Aridity on the Canadian Plains

Sauchyn, David; Barrow, Elaine; Hopkinson, Ron; Leavitt, Peter

doi:https://doi.org/10.7202/009109ar

Introduction

Transient climate change simulations from the Canadian Centre for Climate Modelling and Analysis suggest that the largest CO₂-induced rise in mean surface temperature in southern Canada will occur in the Interior Plains (Boer et al., 2000). Most general circulation models (GCM) also forecast an increase in winter precipitation for this region, coupled with decreased net soil moisture and water resources in summer, and more frequent extreme departures from mean conditions, including severe drought (Hengeveld, 2000). Unfortunately, long-term forecasting of precipitation is problematic for dry environments (Hostetler and Giorgi, 1993; Herrington et al., 1997), where precipitation is variable and occasionally extreme (Hunter et al., 2002). Subhumid ecosystems and soil landscapes are sensitive to fluctuations in the surface and soil water balances (Lemmen and Vance, 1999). Sustained periods of low precipitation and soil moisture lower the resistance to disturbance such that future climatic variability may exceed thresholds for landscape degradation. The recovery of natural systems to a previous or new equilibrium can take decades or centuries (Wolfe et al., 2001).

Both aridity and drought refer to a negative soil and surface water balance. Aridity is a long-term condition and thus is characteristic of a regional climate, while drought is a short-term (seasons to years) departure from mean conditions and thus represents climatic variability. GCM forecasts of increased aridity and more frequent drought have major implications for rates of erosion and sediment yield in the southern Canadian plains. In general, biophysical systems react to short-term climate variability and to extreme events before they respond to gradual changes in mean conditions (Knox, 1984; Hulme et al., 1999). Extreme and anomalous climate can exceed natural and engineering thresholds beyond which the impacts of climate are more severe. The link between aridity and erosion is well established from paleoenvironmental records (Wolfe et al., 2001; Last and Vance, 2002) and from the monitoring of geomorphic processes and regional sediment yields (Knox, 1984). Less protection of the soil surface from wind and rain splash is generally given or implied as the cause of higher rates of erosion in semiarid landscapes, however, plants also reduce runoff erosion through the transpiration of soil water and the positive influence of stems, roots and organic matter on the infiltration or rain and snowmelt water (Thornes, 1985).

Thus aridity lowers resistance to wind and water erosion (Wheaton, 1984; Wolfe and Nickling, 1997; Campbell, 1998). Wheaton (1990) established an exact rank correlation between the number of days with dust and the Palmer Drought Severity Index for the Prairie Province during 1977-88. Wolfe et al. (1995) described the control of soil moisture on the stability of dunes, with the onset of dune reactivation closely corresponding to the incidence of drought during the past 1000 years. Optical dates defined an interval of minimal dune activity between 560 and 220 BP, implying that drought severity did not exceed a threshold for demise of vegetation cover. In contrast, sand dunes were active between about 1000 and 900 BP and since about AD 1800, following droughts of the 1790s known from tree-ring records (Sauchyn and Skinner, 2001; Wolfe et al., 2001). Current sand dune activity in the driest area of the southern Canadian Prairies may serve as an analogue of the potential response of other eolian landscapes in the region to future global warming (Thorpe et al., 2001).

Long-term changes in net precipitation characteristics may also impact fluvial geomorphic systems in the Canadian Prairies (Campbell, 1998; Ashmore and Church, 2001). The climate sensitivity of prairie fluvial geomorphic systems is recorded in late Holocene sedimentary and morphologic records (O’Hara and Campbell, 1993; Rains et al., 1994) with episodes of channel aggradation and incision roughly coinciding with climate fluctuations recorded in lacustrine records (Lemmen and Vance, 1999). Streams that are fed wholly from sources within the plains may be expected to be more responsive to changes in precipitation than the rivers that are derived from glaciers and snowmelt in the Rocky Mountains (Campbell, 1998; Ashmore and Church, 2001). Because the stream channels of the Interior Plains are incised into Quaternary sediments and poorly consolidated sedimentary strata, slope and bank failure have been integral processes in the Holocene history of glacial meltwater channels and tributary valleys (Sauchyn, 1998). Although drier climate and lesser porewater favour slope stability, slope failure during extreme events may occur with greater frequency.

