Successes of soil conservation in the Canadian Prairies highlighted by a historical decline in blowing dust

Blowing dust from agricultural fields has serious health and economic effects, which can be mitigated by soil conservation techniques. However, it is difficult to isolate improved land management in downstream records of airborne dust. In this letter we present multi-decadal (1961–2006) records of airborne dust frequency from seven weather stations across the Canadian Prairies. We related temporal changes in dust frequency to the climatic wind erosion potential and agricultural census data. We identified a statistically significant regime shift in the region-wide dust time series at 1990, with a substantial reduction in dust frequency thereafter. The correspondence between dust frequency and the climatic wind erosion potential improved from 1961–90 (r2 = 0.154, p < 0.001) to 1991–2006 (r2 = 0.429, p < 0.001). We interpret this as indicating that the climate signal was obscured by poor soil conservation practices in 1961–90, leading to dustier conditions. Post 1990, improved land management reduced the impact of land-use practices; only the most severe climate forcings resulted in detectable dust. The dramatic reduction of dust from 1990 onward appears to represent a region-wide threshold crossing, where the effects of soil conservation efforts began to materialize. Overall, the results suggest that soil conservation initiatives have had an impact in reducing airborne dust on the Canadian Prairies.


Introduction
Wind erosion of agricultural soil has numerous negative economic and health effects. Locally, erosion reduces yields by decreasing soil water holding potential (Colacicco et al 1989), soil nutrient content (Larney et al 1998), and often requires increased application of herbicides and pesticides (Wheaton 1992). Adjacent to farms, erosion reduces air quality (Hagen and Woodruff 1973) and visibility (Hagen and Skidmore 1977). Furthermore, airborne dust can act as an allergen (Kellogg and Griffin 2006), increase skin and eye irritations, and augment risk of various cancers (Norton andGunter 1999, Nordstrom andHotta 2004).
Wind erosion and blowing dust on the northern Great Plains and Canadian Prairies (figure 1) is largely an anthropogenic land-use problem. The climate and land cover are generally not conducive, at present, to recurrent, large-scale dust outbreaks as seen in drier settings (e.g., Bodélé depression, Washington et al 2006). The dust footprint caused by anthropogenic land-use activities is evident in many proxy records and direct observations. For instance, lake cores from Colorado indicate a 500% increase in dust deposition coincident with settlement in the late 1800s (Neff et al   [1933][1934][1935][1936][1937][1938], where a severe drought and poor land management resulted in significant soil losses and economic hardship across much of the North American Great Plains (Schubert et al 2004, Marchildon et al 2008. Over the long term, estimates of income foregone due to soil degradation in Canadian Prairie Provinces are as high as $700 million yr −1 (USD), with wind erosion as the costliest component (Dregne 2002). As a result, many initiatives have been instigated to educate farmers on effective wind erosion control techniques (table 1; also see Marchildon et al 2008).
Despite extensive funding to develop and promote wind erosion prevention techniques, especially in the late 1980s, there has been minimal large-scale monitoring of the results or outcomes. Individual farmers may have identified increased yields associated with preventing wind erosion, but improvements in the downstream quantity of airborne dust remain poorly assessed. Both Canada and the United States of America have extremely limited capacity for monitoring airborne dust and assessing changes over long baselines (Wheaton et al 2008, Trimble andCrosson 2000), and as such, researchers are relegated to using meteorological observations not specifically designed for the task. Despite this, several studies have been performed across North America indicating that a reduction of dust events could be related to improvements in field-scale farm management. In the Southern High Plains of Texas a gradual decline in the frequency of historical dust events was identified (Lee et al 1993, Stout andLee 2003). Similarly, in the Red River Valley of North Dakota, Todhunter and Cihacek (1999) found declines in historical dust events. Wheaton and Chakravarti (1990) conducted a short study on the Canadian Prairies  and found no trends. This has left a gap in the research, where the decadal-scale patterns of wind erosion and dust frequency on the Canadian Prairies remain un-assessed.
