uses daily weather observations (temperature, rainfall, relative humidity, and wind velocity) to estimate the moisture content of three different fuel classes and uses these to generate a set of relative indicators of potential rate of fire spread, fire intensity, and fuel consumption. It was in the 1920s that researchers in Canada began relating day-to-day susceptibility of forest fires to weather using rating systems for fire danger (Van Wagner 1987). In subsequent decades, four different fire danger systems were developed for various regions of Canada, with each version based on field research on the fuel types of local importance. By the late 1960s, there was an increasing demand by forest fire control agencies for the development of a new common set of fire danger–rating indexes for the entire country; the result was the FWI System. This system retained a solid link with previous approaches by building on the best features and adding new components where necessary (Van Wagner 1987). Adaptation or adoption of the FWI System has now begun in foreign countries, as methodologies have been developed to electronically gather daily weather data and produce daily fire weather and fire behavior potential maps for large portions of northern Europe and northern Asia (Fire Ecology Research Group 2005; de Groot et al. 2007a). In support of Canada’s National Forest Carbon Monitoring, Accounting and Reporting System, implementation of the FWI System in the Carbon Budget Model of the Canadian Forest Sector (CBM–CFS3) has also begun (de Groot et al. 2007b). In this achievement, forest floor consumption models for each of the fuel types used across the country rely on the FWI System.

As part of the FWI system, the Drought Code (DC) was originally developed to act as a daily index of water stored in the soil (Turner 1972). Initial studies (e.g., Muraro and Lawson 1970) showed that the DC tracked moisture variations in deep compact duff layers reasonably well, and over time it has come to be used as a daily indicator of the moisture content of deep layers of the forest floor (about 20 cm deep on average) or other heavy (large diameter) fuels. Further studies have shown that the DC follows moisture in deep organic layers across a range of stands (Lawson et al. 1996; Wilmore 2001; Abbott et al. 2007; Otway et al. 2007). While the actual relationship between absolute moisture content of the various layers studied and the DC has been found to vary somewhat across stands of significantly different structure, these studies have shown it is a useful indicator of daily moisture change in the deeper layers of the forest floor and, consequently, a good indicator of fire-conducive droughts.

Aside from the daily operations of fire management agencies, the DC has recently been used in the detection of climate change impacts on wildfire risk across Canada (Amiro et al. 2004; Girardin et al. 2004b) and for forecasting wildfire risks over the twenty-first century (Flannigan et al. 2005; Bergeron et al. 2006; Girardin and Mudelsee 2008). The rational behind these efforts is that drought in deep organic material is a determining factor for forest fire severity, as dry conditions allow deep burning and smoldering. Therefore, because of its influence on fire severity, drought is an important agent for postfire ecosystem structure and function through indirect impacts on underground plant roots, reproductive tissues, and soil seed banks (Weber and Flannigan 1997). Climatically induced shifts in drought regimes could subsequently alter forest succession pathways and induce large changes in forest composition (hardwood versus conifers), structure (even versus uneven aged stands), and biodiversity (e.g., de Groot et al. 2003). Also of concern are potential changes in emissions from biomass burning, as the contribution of fires to the net annual atmospheric carbon balance of forest landscapes is not negligible (e.g., Amiro et al. 2001; Coursolle et al. 2006; Kurz et al. 2008). Potential loss of harvestable trees to fires and the expense of fire suppression, which in Canada already costs on average 0.5 billion dollars per year (CDN), must also be considered (period 1985–2006; Canadian Interagency Forest Fire Centre 2007, personal communication). Predicting the risk (probability) of extreme fire-conducive droughts in the regional domain and on midterm time scales (decades) can help forest managers and may provide valuable insight into the future contribution of Canadian forests to the atmospheric greenhouse gas balance (Kurz et al. 2008).

A major obstacle to the completion of climate risk analyses on the DC is that lengthy series of daily temperature and precipitation, which are necessary for the computation of the DC, are not available for large portions of the circumboreal forest (Groisman et al. 2007). As in many long-term climatology studies, particularly those involving the use of long-term climate records or output from general circulation models, modelers must rely on monthly summaries of temperature and total precipitation (among other variables). Here we present a methodological modification to the daily DC to allow its approximation using monthly data. Our approach seeks to provide information on long-term changes in patterns of moisture content of deep layers of the forest floor that may relate to fire activity and observed changes in ecological functions and structures. We recognize that drought and its severity have been numerically defined in various ways, for example, the daily Keetch–Byram drought index (Keetch and Byram 1968), as used in the study by Groisman et al. (2007), the monthly climate moisture index, as used by Hogg et al. (2005), and the monthly Palmer drought severity index (PDSI; Palmer 1965), as used by Dai et al. (2004). Van Wagner (1985) has compared these indexes; however, our work focused on the DC, given that it is the heavy fuel moisture index most commonly used by provincial and national fire agencies across Canada. The current paper is organized as follows: section 2 describes the original formulation of the DC and presents the new Monthly Drought Code (MDC). Section 3 describes the sources of meteorological data used to calculate the DC and MDC in this study. Section 4 presents an example of how the MDC can be used to evaluate changes in summer moisture availability across Canada over the twentieth century and a validation of the MDC formulation against wildfire statistics. Results are presented in section 5, the study is discussed in section 6, and a summary is presented in section 7.

2. Methods

a. Calculation of the daily Drought Code

In Turner’s (1972) original formulation of the DC model, called the soil moisture index (SMI), SMI ranged from 0 to 800 with each unit representing 1/100 of an inch of water (0.254 mm) up to a maximum holding capacity of 8 in. (or approximately 203 mm) for a deep organic soil with a fuel load of 50 kg m−2. Originally these physical limits were carried into the FWI System’s version of the DC (Van Wagner 1974); however, subsequent studies of deep, thick, organic layers in the boreal forests of Canada showed that fuel loads were closer to around 25 kg m−2 with gravimetric moisture contents reaching about 400% at saturation, which corresponds to a water holding capacity of 100 mm (or about 4 in.). These latter values describing the physical characterization of the layer modeled by the DC are currently used in the FWI System. The original scale of the SMI remains internal to the calculation of the index itself; however, a generic conversion to absolute moisture content requires a unit of SMI to be equivalent to 0.5% moisture content (gravimetric). For the sake of comparisons between the daily and the new monthly model, we will briefly present the calculation method for the daily DC value as it is used throughout Canada.