Because aridity is closely linked to soil erosion and degradation, future management of prairie ecosystems and soil landscapes will require an improved understanding of past and future trends and variability in regional aridity. Land use and management practices have more immediate impacts on rates of surface processes than climate change (Jones, 1993), and thus they have the potential to significantly mitigate the impacts of climate; forest and farm management practices enable adaptation to climate change. An expansion of agriculture into the currently forested margins of the Prairie Ecozone will require assessment of the sensitivity of these soil landscapes to both climate change and an altered vegetation cover. To address these issues, we use both instrumental and paleoclimatic records to define regional baseline aridity for the Prairies, and to evaluate the utility of modern climate normals (i.e. 1961-1990) as a metric of environmental change. The proxy and instrumental climate data are compared to a range of GCM forecasts spanning the next 80 years to evaluate the usefulness of historical climate reconstructions as analogs for future climate change.

Data and Methods

Aridity is simply a lack of moisture expressed in terms of mean annual precipitation and potential evapotranspiration as a ratio (P/PET) or deficit (P-PET). P/PET, the Aridity Index (AI), is the basis for classifying drylands as hyperarid (< 0.05), arid (0.05 ≤ AI < 0.2), semiarid (0.2 ≤ AI < 0.5) and dry subhumid (0.5 ≤ AI < 0.65) and for identifying land at risk of desertification, “land degradation in arid, semiarid and dry subhumid areas resulting mainly from adverse human impact” (Middleton and Thomas, 1992). Wolfe (1997) demonstrated that the distribution of sand dunes on the Canadian plains could be modeled with P/PET. Hogg (1994, 1997) was able to map the distribution of vegetation zones in the Canadian interior using the Climate Moisture Index, P-PET. Because our objective was to describe and compare past and future changes in aridity, rather than evaluate different methods of AI calculation, all estimates of PET were based on the widely used Thornthwaite formula (Dunne and Leopold, 1998):

where PET = potential evapotranspiration (cm month^-1), T = mean monthly air temperature (°C) and

Annual heat index, I, was estimated as

With Thornthwaite PET, temperature is the sole measure of the energy available for evapotranspiration. Monthly PET calculated using equation (1) must be adjusted using a correction factor for daylength according to latitude and month. Net radiation and wind data are available for few stations in the Prairie Provinces and thus an energy balance approach to calculating PET would not enable the mapping of aridity. We use Thorntwaite PET for this reason and because it works best in mid-latitude continental climates where air temperature if strongly correlated with net radiation.

To map recent and future aridity we first loaded a gridded database of historical climate observations into a GIS. This database of mean monthly temperature and total precipitation was derived from all of the available monthly climate data in the Canadian Climate Archive for the Prairie Provinces and the surrounding climate regions. The FORTRAN program GRANAL2 used an inverse square distance weighting procedure to interpolate from the station locations to a 50 km grid on a polar stereographic secant projection true at 60^oN. The values at the grid points should be viewed as representative of the region within 25 km of the grid point, although the gridded analysis cannot reproduce the high spatial variability associated with precipitation events. Further information on the database and interpolation procedure is available at http://www.cics.uvic.ca/climate/data.htm. In the GIS, a more dense 5 km grid was constructed to smooth the raster boundaries between map units created by classifying the climate data. The aridity index was computed for each original grid point and interpolated to the 5 km grid by Kriging.

Scenarios of future climate change were applied to the gridded temperature and precipitation data for 1961-1990 to produce maps of future aridity centered on the 2020s (2010-2039), 2050s (2040-2069) and 2080s (2070-2099). The IPCC Task Group on Scenarios for Climate Impact Assessment (IPCC-TGCIA, 1999) strongly recommended the use of more than one GCM for any assessment of climate impacts and that the selected GCMs show a range of changes in the key climate variables for the study region. In a recent study of climate change impacts on the island forests of the northern Great Plains, Henderson et al. (2002) thoroughly analyzed the available GCM scenarios and chose three that captured the range of annual temperature and precipitation changes projected for the region. Given the very similar scenario requirements of this very recent study and the forecasting of future aridity attempted here, we accepted the conclusions reached by Henderson et al. (2002) and used the same climate change experiments listed in Table I. HadCM3 B21 forecasts the least increase in temperature and greatest increase in precipitation. CGCM2 A21 forecasts the smallest change in precipitation and the largest increase in temperature. CSIROMk2b B11 is a mid-range scenario. The names of the models refer to the national modeling centres (Table I), the version of the coupled atmosphere-ocean model, and the emission scenario (the last three digits). The greenhouse gas emission scenarios are those most recently recommended by the IPCC from the Special Report on Emission Scenarios (Nakicenovic and Swart, 2000). The scenario data were downloaded from the web site of the Canadian Climate Impacts and Scenarios project (http://www.cics.uvic.ca/scenarios/).