To this end, we investigated multi-decadal  changes in dust event frequency from seven sites across the southern Canadian Prairies and explored relations with suspected explanatory variables. The specific motivation for this research was to determine whether soil conservation practices and initiatives (see table 1) have been successful in reducing blowing dust. First, we outline recent historical changes in dust frequency, climatic wind erosion potential (CWEP) and agricultural practices. Second, we identify the timing of a regime shift in the annual dust frequency time series. This single changepoint establishes a marker for the onset of a step change in blowing dust frequency. We then relate the time series of dust frequency to the time series of CWEP and agricultural census data. Overall, we find that a reduction of dust event frequency has occurred, coincident with an improvement in agricultural practices. This suggests that soil conservation initiatives have impacted airborne dust.

Site and methods
The Canadian Prairies consist of the southern portions of the provinces of Alberta, Saskatchewan and Manitoba (figure 1). The region represents the northernmost tip of the North American Great Plains and has been under extensive agricultural land use for over 100 years. Four data sets were compiled and analyzed: (i) records of observed dust from prairie weather stations, (ii) homogenized wind data, (iii) homogenized precipitation data, and (iv) land management data. These data were used to produce time series of: (i) observed dust frequency (hours per year and per month), (ii) CWEP, and (iii) a qualitative measure of anthropogenic land-use forcing, in the form of land management data from Canadian Agricultural Census.
Airborne dust records were acquired from the online Environment Canada Historical Weather database for the cities of Edmonton, Red Deer, Calgary, Saskatoon, Regina, Brandon and Winnipeg (figure 1). These sites were selected for their even spatial distribution across the Prairies; each also had a minimum of 45 yr of hourly weather observations . Each observation of 'dust', 'blowing dust', 'duststorm', 'sand', 'blowing sand' or 'sandstorm' was used to denote 1 h of dust. Measurement protocols are standardized across all measurement stations and are strictly defined by the Environment Canada Manual of Surface Weather Observations (Environment Canada 1961. These protocols closely follow World Meteorological Guidelines for observations (O'Loingsigh et al 2010), but are unique to Canada (Environment Canada 1961. A review of archival metadata and station reports indicates that no significant changes in manual observation methods occurred throughout the period of record. To quantify CWEP for a given time period, we developed a metric relating the erosive power or transport capacity (TP) of wind to aridity: where P/PE is a measure of aridity (precipitation divided by potential evapotranspiration). Erosive conditions correspond to higher CWEP values, which indicate either windy and/or arid conditions. First, TP was calculated for every hour using wind data from the Environment Canada 'Homogenized Surface Wind Speed' database (Wan et al 2010). TP was calculated with the Kawamura (1951) transport equation: where Q is streamwise sediment transport, C is a constant (2.78), ρ a is air density (1.22 kg m −3 ), g is the acceleration of gravity (9.81 m s −2 ), u * is surface shear stress, and u * t is wind erosion threshold (constant at 0.185 m s −1 ). Q was related to TP by: where t is the interval of measurement (3600 s), and ρ s is the density of sediment (1600 kg m −3 ). Surface shear stress was calculated with the Law of the Wall using parameters κ = 0.4, z = 10 m, z 0 = 0.01 m (see Shao 2000).
Monthly precipitation (P) data were obtained from the Environment Canada 'Adjusted Precipitation' database, which are standardized and corrected for evaporation, wind, and gauge wetting losses (Mekis and Hogg 1999). Potential evapotranspiration (PE) was calculated using the Thornthwaite (1948) method. Average monthly temperatures used to derive PE were acquired from the Environment Canada Historical Weather database. This produces a CWEP for each month and site of the study. Although specific to this study, our CWEP metric effectively describes the climatic forcing on wind erosion and builds on the widely used dune mobility index (Lancaster 1988).
Land management data came from the Agricultural Census data available from Statistics Canada. Data were collected during the agricultural censuses of 1976-2006, which are conducted at 5 yr intervals. Over the period of study, an average of 283 270 farms reported on activities in each census, each representing an average of 242 hectares. Two of the most common practices for reducing wind erosion are reduction of summerfallow and the use of reduced tillage direct seeding systems. Summerfallow is the practice of leaving fields unused through the growing season. These fields are often exposed and susceptible to wind erosion. Direct seeding is a method of injecting seeds directly into the soil with minimal or no tillage, which helps reduce erosion by minimizing disturbance of the soil. Summerfallow data were available from 1976 to 2006. Direct seeding data were only available from 1991 for Canada, and were extrapolated to 1976 using data from the United States of America (Coughenour and Chamala 2000). We base this extrapolation on the assumption that agricultural technology has seen roughly similar adoption rates between the United States and Canada.