The DC is a simple moisture bookkeeping system that uses an estimate of maximum daily temperature to estimate a day’s potential evapotranspiration, following the method of Thornthwaite and Mather (1955), and daily rainfall to track increases in wetness of the layer. Potential evapotranspiration E (which in the DC has been scaled to be a unitless quantity) during day d is given by
i1558-8432-48-3-517-e1a
where T is the 1200 LST observation of air temperature (°C) and Lf is the standard day length adjustment factor. Operational fire weather observations are traditionally made at 1200 LST, and so Turner (1972) and Van Wagner (1987) used a simple estimate of the difference between noon and maximum temperature in the formulation of the DC: maximum temperature was estimated to be noon observation plus 2.8°C. Thus, if observations of the maximum temperature (Tmx) are available, Eq. (1a) is replaced with
i1558-8432-48-3-517-e1b
The Lf factor varies by month and is equal to January–March = −1.6, April = 0.9, May = 3.8, June = 5.8, July = 6.4, August = 5.0, September = 2.4, October = 0.4, and November and December = −1.6. Only values of T > −2.8°C are used in the model; if temperature is less than this value, then it should be set to −2.8°C. Similarly, using the formula in Eq. (1b), only values of Tm > 0 should be used. If E becomes negative because of negative values of Lf , then it should be set to 0.
The effect of any rain r (mm) is added to yesterday’s value of the Drought Code, DC0 (unitless), and the moisture equivalent in the layer after rain Qr (unitless) is calculated as
i1558-8432-48-3-517-e2a
or
i1558-8432-48-3-517-e2b
In these equations, r is the total daily rainfall (mm) and is reduced to an effective rainfall after canopy and surface fuel interception using REFF = 0.83r − 1.27. If Qr > 800 in Eq. (2a), then Qr = 800.
The new Drought Code value is then calculated as
i1558-8432-48-3-517-e3
There are no absolute guidelines as to the meaning of the DC values but, generally speaking, values below 200 are considered low while values around 300 may be considered moderate in most parts of the country. A rough rule of thumb used by fire managers across the country is that a DC rating of 400 or more indicates that fire will involve burning of deep subsurface and heavy fuels. The DC generally peaks in mid- to late August, beyond which it either declines or maintains the same value (during extreme years with late-season fires, this does not always hold true) (McAlpine 1990; Girardin et al. 2004b). The reversal in August is only attributed to a change in day length, and is not a function of seasonal precipitation.
In boreal forests with considerable overwinter precipitation, the DC is considered saturated (reset to zero) because of melting snowpack in the spring and it is typically assigned a value of 15 (indicating 3 days of drying under standard conditions) at the date of weather station startup. In regions or cases where this assumption cannot be counted on (e.g., western Canada), an overwintering adjustment is calculated and the DC starting value (DCs) for the season may be obtained from
i1558-8432-48-3-517-e4
where Qf is the final autumn moisture equivalent (normally set at 31 October: the end of the official fire season in many jurisdictions), rw is winter precipitation (mm), Qs is the starting spring moisture equivalent, and a and b are empirical constants representing the carryover fraction of last autumn’s moisture and the effectiveness of winter precipitation in recharging moisture reserves in spring, respectively (Van Wagner 1987; Lawson and Dalrymple 1996). The Qf is calculated from the conversion equation
i1558-8432-48-3-517-e5
where DCf is final autumn DC. The spring starting value DCs is obtained from the conversion equation
i1558-8432-48-3-517-e6
Because of insufficient data on the soil properties needed to determine constants a and b at each location, we did not apply the “overwintering” model in this work. In this analysis, the DC layer was assumed to be fully recharged on 30 April of each year. We did, however, carry out a sensitivity analysis (described later) to examine the effect of this adjustment.

b. Monthly version of the drought code

With the long response time in the DC model (e.g., 62 days at 15°C, 44 days at 30°C), and the linear relationships between both temperature and evapotranspiration as well as rainfall and SMI, it seems logical that the calculation of the DC could be generalized using monthly means and still retain its ability to capture moisture trends in deep organic layers. In fact, the Thornthwaite model of potential evapotranspiration used as the basis for the DC was developed for monthly mean temperature and total precipitation; thus it is not too great a leap to develop a monthly Drought Code value using the existing formula. We propose the following methodological modification to the daily DC to allow its approximation using monthly data.

Potential evapotranspiration E (unitless quantity) over month m is given by
i1558-8432-48-3-517-e7
where Tmx is the monthly mean of daily maximum temperatures (°C), Lf is the standard day length adjustment factor, and N is the number of days in the month. In this formula, instead of using an approximate adjustment for the difference between noon and maximum temperature, we use the average maximum daily temperature directly. Only values of Tmx > 0°C are used in the model; if a temperature mean is less than this value, then it should be set to 0°C. If Em becomes negative because of negative values of Lf , then it is set to 0.
To try to reduce the potential bias that may arise if the forest floor becomes saturated in spring after heavy rain, we assume that the total monthly rainfall (rm) occurs in the middle of the month. While it is of course not generally true, this assumption seems more appropriate than placing rainfall either at the start or end of the calculation period. This assumption should only influence early spring values when the forest floor is near saturation. By assuming that only drying takes place over the first half of the month, we calculate
i1558-8432-48-3-517-e8
The effect of rain is added to the DC calculated for the middle of the month, DCHALF:
i1558-8432-48-3-517-e9
where RMEFF = 0.83rm. If Qmr > 800 in Eq. (9), then Qmr should be set equal to 800.
The MDCm value (unitless) would then be calculated as
i1558-8432-48-3-517-e10
MDCm represents an estimate of the DC value at the end of the month for which total rainfall and mean temperature apply. To find a mean value for the month, MDC0 (the MDC from the end of the previous month) and MDCm should be averaged:
i1558-8432-48-3-517-e11
The MDC quantity becomes more easily comparable to monthly means of the daily DC obtained from Eq. (3). When calculating the next month’s MDC, the value of MDCm from the previous month then becomes the new MDC0. As in the DC formula, the MDC layer was assumed in this work to be fully recharged in April of each year.

3. Temperature and precipitation data

We compared the original formula of the DC [Eq. (3)] with the modified calculation [Eq. (11)] using the following data. Daily maximum temperatures (°C) and precipitation (mm) were collected over the period 1900–2006 at four locations across Canada (Table 1). Missing daily data were obtained for the location by interpolating daily weather data from nearby weather stations using the weather generator of BioSIM, which adjusts the weather data for differences in latitude, longitude, and elevation (Régnière and Bolstad 1994; Régnière 1996). Monthly means of daily maximum temperature and total monthly precipitation were obtained from the average and sum of the daily quantities, respectively.

4. MDC across Canada over 1901–2002

In our application of the MDC to evaluate long-term changes in climate and wildfire risk in Canada, we used Climatic Research Unit (CRU) total monthly precipitation and monthly maximum temperature data (CRU TS 2.1 data; Mitchell and Jones 2005) interpolated into a 0.5° latitude × 0.5° longitude grid covering the 1901–2002 period. The MDC over Canada was calculated for each grid cell (total of 6171 cells) and the percentage of areas experiencing July drought (defined as any areas where July MDC exceeds a given threshold) was computed for each year over 1901–2002. Long-term changes in the percentage index were detected using regime shift detection on a sliding 30-yr window with correction for serial persistence in data (Rodionov 2006). This analysis allows us to verify that changes in the mean from one period to another are not just the manifestation of a red noise [i.e., AR(1)] process (probability σ = 0.05; outliers weight parameter = 6; IP4 method for red noise correction). Regime shifts detection was also carried out on the countrywide mean of MDC and on regional records of MDC.