Table I

**The climate change experiments (from Henderson et al., 2002) used for the forecasting of future aridity**

High-resolution records of paleo-aridity were derived from analyses of lake sediments and tree rings. The tree ring chronology is from Pinus contorta (lodgepole pine) in the Cypress Hills of southwestern Saskatchewan and southeastern Alberta (Fig. 1). Standard procedures were used to process the tree cores and disks and measure ring width to within 0.001 mm (Cook and Kairiukstis, 1990). Cross dating of all tree-ring series enabled the detection of anomalous series and missing or false rings and ensured that the correct calendar year was assigned to each tree ring. There are 53 ring-width series with an inter-correlation of 0.481 and a mean sensitivity of 0.231. The raw ring-width data were converted to index chronologies (averaged standardized ring widths) using the program ARSTAN (Cook and Holmes, 1999). The raw data were detrended with a best-fit negative exponential curve or linear regression and filtered to remove first-order autocorrelation. A biweight robust estimate of the mean ring-width index for each year comprised the tree-ring index chronology. The reconstruction of June-July P/PET from standardized residual ring width was validated using a split sample approach; two regression models were calibrated using different halves of the instrumental record and then cross-validated using the remainder of the record.

Figure 2

**Aridity Index (P/PET), 1961-90**

Chauvin Lake (Fig. 1) is ideally suited for climatic research because it is located within a very small effective drainage. Analysis of aerial photographs between 1940 and the present demonstrate that lake levels have varied in accordance with historical changes in precipitation. Lake sediments cores were collected using standard piston-coring procedures and sediments were cut into 2.5 mm intervals equivalent to an average of 1-3 years of lake history. Fossil diatoms were isolated, identified to species and quantified for alternate sediment levels, and changes in species composition were used in published inference models (Wilson et al., 1996) to reconstruct past changes in lake water salinity at intervals of 3-5 years. Analysis of past changes in chemistry show that lake water salinity accurately records the occurrence of major drought events (Laird et al., 1996), with greatly elevated salinity representing the most arid conditions. Sediment age was determined by analysis of naturally-occurring radioisotopes in recent bulk sediments (²¹⁰Pb) and individual plant macrofossils (AMS ¹⁴C).

Diatom-inferred changes in lake salinity were compared to instrumental climate records from 1910 to 1990 to evaluate the ability of fossil reconstructions to infer changes in aridity. PET data were filtered using a three-year running mean to match the temporal resolution of the fossil time series. All time series were detrended by linear regression, and residual curves compared using correlation analysis. Lags of 1-5 years were applied to the fossil data to evaluate the possibility of delayed hydrologic response to climatic forcing functions.

Results and Discussion

A map of “normal” (i.e. 1961-1990) aridity for the Prairie Provinces (Fig. 2) illustrates that, at 50 km resolution, the “normal” climate of the Prairie Provinces cannot be described as semiarid anywhere (P/PET < 0.5), even though this terminology is often used to describe the southern Prairies. The driest region, the dry subhumid (0.5 < P/PET < 0.65) landscape of southwestern Saskatchewan and southeastern Alberta, corresponds to the mixed grass prairie ecoregion. The moist subhumid Cypress Hills are near the geographic center of this driest region. Analysis of the longest instrumental record (Medicine Hat, Alberta) showed that annual P/PET varied from 0.34 (1943) to 1.24 (1927), with 26 years of semiarid conditions (Fig. 3). On average, these dry events occurred approximately every third year, although only eight semiarid years were seen after 1950.

This large inter-annual variation in the aridity index translates into varying geographic extent of annual drought. Each documented drought had a unique geographic extent and intensity, although semiarid conditions were consistently recorded in southeastern Alberta and southwestern Saskatchewan (Fig. 4). The “normal” dry subhumid climate of this region reflects the frequency of low annual precipitation and P/PET. The most extensive drought was in 1961.