The ratio of precipitation to potential evapotranspiration is of limited value in winter in Canada because plants are non-responsive and non-contributing (Allen et al 1998), thus we only compared dust frequency and CWEP in April and May of each year. The average cumulative proportion of dust hours in April and May for all seven sites is 71.0 ± 17.7% (figure 2(D)), indicating that these months are representative of the most prevalent dust conditions since they experience the overwhelming majority of dust events.

Results
Between 1961 and 2006, a total of 1342 h of springtime airborne dust were reported at the seven sites across Alberta, Saskatchewan and Manitoba (figure 2(A)). Sites in Saskatchewan experienced the greatest number of dust events (Saskatoon: 380 h; Regina: 393 h), followed by Manitoba (Brandon: 235 h; Winnipeg: 185 h) and Alberta (Edmonton: 54 h; Red Deer: 66 h; Calgary: 29 h). When all sites are considered together, there is a subtle declining trend over the period of record, suggesting a broad-scale decline in recorded dust across the southern prairies.
Dust event frequency shows variability at a number of temporal scales ( figure 2(A)). First, a high degree of inter-annual variability suggests that the average frequency of dust hours in a given year does not closely relate with adjacent years ( figure 2(A)). Most stations showed correspondence in the frequency of dust events, indicating that certain years were dustier across the prairies, rather than just at one site. Dust frequency distributions are highly skewed; many years have little to no dust recorded, the mean values are influenced by highly dusty years (e.g., 1981). Broad-scale trends emerge at the decadal scale, with clusters of heightened activity occurring prior to 1991 and less dusty conditions occurring from 1991 to 2006. Visual inspection of the series in figure 2(A) indicates a change in dust regime sometime in the early 1990s. Individually, the time series for each station do not reveal statistically significant changepoints; however, when the data from all stations are totaled for each year, the combined time series reveals a significant changepoint or regime shift in dust frequency at 1990 based on two separate homogeneity tests (Pettitt and Buishand). This is also demonstrated by the dramatic difference in means (µ) for 1961-90 (µ = 2.99 h month −1 ) and 1991-2006 (µ = 0.38 h month −1 ). Thus, average dust frequency after 1990 was statistically different from the preceding period of the record.
The CWEP shows a similar long-term trend to dust frequency in that CWEP forcing has reduced; however, the difference in distribution is less pronounced (means: 1961-90: 15.35, 1991-2006: 11.13; medians: 1961-90: 6.71, 1991-2006: 5.26; all units m 3 m −1 mo −1 P −1 PE). Results of the homogeneity tests do not reveal any evidence of changepoints in the individual CWEP series from each station, or in the combined series from all stations. The CWEP shows inter-annual patterns in variability similar to the dust series ( figure 2(B)). There is temporal correspondence between years with high CWEP and years with more dust hours (e.g., 1968, 1977, 1980-81, 1987-88). Throughout the record, values showed similar decadal-scale levels, although there was a notable drop in CWEP between 1991 and 1996. To examine the potential role of CWEP in explaining dust frequency, we split the dataset at the sharp reduction of dust frequency in 1990 (figure 2(A)). We base this distinction on a visual assessment of a step reduction in dust frequency after 1990, which is supported by results of the homogeneity tests. Results in figure 3 show that the relation between the CWEP and dust frequency was poor prior to 1990 (r 2 = 0.154, p < 0.001), but improved following 1991 (r 2 = 0.429, p < 0.001) (figure 3). Figure 3 also shows that under a given climate forcing, the average expected response in dust frequency was higher on average during 1961-90 than 1991-2006.