Linear trends of MDC for each grid cell and on regional records of MDC were also examined for 1901–2002 and 1951–2002 using least squares linear regressions (Zar 1999). Goodness of fit was described by the correlation coefficient R. Significance was tested against the null hypothesis that the trend is different from zero, using a variant of the t test with an estimate of the effective sample size that takes into account the presence of serial persistence in data (von Storch and Zwiers 1999, their sections 8.2.3 and 6.6.8). When necessary, MDC data were ranked prior to analysis to satisfy the normality distribution requirement in model residuals (Zar 1999).

One of the prerequisites for the development of the MDC was that it could be used reliably as an approximation of fire-conducive climate variability over broad areas and over time. This implies that MDC should correlate well with the area burned at locations where wildfire risk is high (i.e., northwestern Canada; Stocks et al. 2003). As a means of verification, large forest fire (size > 200 ha) statistics from the Canadian Large Fire Data Base (LFDB; Stocks et al. 2003) were used in a composite-plus-scale analysis (Mann et al. 2005; Lee et al. 2008). The LFDB contains information on start location, estimated ignition date, cause, and final size of each fire. These large fires represent only a very small percentage of the fires in Canada (about 3%) but account for most (>97%) of the area burned (Stocks et al. 2003). First, fires were compiled for Canada, producing a time series of annual area burned (AAB, in ha) covering the period 1959–99. Shapiro–Wilk normality tests indicated right-skewness in AAB frequency distributions (P = 0.000, where P values below 0.05 are considered small enough to declare the fit with the normal curve poor). The logarithmic transformation (LOG) was found to provide an adequate data transformation to meet the normality requirement (after LOG transformation: P > 0.200). Next, Spearman rank correlation coefficients k were computed between July MDC grid cells and the AAB time series. The annual values xj of a spatial average of all July MDC grid cells, weighted with the coefficients k and the cosine of latitude, were then linearly fitted with the LOG-transformed AAB yj using least squares linear regression (Zar 1999):
i1558-8432-48-3-517-e12
where β1 is the regression coefficient and εj is the error. Model residuals were tested for the presence of serial persistence, of a constant variance, and of a normal distribution (Zar 1999). Finally, the LOG-transformed AAB estimates obtained from Eq. (12) were translated into arithmetic units (ha) by applying a conversion function with correction for the skewness bias in AAB (Baskerville 1971; Girardin et al. 2006a):
i1558-8432-48-3-517-e13
where σ̂2 is the variance of LOG-transformed AAB estimates ŷj (log ha), and υ̂j is the estimated AAB (ha).

The MDC datasets for Canada may be obtained by contacting Martin P. Girardin. A proprietary Statistical Analysis Software (SAS) software source code and an executable file (program SimMDC, compiled using Visual C++ 8.0) for computation of MDC on CRU TS 2.1 data are also available on request.

5. Results

a. DC versus MDC

Biases associated with changes from daily to monthly calculations of the DC were explored by regressing time series of monthly MDC against those of monthly means of the daily DC at four locations across Canada (period of analysis is 1900–2006). Averages and variances of the monthly means of the daily DC and of the MDC for the whole period were also computed (hereinafter referred to as climatological average and variance statistics). Shapiro–Wilk normality tests were also computed to verify that the distributions of monthly averages of the daily DC and of MDC did not depart from normality. In this test, P values of the order of 0.050 were considered small enough to declare the fit with the normal curve poor.

Globally, we found that the MDC yielded an excellent fit against the monthly means of the daily DC, with model R2 ranging from 0.87 to 0.95 (Fig. 1). The amount of shared variance was lower in locations characterized by higher precipitation regimes (i.e., Bagotville and Thunder Bay). An analysis at the monthly time step (Fig. 2) showed that the prediction error was essentially distributed during late spring and early summer months (May–July). For locations in central and western Canada, for instance, the amount of shared variance remained above 70% in all months of the year, and in some circumstances reached 93%. In eastern locations, however, R2 values were as low as 0.58 (Fig. 2).

Shapiro–Wilk normality tests allowed us to identify a tendency of the MDC calculation to yield a skewed distribution of the monthly values, more so than the DC calculation (Table 2). The bias is particularly important in the eastern locations under study (i.e., Bagotville and Thunder Bay) (Table 2). Analyses of the long-term monthly climatological averages further allowed us to identify a systematic underestimation of drought severity at the Fort McMurray location during all months of the year (Fig. 3; Table 2). Underestimation of drought severity was also apparent at the Dauphin location during late-summer months. At all locations, variance was systematically underestimated by the MDC (Table 2). In section 5b, we present an application of the MDC to the analysis of summer moisture availability across Canada over the twentieth century.

b. MDC values for Canada

The aim of this section is to present an initial analysis of maps and trends of the July MDC for Canada for the period of 1901–2002. The time lag of the DC, and hence of the MDC, is long enough so that July records integrate the influence of the two previous months [i.e., May and June (see Girardin et al. 2004b)]. Over 82% of the area burned (or 76% of all fires of size > 200 ha) in Canada from 1959 to 1999 did so during these three months (Stocks et al. 2003). Girardin et al. (2006a) also found that the July DC was a good predictor of area burned and of the number of fires on the Boreal Shield, explaining about 45% of the variance in their annual calculations.

First, for the four locations outlined in Table 1, July MDC computed from meteorological station data and from CRU data were entered in least squares linear regressions (on ranked data) to verify the existence of potential deviation. Goodness of fit was described by the coefficient of determination (R2). Globally, the MDC values computed from CRU data did not differ much from data obtained using the weather generator of BioSIM, except in periods when data were scarce (prior to the 1920s). For the 1901–2002 period, the amount of shared variance varied from 50% at Bagotville to 67% at Dauphin. Over 1951–2002, it ranged from 93% at Fort McMurray to 68% at Bagotville and 51% at Thunder Bay. For the whole 1901–2002 period, Durbin–Watson statistics revealed small trends in model residuals at the Fort McMurray and Thunder Bay locations (D-W > 1.48; AR(1) < 0.22); no trend was detected over the shorter period of 1951–2002.

The map of the 1901–2002 mean of July MDC shows a distinct geographic pattern (Fig. 4). Locations along the Pacific Coast and in northeastern Canada are characterized by climates that are less conducive to fires (low July MDC), whereas locations east of the Rockies and in the southern interior of British Columbia are found to be more drought-prone (Fig. 4). It is important to remember that in our analysis we have not accounted for overwintering effects of the DC. Therefore, in regions where drought carries over from year to year (discussed later), MDC values may appear lower than mean DC values commonly encountered by fire agencies.

As demonstrated in Fig. 5a, the July MDC averaged over Canada experienced strong decadal-scale variability. The 1920s to the early 1960s stand out as a persistent and exceptionally dry period, whereas the mid-1960s to the late 1980s were relatively humid. The percentage of area experiencing extreme drought (MDC > 280 units) has not changed over the course of the past century (Fig. 6). In contrast, the percentage of area affected by moderate drought (MDC > 120) declined around the 1960s, and increased back to previous levels during the 1990s (Fig. 6). Note that the upward trend in the first decade (Figs. 5, 6) could partially be related to a bias associated with a decrease in the variance in precipitation data. Meteorological data at high spatial resolution are not available for this period, and data were interpolated from surrounding stations. This procedure tends to suppress extreme values in the datasets (van der Schrier et al. 2006). There was otherwise no evidence of change in spatial variability (standard deviation) of MDC over the 1901–2002 period (Fig. 5b).