Figure 3

**Annual P/PET, Medecine Hat Alberta, 1900-97**

Scenarios of future aridity centered on the 2020s, 2050s and 2080s are mapped for the Prairie Provinces in Figures 5-7 for the CGCM2 A21 (warm-dry), CSIROMk2b B11 (mid-range), HadCM3 B21 (cool-wet) climate change scenarios. Under the CGCM2 A21 scenario, a small area of continuously semiarid climate (P/PET < 0.5) appears in western Saskatchewan by the 2020s and expands significantly by the 2050s. In contrast, forecasts using the CSIROMk2b B11 mid-range scenario (Fig. 6) suggest that perpetually semiarid conditions will be evident only by 2080, but that a dry subhumid climate will extend to southwestern Manitoba. The HadCM3 B21 scenario indicates the least climate change, with a marginal increase in the area of subhumid climate and maximum aridity in mid century. Taken together, these forecasts suggest a general increase in the geographic extent of dry conditions, with possible risks of desertification in the more extreme scenarios (CGCM2 A21).

Analysis of tree-ring indices of past climate suggest that very dry conditions have been common on the Prairies during the past 300 years (Fig. 8). The residual index tree-ring widths for Pinus contorta from the Cypress Hills are significantly correlated (p < 0.01) with June, July, annual and plant year (August to July) precipitation for Medicine Hat, Alberta for the period 1900-1999 (Table II). The correlation of ring width with potential evapotranspiration (PET) is weak (p > 0.05) because Thornthwaite PET is a function of temperature and heat has both positive and negative effects on tree growth depending on the ambient air temperature and antecedent moisture conditions. The negative influence would be confined to warm dry days. Thus the aridity (P/PET) signal in the tree rings mostly reflects large interannual variability in annual, and especially summer, precipitation. Not only is the response to precipitation direct and unambiguous, but amounts of rain and snow are much more variable from year to year than mean temperature and in turn evapotranspiration.

The reconstruction of summer (June-August) aridity for 1723-2000 is plotted in Figure 8. The similar standard errors of estimation (SE_e) and validation (RMSE_v) in Table III suggest an acceptable regression model. Events of high aridity occurred regularly during the past three centuries, with major droughts indicated for 1768-1782, 1786-1796 and 1851-1864. In addition, analysis of the time series suggested that mean interannual variance (as standard deviation) was greater during the instrumental period (post-1884; SD = 0.080) than in the preceding period (1723-1883; SD = 0.063). This prominent increase in inter-annual variability may be related to a mid-19^th century shift in the Pacific Decadal Oscillation (D’Arrigo et al., 2001). The consecutive years of summer aridity highlighted in Figure 8 are further evidence of prolonged aridity prior to Euro-Canadian settlement of the western Canadian plains and the instrumental observation of climate. The impacts of these droughts are described in Wolfe et al. (2001) and Sauchyn et al. (2003). Their relevance here is as possible analogues of anomalous precipitation expected with global warming in contrast to the climate of the 20^th century, and particularly 1961-1990, the basis for our definition of the normal climate.

Paleoclimatic reconstructions based on the 2000-year salinity record from Chauvin Lake, Alberta, also suggest that past drought severity was greater than that recorded during the instrumental period. Validation using modern climatic records showed that detrended, diatom-inferred salinity was significantly and positively correlated (p < 0.05) with PET during 1910-1990 (Fig. 9). Correlations were also positive and significant at lags of one or two years. Using this observation, and the drought of 1988-1989 as a benchmark, analysis of past salinities revealed that climatic conditions were continuously arid during the 4^thto 8^th centuries and regularly dry during CE 1300-1750 (Fig. 10). Severe events were also recorded during 1860-1890, an interval of broadscale droughts in central North America. Salinities declined substantially during the 20^th century, reaching values during the 1960s and 1970s that were the lowest of the past 750 years.

Other recent studies of the past hydroclimate of the northern Great Plains (Woodhouse and Overpeck, 1998; St. George and Nielsen, 2002) have similarly concluded that the droughts of the 20^th century were of comparatively short duration. Much of the previous paleo drought research has focused on the United States (Cook et al., 1999). Our results fit with the timing of droughts (e.g., 1790s, 1860s, 1880s) in eastern Montana, immediately south of our study area, but generally do not correlate with tree-ring data from the eastern plains (St. George and Nielsen, 2002). The lower frequency variation in hydroclimate (i.e., aridity), on the other hand, tends to be consistent over a larger area. Our tree-ring and diatom data share decadal to century scale variability with dendroclimatic (Case and MacDonald, 1995; St. George and Nielsen, 2002) and lake sediment (Laird et al., 2003) records from across the northern Great Plains. Common periods of sustained aridity encompass much of the 16^th and 18^th centuries.