Discussion and conclusions
The purpose of this study was to examine long-term trends in dust emissions on the Canadian Prairies and determine whether the efforts of soil conservation initiatives have had a measurable impact on downstream airborne dust frequency. Dust frequency before 1991 showed higher peaks and sustained clusters of dust observations (e.g., 1980s, figure 2(A)). Dust frequency post 1991 showed fewer hours of dust. In particular, the droughts of 2001-02  were not as dusty as in previous years despite erosive conditions. We attribute this to changes in farming practices, which have trended toward methods that reduce tillage and decrease soil erosion losses (figure 2(C)). This could be the result of numerous initiatives that promoted soil conservation techniques in the late 1980s (table 1).
Our study shows similar results to those by Todhunter and Cihacek (1999) and Stout and Lee (2003) whereby a reduction in observed dust was noted in the past couple decades and attributed to improved land management. However, we provide further evidence to support this by exploring systematic changes in farming practices with census data and establishing a changepoint in the dust series. We also differ in that we use a metric of climate forcing that synthesizes both wind power and aridity, both which can be related more directly to wind erosion potential.
Although changes in farming practices have occurred (figure 2(C)), it is difficult to attribute them directly and quantitatively to the soil conservation initiatives. For example, direct seeding is also promoted to preserve soil moisture in the early season. However, regardless of the rationale made by individual farmers, the techniques do have the overall effect of reducing soil erosion.
The closer correspondence between dust frequency and CWEP post 1990 (figure 3) suggests that climate now plays a larger role in determining wind erosion. From 1961 to 1990 it is likely that both poor land management and climate were responsible for airborne dust. This may be viewed as a heightened state of land susceptibility to wind erosion, whereby the magnitude of climate forcing required to initiate dust emission was lower than 1991-2006. However, with improved land management post 1990, only the most severe wind erosion conditions (dry and windy) resulted in measurable observations of dust. This suggests that farmers on the Canadian Prairies are better equipped to handle conditions with moderate climatic erosion forcing now than in the 1980s. However, results also show that it may be difficult to completely eliminate wind erosion; extreme droughts and/or windy periods will always result in some wind erosion as in the droughts of 2001-2, for example .
Our interpretation is that the strong reduction of dust after 1990 represents a region-wide threshold crossing, whereby the progressive shift in soil conservation practices began effectuating a change in the dust frequency. From our analysis, CWEP does not appear to explain the step change. Indeed, there are other factors that could be involved, but in this environment, where wind erosion is naturally restricted by the presence of vegetation cover, the most important factor is whether the soil is exposed to wind by agricultural practices. Thus, any land-use activity that exposes soil will render the surface susceptible to wind erosion, and conversely, any land use that reduces this exposure should decrease dust frequency. We suggest that the landscape was close to the threshold crossing prior to 1990, but required greater effort in soil conservation to tip the scale in favor of reducing dust.
Similar to other studies (Wheaton and Chakravarti 1990, Todhunter and Cihacek 1999, Stout and Lee 2003, our study detail is constrained by data quality. Meteorological measurements are not ideal indicators of airborne dust. Observations of visibility and meteorological conditions are performed with a variety of people and although concretely defined (Environment Canada 1961, there is room for variability in reporting. Dust events can be secondary to other weather and therefore can be missing from the record (see O'Loingsigh et al 2010). Additionally, meteorological measurements have poor spatial coverage and are likely to be influenced preferentially by dust sources directly upwind from the observation station. Census data describing direct seeding and summerfallow also lack the ability to fully describe all changes in land management that have reduced wind erosion. It is clear that better quality data are required to answer the questions posed in this study with finer spatial and temporal resolution. However, regardless of the limitations of the data, it is clear from our analysis that airborne dust has declined in the southern Canadian Prairies, beginning notably after 1990 (figure 2(A)), and that this decline is coincident with improved farming practices (figure 2(C)), which may signal the success of soil conservation initiatives (table 1), especially those initiated in the 1980s and thereafter.
Given the potential negative economic and health effects of wind erosion, a case for more detailed monitoring is easy to justify. Severe droughts and climatic wind erosion conditions equal to or exceeding those in the 'dust bowl ' and 1980s have the potential to occur in the future. Data from our study suggest that improved farming techniques could reduce (but not eliminate) airborne dust in these events. Future studies and increased monitoring of dust frequency are required to further refine the controls of agriculturally derived airborne dust on the Canadian Prairies. Despite this, our synoptic study demonstrates that dust frequency has reduced and can be attributable to soil conservation.