Spatial patterns of July MDC changes for the 1901–2002 and 1951–2002 periods are illustrated in Fig. 7. For the century-long period, drying is statistically significant in northern Canada. As indicated by the regime shift detection, drought severity has increased by nearly 20 units in northwestern Canada since the 1980s (Fig. 8b). The trend at that location was also significant using a least squares linear fit (P = 0.030). Conversely, locations south of the Hudson Bay, in the eastern Maritimes, and in western Canada recorded a trend toward decreasing dryness (Fig. 7). This holds for individual grid cells and for regional records (Figs. 8a,c; least squares linear fit: P < 0.050). Patterns of changes over 1921–2001 (not shown) were similar to those of 1901–2002, albeit the trend toward increasing dryness northwest of the Hudson Bay failed to be significant at the 5% level (passed the 10% level).

When analyzed over the recent period, only a few locations showed significant changes in drought severity (less than 4% of all cells, hence less than expected by chance alone). The trend over western Canada for individual grid cells was not as spatially extensive as that detected for the century-long period, and did not hold for western Canada as a whole (Fig. 8a; least squares linear fit over 1951–2002: P = 0.188). In eastern Canada, trends in drought severity during the recent period also failed to pass the 5% significance level. This holds for individual grid cells (Fig. 7) and for eastern Canada as a whole, despite a positive shift in mean MDC in 1995 (Fig. 8c; least squares linear fit: P = 0.191). Drought severity in northwestern Canada recorded a significant increase from 1951 to 2002 (least squares linear fit: P = 0.048) with a significant change in the mean MDC in 1983 (Fig. 8b; regime shift detection).

In general, the maps indicate that differences in trends between locations and periods can be important, and that trends can hardly be distinguished from the (multi) decadal variability when analyzed over a short period. In section 5c, we demonstrate that the annual area burned in Canada from 1959 to 1999 was closely related to summer variability in the moisture content of deep layers of the forest floor.

c. MDC and area burned in Canada

The correlation pattern in Fig. 9 shows how the annual area burned in Canada is related to July MDC. Over 18% of the grid cells showed significant correlation with AAB, which is more than expected by chance alone. Significant correlations were found across broad areas of the Canadian boreal forest, including areas of the Boreal Shield (southwest of the Hudson Bay), Taiga Shield West, Taiga Plains, Boreal Cordillera, and Taiga Cordillera (all located in northwestern Canada). The west-to-east pattern highlighted by the positive correlation field was essentially located in ecozones where about 75% of AAB occurred during the period of analysis. Furthermore, the AAB estimates obtained from composite-plus-scale analysis [Eq. (12)] show excellent fit against the observations: 63% of the variance (P < 0.001) in LOG-transformed AAB observations was recovered by the estimates (Table 3). If scaled into arithmetic units [Eq. (13)], it is 60% of the variance that is recovered (Fig. 9). The positive linear trend that characterizes AAB observations over 1959–99 (P < 0.001) is also well recovered by our estimates (albeit weaker at P = 0.028). In comparison, 22% of the variance (P = 0.002) in the LOG-transformed AAB was explained by the percentage of area experiencing extreme July drought (MDC > 280; Fig. 6), and 16% (P = 0.011) by the percentage of area experiencing MDC > 200. In another analysis, we found that the substitution in Eq. (12) of AAB for the annual number of large forest fires in Canada (FireOcc) also yielded excellent results: 67% of the annual variance in Canadian FireOcc was recovered by the MDC data.

6. Discussion

a. Predictive skills of the MDC

Globally, we found that the MDC yields an excellent fit against the monthly means of the daily DC. The goodness of fit was lower at locations characterized by higher precipitation regimes (i.e., eastern Canada) and during late spring and early summer months. The calculated MDC may be different from the calculated value of monthly mean of daily DC for the same dataset for several reasons. First, in early spring, the DC can become saturated after a heavy rainfall, and any excess moisture that falls as rainfall at that point is lost to the system as runoff (Girardin et al. 2004b). In the daily model it is assumed that all rainfall occurred prior to the drying for that day (i.e., any rainfall measured occurred overnight). The monthly model must use a date with which to associate rainfall, and we have chosen the middle of the month to try to reduce bias. This bias may explain the lower R2 observed in spring in eastern locations where precipitation is more abundant. Second, there tends to be some differences between MDC and monthly means of the daily DC because the effective rainfall function (REFF) differs. We have maintained the reduction in rainfall that is used in the DC model (83%), which overall should lead to reasonable correspondence between effective rainfall in the daily and monthly models. However, in the daily model, light rain (≤2.8 mm) does not influence the value of the DC [Eq. (2b)]. When summarized over a month, however, the character of these daily rain events gets lost. Thus, a series of light rainfalls of <2.8 mm each but totaling over 2.8 mm would have no effect on the daily DC value, but their cumulative effect would lower the MDC if we used the same rainfall function in the two models. Therefore, because the MDC includes these rainfalls, we would expect it to slightly underestimate the corresponding monthly mean of daily DC values in general. Furthermore, the intercept term in the rain function in Eq. (2a) (the value of 1.27 mm that is subtracted from rainfall) is applied to each rainfall case above 2.8 mm in the daily model; if we used the original rainfall function in the monthly model it would be applied only once, and as such it would have a very small impact.

In general, the influence of these light rainfalls included in the monthly rainfall total could lead to MDC underestimating DC as demonstrated in Table 2. Nonetheless, the high R2 values obtained in the monthly correlation analyses tend to suggest that year-to-year variability in monthly quantities remains relatively unaffected by the various sources of error through most of the year (except in spring). Therefore, the modified calculation model is well suitable for generating summer drought indexes that can be submitted for climate risk analysis (i.e., detection of long-term changes in mean, in the occurrence rate of extreme drought events, or in drought spatial extents). However, the proposed MDC formula tends to frequently produce a skewness bias in monthly time series, more so than the DC formula. For this reason, nonparametric tests (Zar 1999), or parametric tests that correct the skewness bias (Mudelsee 2000; Mudelsee and Alkio 2007), are more likely to be necessary when analyzing these data (albeit the bias is less important in the summer-months MDC time series).