Figure 8

**Summer (JJA) P/PET Medecine Hat, 1723-2000**

Table II

Correlations between the residual index chronology for Pinus contorta (PICO) from the Cypress Hills (CH) and precipitation (P), potential evapotranspiration (PET) and P/PET for Medicine Hat, Alberta for the period 1900-1999. The bold coefficients are significant at p <0.01

Table III

Tree-ring regression model and validation statistics. The predictand is June – July P/PET at Medicine Hat and the predictor is standardized residual ring widths from Pinus contorta from the Cypress Hills

Figure 10

**Paleosalinity (g/l), Chauvin Lake, Alberta**

Conclusions and Implications

Analysis of climate normals (1961-1990) revealed that a large part of the southern Canadian Plains usually has dry subhumid conditions, and that P/PET was rarely < 0.5 (semiarid) when analyzed on a 50 x 50 km grid. In contrast, semiarid conditions were common, on average, one of every three years during the instrumental record, although the high degree of spatial variability suggested that any given site could experience a wide range of climate, from semiarid (P/PET < 0.5) to humid conditions (> 1.0). High temporal variability was evident in the tree-ring (Fig. 8) and lake sediment (Fig. 10) reconstructions, however, these records also demonstrated that more prolonged, intensely arid events were common in the past, and showed further that modern climate normals are not representative of the full range of potential climatic conditions, even in the absence of global warming.

The increasing aridity projected by the GCM scenarios represents a higher frequency of dry years over a larger area. As a result of an increase in mean annual temperature and, in turn, potential evapotranspiration, with even the coolest and wettest climate change scenario, there is a marginal increase in the area of subhumid climate in the Prairie Provinces by the 2080s. The Canadian GCM forecasts the least increase in precipitation and the largest increase in temperature and therefore an approximately 50 % increase in the area of subhumid climate and a significant area of semiarid climate by the 2050s. Past droughts may provide important benchmarks to evaluate potential impacts of increasing aridity forecast by GCMs. For example, comparison of climatic conditions observed in the pre-drought year 1987 shows a marked similarity to conditions forecast by the CGCM2 A21 scenario for 2050, in terms of the geographic extent of semiarid and dry subhumid climate (Fig. 11a). Dry conditions during 1987 are viewed as partly responsible for establishing more extensive drought during 1988-1989 (Fig. 4) and suggest that future increases in aridity may generate conditions that favour the regular development of large-scale droughts.

The implication of more frequent and widespread droughts is that they place a larger area at risk of desertification more often unless counteracting management strategies are enacted. The critical variable will be the duration of the semiarid conditions in any given area. The scenario of increased climate extremes includes unusually wet years during which social and biophysical systems may recover from drought. If aridity is prolonged, however, the resistance of systems could drop below critical thresholds, such that they will not recover during the next wet year. Based on evidence derived from climatic proxies, our analysis suggests that the climate of the 20^th century was anomalous in terms of the absence of sustained drought. Further, comparison of instrumental records and climatic reconstructions suggests that the climate normal period of 1961-1990 may have been among the most benign of the past 750 years. Possibly, the climate of 1921-1950 (Fig. 11b) is more characteristic of the past and future climate of the Prairie Provinces than the “normal” climate of 1961-1990. Because both lake and tree-ring analyses recorded an abrupt amelioration of climatic conditions at near the start of the instrumental record, we suggest that the immediate impacts of future global warming may be to return the prairie environment to past conditions in which persistent aridity was recorded for intervals of decades or longer.

Figure 11

-> Voir la liste des figures

Aridity on the Canadian PlainsL’aridité dans les PrairiesAridez de la praderas canadienses

Abstract

Résumé

Resumen

Introduction

Data and Methods

Results and Discussion

Conclusions and Implications

Acknowledgements

References

Liste des figures

Liste des tableaux

Résumés

Abstract

Résumé

Resumen

Corps de l’article

Introduction

Data and Methods

Results and Discussion

Conclusions and Implications

Parties annexes

Acknowledgements

References

Liste des figures

Liste des tableaux

Outils de citation

Citer cet article

Exporter la notice de cet article