The applicability of the MDC formula to wildfire studies is well demonstrated by our composite-plus-scale analysis. Indeed, the annual values of the spatial average of all July MDC grid cells showed an excellent fit against Canadian fire statistics: 60% of the variance in the Canada-wide annual area burned and 67% in the annual number of large forest fires from 1959–99 was explained by July moisture availability (Fig. 9). The goodness of fit of MDC to the area burned and number of fires is better than reported in other studies that made use of the DC (e.g., Flannigan et al. 2005; Girardin et al. 2006a). This is largely due to the fact that using the countrywide aggregate provided a “statistical smoothing” of the fire and weather statistics, attenuating local extremes and other factors affecting fire behavior (e.g., probability of ignition), and enhancing climate patterns that were synoptic and seasonal in scale.

b. Overwintering adjustment

Drought and its severity have been numerically defined in various ways using indices that integrate temperature, precipitation, and other variables that affect evapotranspiration and soil moisture. For instance, the Palmer drought severity index is a measure of regional moisture availability that has been used extensively to study droughts and wet spells in the contiguous United States and in other parts of the world (Briffa et al. 1994; Dai et al. 2004; van der Schrier et al. 2006). In Canada, the summer PDSI averaged for the entire country indicates dry conditions during the 1940s and 1950s, generally wet conditions from the 1960s to 1995, and dry conditions after 1995 (Shabbar and Skinner 2004). This long-term pattern of drying and wetting inferred from the PDSI is qualitatively consistent with our countrywide averages of MDC grid cells (presented in Fig. 5). In fact, the two records integrate the same field capacity of 100 mm, which may explain the close agreement. However, the MDC time series show more year-to-year variability (less red noise) than the PDSI (e.g., van der Schrier et al. 2006) due to the fact that, unlike the PDSI, the formula of the MDC does not overwinter drought from one year to the next, nor does it integrate winter precipitation. Although overwintering capacity is implemented in the PDSI, the calculation of the index does not account for precipitation in the form of snow; it assumes all precipitation to be in the liquid phase and made available to the soils in winter (van der Schrier et al. 2006). Also, the PDSI calculation takes no account of changes in the potential water-holding capacity of the soils when the ground freezes (van der Schrier et al. 2006).

While the assumption of a fully recharged MDC layer in spring may hold for large parts of the country (where total November–April precipitation exceeds 200 mm; Table 1; Lawson and Dalrymple 1996), in western and northern Canada, overwintering is commonly required. Girardin et al. (2006a) found that on the Boreal Shield, the area burned and the number of large forest fires was significantly correlated with the previous year’s DC values. The authors attributed this lagged effect to a potential overwintering of drought. Xiao and Zhuang (2007) have also linked the annual area burned in Canada to the percentage of area affected by droughts as inferred from the PDSI (as done in this study). Their correlation (R2 = 0.24) was similar to ours (R2 = 0.22) despite the seasonal carryover implemented in the Dai et al. (2004) PDSI data. So seemingly the addition of a drought overwintering does not improve the fit between soil moisture availability and countrywide area burned.

As a means of evaluating the importance of the previous year’s moisture and winter precipitation on current year wildfires, we recorrelated the countrywide AAB record with the previous year’s MDC and winter (November–April) precipitation for each grid cell (analysis not shown). Less than 2% of all cells showed significant positive correlation with the previous year’s July MDC (results not shown); these correlations were essentially distributed across the Boreal Shield and Boreal Plains ecozones. Negative correlations were recorded in 4.7% of all grid cells; these were distributed in regions not (or minimally) affected by fire. These results were no better when the previous year’s October MDC was used as a predictor: 1.5% of all grid cells showed significant positive correlation with AAB and 6.7% showed negative correlation (distributed in the vicinity of the Hudson Bay area in regions minimally affected by fires). Results on winter precipitation were not conclusive either. That being said, it is possible that at regional levels the local fire activity is closely related to the previous year’s drought conditions. Follow-up analyses should attempt to apply the DC overwintering model [Eq. (4)] to the MDC across broad areas. This parameterization will require the development of datasets for each of the 0.5° × 0.5° grid cells on soil moisture depletion in fall and winter, drainage class (poor, moderate, or well drained), time of ground frost, possibility of infiltration of the melting snowpack, and possibility of rapid runoff prior to melting of ground frost (Lawson and Dalrymple 1996). A sensitivity analysis in which empirical constants [Eq. (4)] were subjectively set suggested that the evaluation of long-term trends in MDC for the past century was not critical to the overwintering adjustment (Fig. 10). The main source of difference was in the presence of a red noise bias in the adjusted MDC. In contrast, the magnitude of the climatological average was shown to be severely impacted in regions of insufficient winter precipitation: adjusted MDC at the Fort McMurray location (Fig. 10a) nearly reached the 280 mark, which is considered extreme for July. While the adjustment might not significantly impact evaluation of year-to-year July MDC variability and synoptic-scale features, the information may be valuable for the characterization of drought regimes in northwestern locations. An overwintering adjustment may also be critical for evaluation of the impact of climate change on fire regimes if important changes in winter precipitation amounts should occur.

c. Linear trends of MDC across Canada

Linear trend analysis of the July MDC grid cells and regional series (Figs. 7, 8, 9) showed heterogeneous patterns of drought changes across Canada since the early twentieth century. Both eastern and western Canada have recorded significant decreases in July drought severity. On the other hand, northwestern Canada has recorded a positive shift in its mean July drought severity in the 1980s.

In general, the long-term changes in July MDC recorded in this study are coherent with the reported changes in fire activity in Canada. In the northwest, the frequency of large fire years and, consequently, annual area burned have been steadily increasing since 1970 (Kasischke and Turetsky 2006). Deeper seasonal thawing as a consequence of global warming is anticipated to further accelerate this trend. Conversely, the most important trend in July MDC was recorded in western Canada, with an estimated reduction of 0.37 unit yr−1. Synchronous to this trend is a decrease in the amount of area burned by forest fires in pine forests in British Columbia since about 1920, as reconstructed by Taylor et al. (2006). This later phenomenon has created ideal conditions for an unprecedented mountain pine beetle (Dendroctonus ponderosae Hopk.) outbreak in western Canada by changing forest composition and structure (age/size classes) to states more susceptible to the disturbance agent (Taylor et al. 2006). Taylor et al. (2006) attributed the area burned trend to the effectiveness of fire suppression, which has steadily increased with the greater availability of aircraft. However, based on our analyses of the regional July MDC record (Fig. 8a), we believe that it is reasonable to suggest that climate change may have played a leading role in regulating fire activity over the past century in this province (an opinion already expressed by Masters 1990). A correlation analysis between the regional July MDC and area burned in British Columbia (data from Stocks et al. 2003) has indeed indicated that both datasets shared significant common variance, with an R2 = 0.23 (P < 0.001). We were able to increase this amount of shared variance to R2 = 0.42 using the composite-plus-scale analysis (results not shown). We nonetheless feel the need to specify that our findings might not apply to locations where there is a distinctly thin or absent deep duff layer. Eastern Canada also recorded a decrease in fire activity, a trend that had already been attributed to the combined influence of climate change and fire suppression (Bergeron and Archambault 1993; Bergeron et al. 2004; Girardin et al. 2006b). The elevated July MDC for eastern Canada in 1900–09 and the 1920s (Fig. 8b) notably coincides with a period of important stand recruitment in the mixed wood boreal forest of southwestern Quebec: about 22% of the forested areas’ origin from these two decades (Bergeron et al. 2004; Girardin et al. 2006a).

7. Summary

We presented a methodological modification to the daily DC to allow its approximation using monthly data. It was shown that this new Monthly Drought Code has the ability to capture moisture trends in deep organic layers. The MDC is not intended to be used in operational situations where daily weather is available to fire managers; in this case, the daily version of the model would be far superior. We propose that this model be used when carrying out seasonal drought characterization analyses where daily data are not available. Such applications include paleoclimatic reconstruction of past droughts using tree-ring increment data (e.g., Stahle et al. 1985; Cook et al. 2004; Girardin et al. 2004a, 2006b; Li et al. 2007) or modeling of past and future changes in drought regimes using data obtained from global climate model simulations (Brewer et al. 2007).

According to our application of the MDC model to high-resolution Climate Research Unit (CRU) temperature and precipitation data, mean countrywide July MDC experienced strong (multi) decadal-scale variability. The 1920s through the early 1960s stand out as a persistent and exceptionally dry period in Canada, whereas the mid-1960s through the late 1980s were relatively humid. These changes were not uniformly distributed. When analyzed over the 1901–2002 period, drying was found to be statistically significant in northern Canada. Conversely, locations south of the Hudson Bay, in the eastern Maritimes, and in western Canada recorded a trend toward decreasing dryness (Fig. 7). When analyzed over the 1951–2002 period, trends could hardly be distinguished from the (multi) decadal variability.

Data quality at times when meteorological stations across Canada were rare (particularly north of 55°N prior to 1950) is certainly an issue in our analyses (McKenney et al. 2006; Trenberth et al. 2007). Furthermore, our analyses did not take into account overwintering effects of drought (i.e., where drought carries from year to year). For these two reasons, the century-long trend toward increasing dryness in northwestern Canada should be interpreted with caution. In contrast, for eastern Canada and British Columbia, meteorological stations are sufficiently replicated and precipitations are sufficient to saturate the DC layer in spring. The trend toward increasing summer moisture content in deep layers of the forest floor in these two regions is hence likely to be a robust feature. Trends over these two regions are also up to now well supported by reconstructions of wildfire activity that indicate a decrease in the extent of area burned throughout the twentieth century (Masters 1990; Bergeron and Archambault 1993; Bergeron et al. 2004, 2006; Taylor et al. 2006). Follow-up studies should look at how the MDC trend over northwestern Canada holds out when a drought-winter carryover model (Lawson and Dalrymple 1996) is implemented in the MDC formula using empirical constants specific to each location.

Acknowledgments

We thank Rémi St-Amant and Dominique Boucher for programming assistance. Thanks are extended to Isabelle Lamarre and three anonymous reviewers for their comments and for reviewing the manuscript. This work was financed by Canadian Forest Service funds.

REFERENCES

  • Abbott, K. N., , M. E. Alexander, , D. A. MacLean, , B. Leblon, , J. A. Beck, , and G. C. Staples, 2007: Predicting forest floor moisture for burned and unburned Pinus banksiana forests in the Canadian Northwest Territories. Int. J. Wildland Fire, 16 , 7180.

    • Search Google Scholar
    • Export Citation
  • Amiro, B. D., and Coauthors, 2001: Direct carbon emissions from Canadian forest fires, 1959 to 1999. Can. J. For. Res., 31 , 512525.

  • Amiro, B. D., , K. A. Logan, , B. M. Wotton, , M. D. Flannigan, , J. B. Todd, , B. J. Stocks, , and D. L. Martell, 2004: Fire weather index system components of large fires in the Canadian boreal forest. Int. J. Wildland Fire, 13 , 391400.

    • Search Google Scholar
    • Export Citation
  • Baskerville, G. L., 1971: Use of logarithmic regression in the estimation of plant biomass. Can. J. For. Res., 2 , 4953.

  • Bergeron, Y., , and S. Archambault, 1993: Decreasing frequency of forest fires in the southern boreal zone of Québec and its relation to global warming since the end of the ‘Little Ice Age’. Holocene, 3 , 255259.

    • Search Google Scholar
    • Export Citation
  • Bergeron, Y., , S. Gauthier, , M. Flannigan, , and V. Kafka, 2004: Fire regimes at the transition between mixed wood and coniferous boreal forest in northwestern Quebec. Ecology, 85 , 19161932.

    • Search Google Scholar
    • Export Citation
  • Bergeron, Y., and Coauthors, 2006: Past, current, and future fire frequencies in Quebec’s commercial forests: Implications for the cumulative effects of harvesting and fire on age-class structure and natural disturbance-based management. Can. J. For. Res., 36 , 27372744.

    • Search Google Scholar
    • Export Citation
  • Brewer, S., , S. Alleaume, , J. Guiot, , and A. Nicault, 2007: Historical droughts in Mediterranean regions during the last 500 years: A data/model approach. Climate Past, 3 , 355366.

    • Search Google Scholar
    • Export Citation
  • Briffa, K. R., , P. D. Jones, , and M. Hulme, 1994: Summer moisture variability across Europe, 1892–1991: An analysis based on the Palmer Drought Severity Index. Int. J. Climatol., 14 , 475506.

    • Search Google Scholar
    • Export Citation
  • Cook, E. R., , C. A. Woodhouse, , C. M. Eakin, , D. M. Meko, , and D. W. Stahle, 2004: Long-term aridity changes in the western United States. Science, 306 , 10151018.

    • Search Google Scholar
    • Export Citation
  • Coursolle, C., and Coauthors, 2006: Late-summer carbon fluxes from Canadian forests and peatlands along an east–west continental transect. Can. J. For. Res., 36 , 783800.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , K. E. Trenberth, , and T. Qian, 2004: A global dataset of Palmer Drought Severity Index for 1870–2002: Relationship with soil moisture and effects of surface warming. J. Hydrometeor., 5 , 11171130.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., , P. M. Bothwell, , D. H. Carlsson, , and K. A. Logan, 2003: Simulating the effects of future fire regimes on western Canadian boreal forests. J. Veg. Sci., 14 , 355364.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., , R. D. Field, , M. A. Brady, , O. Roswintiarti, , and M. Mohamad, 2007a: Development of the Indonesian and Malaysian fire danger rating systems. Mitigation Adapt. Strategies Global Change, 12 , 165180.

    • Search Google Scholar
    • Export Citation
  • de Groot, W. J., and Coauthors, 2007b: Estimating direct carbon emissions from Canadian wildland fires. Int. J. Wildland Fire, 16 , 593606.

    • Search Google Scholar
    • Export Citation
  • Fire Ecology Research Group, 2005: Experimental Climate Prediction Center - Fire Weather Index Forecast. Global Fire Monitoring Center, University of Freiburg, Freiburg, Germany. [Available online at http://www.fire.uni-freiburg.de/.].

    • Search Google Scholar
    • Export Citation
  • Flannigan, M. D., , K. A. Logan, , B. D. Amiro, , W. R. Skinner, , and B. J. Stocks, 2005: Future area burned in Canada. Climatic Change, 72 , 116.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , and M. Mudelsee, 2008: Past and future changes in Canadian boreal wildfire activity. Ecol. Appl., 18 , 391406.

  • Girardin, M. P., , J. Tardif, , M. D. Flannigan, , and Y. Bergeron, 2004a: Multicentury reconstruction of the Canadian Drought Code from eastern Canada and its relationship with paleoclimatic indices of atmospheric circulation. Climate Dyn., 23 , 99115.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , J. Tardif, , M. D. Flannigan, , B. M. Wotton, , and Y. Bergeron, 2004b: Trends and periodicities in the Canadian Drought Code and their relationships with atmospheric circulation for the southern Canadian boreal forest. Can. J. For. Res., 34 , 103119.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , Y. Bergeron, , J. C. Tardif, , S. Gauthier, , M. D. Flannigan, , and M. Mudelsee, 2006a: A 229-year dendroclimatic-inferred record of forest fire activity for the boreal shield of Canada. Int. J. Wildland Fire, 15 , 375388.

    • Search Google Scholar
    • Export Citation
  • Girardin, M. P., , J. C. Tardif, , M. D. Flannigan, , and Y. Bergeron, 2006b: Synoptic-scale atmospheric circulation and boreal Canada summer drought variability of the past three centuries. J. Climate, 19 , 19221947.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Ya, and Coauthors, 2007: Potential forest fire danger over Northern Eurasia: Changes during the 20th century. Global Planet. Change, 56 , 371386.

    • Search Google Scholar
    • Export Citation
  • Hogg, E. H., , J. P. Brandt, , and B. Kochtubajda, 2005: Factors affecting interannual variation in growth of western Canadian aspen forests during 1951-2000. Can. J. For. Res., 35 , 610622.

    • Search Google Scholar
    • Export Citation
  • Kasischke, E. S., , and M. R. Turetsky, 2006: Recent changes in the fire regime across the North American boreal region - Spatial and temporal patterns of burning across Canada and Alaska. Geophys. Res. Lett., 33 , L09703. doi:10.1029/2006GL025677.

    • Search Google Scholar
    • Export Citation
  • Keetch, J. J., , and G. M. Byram, 1968: A drought index for forest fire control. U.S. Dept. of Agriculture, Forest Service Research Paper SE-38, Southeastern Forest Experiment Station, Asheville, NC, 32 pp.

    • Search Google Scholar
    • Export Citation
  • Kurz, W. A., , G. Stinson, , G. J. Rampley, , C. C. Dymond, , and E. T. Neilson, 2008: Risk of natural disturbances makes future contribution of Canada’s forests to the global carbon cycle highly uncertain. Proc. Natl. Acad. Sci. USA, 105 , 15511555.

    • Search Google Scholar
    • Export Citation
  • Lawson, B. D., , and G. N. Dalrymple, 1996: Ground-truthing the Drought Code: Field verification of overwinter recharge of forest floor moisture. Canada-British Columbia Partnership Agreement on Forest Resources Development, FRDA Rep. 268, Pacific Forestry Centre and Research Branch, Victoria, BC, Canada, 21 pp.

    • Search Google Scholar
    • Export Citation
  • Lawson, B. D., , O. B. Armitage, , and W. D. Hoskins, 1996: Diurnal variation in the Fine Fuel Moisture Code: Tables and computer source code. Natural Resources Canada, Canadian Forest Service and British Columbia Ministry of Forests, FRDA Rep. 245, Pacific Forestry Centre and Research Branch, Victoria, BC, Canada, 20 pp.

    • Search Google Scholar
    • Export Citation
  • Lee, T. C. K., , F. W. Zwiers, , and M. Tsao, 2008: Evaluation of proxy-based millennial reconstruction methods. Climate Dyn., 31 , 263281.

    • Search Google Scholar
    • Export Citation
  • Li, J., , F. Chen, , E. R. Cook, , X. Gou, , and Y. Zhang, 2007: Drought reconstruction for north central China from tree rings: The value of the Palmer drought severity index. Int. J. Climatol., 27 , 903909.

    • Search Google Scholar
    • Export Citation
  • Mann, M. E., , S. Rutherford, , E. Wahl, , and C. Ammann, 2005: Testing the fidelity of methods used in proxy-based reconstructions of past climate. J. Climate, 18 , 40974107.

    • Search Google Scholar
    • Export Citation
  • Masters, A. M., 1990: Changes in forest fire frequency in Kootenay National Park, Canadian Rockies. Can. J. Bot., 68 , 17631767.

  • McAlpine, R. S., 1990: Seasonal trends in the Drought Code component of the Canadian Forest Fire Weather Index System. Forestry Canada, Petawawa National Forestry Institute, Chalk River, ON, Canada, Information Rep. PI-X-97 E/F, 10 pp.

    • Search Google Scholar
    • Export Citation
  • McKenney, D. W., , J. H. Pedlar, , P. Papadopol, , and M. F. Hutchinson, 2006: The development of 1901–2000 historical monthly climate models for Canada and the United States. Agric. For. Meteor., 138 , 6981.

    • Search Google Scholar
    • Export Citation
  • Mitchell, T. D., , and P. D. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25 , 693712.

    • Search Google Scholar
    • Export Citation
  • Mudelsee, M., 2000: Ramp function regression: A tool for quantifying climate transitions. Comput. Geosci., 26 , 293307.

  • Mudelsee, M., , and M. Alkio, 2007: Quantifying effects in two-sample environmental experiments using bootstrap confidence intervals. Environ. Model. Software, 22 , 8496.

    • Search Google Scholar
    • Export Citation
  • Muraro, S. J., , and B. D. Lawson, 1970: Prediction of duff moisture distribution for prescribed burning. Canada Department of Fisheries and Forestry, Canadian Forestry Service, Information Rep. BC-X-46, Pacific Forest Research Centre, Victoria, BC, Canada, 13 pp.

    • Search Google Scholar
    • Export Citation
  • Otway, S. G., , E. W. Bork, , K. R. Anderson, , and M. E. Alexander, 2007: Predicting sustained smouldering combustion in trembling aspen duff in Elk Island National Park, Canada. Int. J. Wildland Fire, 16 , 690701.

    • Search Google Scholar
    • Export Citation
  • Palmer, W. C., 1965: Meteorological drought. U.S. Department of Commerce, Weather Bureau Research Paper 45, 58 pp.

  • Régnière, J., 1996: A generalized approach to landscape-wide seasonal forecasting with temperature-driven simulation models. Environ. Entomol., 25 , 869881.

    • Search Google Scholar
    • Export Citation
  • Régnière, J., , and P. Bolstad, 1994: Statistical simulation of daily air temperature patterns in eastern North America to forecast seasonal events in insect pest management. Environ. Entomol., 23 , 13681380.

    • Search Google Scholar
    • Export Citation
  • Rodionov, S. N., 2006: Use of prewhitening in climate regime shift detection. Geophys. Res. Lett., 33 , L12707. doi:10.1029/2006GL025904.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., , and W. Skinner, 2004: Summer drought patterns in Canada and the relationship to global sea surface temperatures. J. Climate, 17 , 28662880.

    • Search Google Scholar
    • Export Citation
  • Stahle, D. W., , M. K. Cleaveland, , and J. G. Hehr, 1985: A 450-year drought reconstruction for Arkansas, United States. Nature, 316 , 530532.

    • Search Google Scholar
    • Export Citation
  • Stocks, B. J., and Coauthors, 2003: Large forest fires in Canada, 1959–1997. J. Geophys. Res., 108 , D18149. doi:10.1029/2001JD000484.

  • Taylor, S. W., , A. L. Carroll, , R. I. Alfaro, , and L. Safranyik, 2006: Forest, climate and mountain pine beetle outbreak dynamics in western Canada. The Mountain Pine Beetle: A Synthesis of Biology, Management, and Impacts in Lodgepole Pine, L. Safranyik and W. R. Wilson, Eds., Natural Resources Canada, 67–94.

    • Search Google Scholar
    • Export Citation
  • Thornthwaite, C. W., , and J. R. Mather, 1955: The water balance. Publ. Climatol., 8 , 186.

  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

    • Search Google Scholar
    • Export Citation
  • Turner, J. A., 1972: The drought code component of the Canadian Forest Fire Behaviour System. Environment Canada, Canadian Forest Service Publication 1316, Ottawa, ON, Canada, 14 pp.

    • Search Google Scholar
    • Export Citation
  • van der Schrier, G., , K. R. Briffa, , P. D. Jones, , and T. J. Osborn, 2006: Summer moisture variability across Europe. J. Climate, 19 , 28182834.

    • Search Google Scholar
    • Export Citation
  • Van Wagner, C. E., 1974: Effect of duff weight on drying rate. Bi-Mon. Res. Notes Can. For. Serv., 30 , 1112.

  • Van Wagner, C. E., 1985: Drought, timelag and fire danger rating. Proc. Eighth Conf. on Fire and Forest Meteorology, Publication 85-04, Detroit, MI, Society of American Foresters, 132–149.

    • Search Google Scholar
    • Export Citation
  • Van Wagner, C. E., 1987: Development and structure of the Canadian Forest Fire Weather Index System. Forestry Tech. Rep. 35, Canadian Forest Service, Ottawa, ON, Canada, 37 pp.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., , and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Weber, M. G., , and M. D. Flannigan, 1997: Canadian boreal forest ecosystem structure and function in a changing climate: Impact on fire regimes. Environ. Rev., 5 , 145166.

    • Search Google Scholar
    • Export Citation
  • Wilmore, B., 2001: Duff moisture dynamics in black spruce–feather moss stands and their relation to the Canadian Forest Fire Danger Rating System. M.S. thesis, University of Alaska, Fairbanks, 84 pp.

  • Xiao, J., , and Q. Zhuang, 2007: Drought effects on large fire activity in Canadian and Alaskan forests. Environ. Res. Lett., 2 , 044003. doi:10.1088/1748-9326/2/4/044003.

    • Search Google Scholar
    • Export Citation
  • Zar, J. H., 1999: Biostatistical Analysis. 4th ed. Prentice Hall, 929 pp.

Fig. 1.
Fig. 1.

MDC [Eq. (11)] vs monthly means of the daily DC mean [Eq. (3)] for four locations across Canada over the 1900–2006 period. Months under analysis are May–October (n = 642 data points). A least squares linear regression (gray line) is shown along with the model R2. Refer to Table 1 for information on these locations.

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 2.
Fig. 2.

Spearman rank correlation coefficients squared (R2) computed between MDC and monthly means of the daily DC over 1900–2006 for each calendar month [May (M)–October (O)]. Refer to Table 1 for information on these locations.

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 3.
Fig. 3.

Climatological averages of MDC and of monthly means of the daily DC, per month, for the 1900–2006 period. Months under analysis are May (M)–October (O).

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 4.
Fig. 4.

Mean July MDC over Canada, 1901–2002 period. Scale (unitless) ranges from moist (blue, MDC < 139) to dry (yellow, MDC > 245).

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 5.
Fig. 5.

(a) Mean Canadian July MDC. The values are a spatial average of all grid cells, weighted with the cosine of the lat. The solid line shows a 10-yr low-pass filter (order 4); the dashed line shows long-term changes in mean as detected using the sequential algorithm method {Rodionov 2006; 30-yr window with a red noise [i.e., AR(1)] parameter set to 0.07}. The three driest and wettest years on record are 1961, 1955, 1958, and 1957, 1978, 1993, respectively (omitting the portion earlier than 1920). (b) Std dev.

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 6.
Fig. 6.

Percentage of area experiencing moderate (MDC > 120) to extreme (MDC > 280) July droughts in Canada. The solid line shows a 10-yr low-pass filter (order 4). The dashed line shows long-term changes in mean as detected using the sequential algorithm method [Rodionov 2006; 30-yr window with AR(1) parameters from top to bottom set to 0.24, 0.15, 0.00, and 0.00].

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 7.
Fig. 7.

Linear trends of July MDC for (a) 1901–2002 and (b) 1951–2002 as detected using the Spearman rank correlation coefficient (for skewness bias). Correlation scale ranges from decreasing drought severity (blue) to increasing dryness (yellow). Trends significant at the 5% level are indicated by black dots.

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 8.
Fig. 8.

Regional July MDC records. Coordinates refer to upper-left and lower-right corners of the gridcell domains. The solid line shows a 10-yr low-pass filter (order 4). The dashed line shows long-term changes in mean as detected using the sequential algorithm method [Rodionov 2006; 30-yr window with AR(1) parameters set to (a) 0.00, (b) 0.13, and (c) 0.00]. The dotted line shows linear trends of MDC for 1901–2002 (P value is indicated; all three regression models passed the Durbin–Watson, normality, and constant variance tests): the linear trends are, respectively, (a) −0.37, (b) +0.20, and (c) −0.26 MDC (units yr−1).

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 9.
Fig. 9.

(a) Correlation pattern between the total annual area burned in Canada by large forest fires (size > 200 ha) and July MDC (1959–99). Correlations significant at the 5% level are indicated by black dots. (b) Total annual area burned in Canada predicted from July MDC using the composite-plus-scale analysis, plotted against the observations.

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Fig. 10.
Fig. 10.

July MDC records for (a) Fort McMurray and (b) Thunder Bay (Table 1) before and after applying an overwintering adjustment using program SimMDC [Eq. (4)]. Overwintering empirical constants a and b were subjectively set to 0.75 (area subject to bare ground in winter) and 0.50 (well-drained soils), respectively (see Lawson and Dalrymple 1996). The Spearman R2 between MDC records over the 1901–2002 period is shown. Linear trends of MDC are also shown (P < 0.01 for the Thunder Bay location).

Citation: Journal of Applied Meteorology and Climatology 48, 3; 10.1175/2008JAMC1996.1

Table 1.

Study locations. Mean of seasonal averages of daily maximum temperatures and mean of seasonal precipitation totals are also shown (period: 1900–2006).

Table 1.
Table 2.

Average and variance of monthly means of daily DC and of MDC, 1900–2006 period. The SW statistics and P values refer to the Shapiro–Wilk normality test; the fit with the normal curve is declared poor (test “failed”) if P value < 0.050.

Table 2.
Table 3.

Summary statistics of the least squares linear fit between the total annual area burned in Canada by large forest fires (size > 200 ha; logarithmic transformed) and annual values of a spatial average of all July MDC grid cells.

Table 3.
Save