An analysis of the movement of dairy cattle through 2 large livestock markets in the province of Ontario, Canada

Abstract

Data pertaining to the movement of dairy cattle through 2 large livestock markets in the province of Ontario were collected for 1 week per month throughout 2004. Counts and postal codes of sellers and buyers of adult dairy cattle, veal calves, and dairy calves were obtained. Three assumptions were made to represent the level of mixing among animals that could take place at the markets. We estimated the number of livestock holdings that could be exposed to a highly contagious disease agent, should infected animals have been sold through the market in the same week. The estimates ranged from 8 to 20 holdings, when assuming no mixing at the market, to 51 to 171 holdings when assuming complete mixing. These markets are important hubs in the dairy cattle movement network in Ontario and pose the risk of infecting a large number of livestock holdings should animals infected with a highly contagious disease agent pass through them.

Introduction

Livestock markets play a central role in dairy cattle movements in the province of Ontario. Almost all adult slaughter dairy cows and finished veal calves in Ontario are sold to processing plants at livestock markets. In addition, these markets are an important selling place for bob calves, which are male calves sold by dairy farms to veal producers to be either milk- or grain-fed. Dairy heifers may also be purchased by farmers and dealers at livestock markets. By mixing animals from various sources and then dispersing them to a potentially high number of recipients located in different, sometimes distant, geographical locations, there is a high risk of spreading a disease such as foot-and-mouth disease (FMD), should infected animals have been sold at a market.

One well-known example of the role played by livestock markets and dealers in spreading an infectious disease occurred in the 2001 FMD outbreak in the United Kingdom (UK) (1). One livestock dealer purchased infected animals at one market and assembled these animals with a large group of healthy sheep for sale at another market in the country. This activity resulted in the spread of the FMD virus to at least 25 other holdings, including another 8 dealers, and across 9 of the 12 major geographical areas subsequently involved in the outbreak. It is likely that because of the livestock management structure in Canada, livestock markets could play a similar role in spreading infected animals throughout the country. However, there have been very few published studies on the activities of livestock markets in Canada (2).

Studies evaluating the role that livestock markets and dealers play in moving livestock have been published in the UK (3,4). These studies used network analysis techniques, a process that classifies livestock holdings as nodes, and the movements of animals among holdings served as links between the nodes of the network. Such network analyses were used to evaluate the importance of various holdings in the movement of cattle. The authors reported that markets are important “hubs” in networks of livestock movements as they play a central role in linking pairs of livestock holdings. Quickly stopping movements and sales at markets and quickly determining if infected animals have gone through a market may therefore reduce the extent of widespread epidemics.

We previously studied the movements of adult milking cows among Ontario dairy farms enrolled in the Dairy Herd Improvement (DHI) Program (5), currently the only readily available source of livestock movement information in the province. In order to improve our understanding of dairy cattle movements in Ontario, the movement of adult dairy cows, dairy calves, and veal calves and the activities of dealers and markets should be evaluated. One of the best ways to obtain this information is to study the movements of dairy cattle through livestock markets.

The main objective of this study was to describe the movements of dairy cattle through 2 large markets in the province of Ontario, Canada. A secondary objective was to determine the potential number of livestock holdings that could be exposed should infected animals have been sold at a livestock market in a given week of sales.

Materials and methods

All sales of dairy-type animals through 2 large auction markets (named Market 1 and Market 2) in Ontario were obtained for 1 wk per mo in 2004, typically from the 12th to 18th day of each month. These markets were purposely selected based on their location in the province (Market 1 in southwestern Ontario and Market 2 in eastern Ontario) and their sales activity, both being important large markets in the province. Records provided by the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA) showed that out of 1 166 791 livestock that were sold in 25 markets in Ontario in 2004, 84% and 16% were sold in western and eastern Ontario, respectively. Market 1 was third largest in the west, selling 27% of livestock. Market 2 had the largest volume of sales in the east, selling 30% of livestock sold in that region. Both markets sold dairy-type animals 2 d per wk. These 2 markets were chosen initially because of their size and because they kept both paper and electronic records of sales. It was not until the data collection portion of the study began that we discovered that both auction markets overwrote their electronic sales databases after only 1 wk. As a result, sales transactions for 2004 at both markets were available on paper invoices only, so all data had to be re-entered into an electronic database. To evaluate sales trends for the entire year, we chose to sample sales information for 1 complete week per month in 2004. Information recorded included: the postal code, a unique identifier, the number and type of animals sold per seller, and the postal code and unique identifier of the buyer. Sellers were linked to buyers in the database so that the destination of shipments from each seller could be followed. A transaction was then identified as the sale of ≥ 1 animal from a seller to an individual buyer that week.

A list of permanent buyers was obtained from each market that allowed us to identify the buyers as one of the following: processing plants, farmers, dealers, truckers, auction markets, food markets, and “other” such as insurance companies and universities. The type of seller was available at Market 2, but had to be deduced at Market 1. At Market 1, the unique identifier and postal codes of sellers were matched against: 1) the list of permanent buyers obtained at that market; 2) the list of licensed dealers and truckers obtained from the Ontario Cattlemen’s Association (http://www.cattle.guelph.on.ca/cgi-bin/text.pl?Zone=1&Type=); and 3) the list of dairy farms in the Dairy Herd Improvement (DHI) database for Ontario. The DHI database stores lactation information for all cows in registered herds and includes names and postal codes of producers; this permitted matching of buyers with sellers. Sellers were classified as farmers, dealers, truckers, unknown, and other, such as insurance companies, universities, or the market itself when it purchased animals. In addition, employees at Market 2 were asked to categorize farms that sold feeder calves as dairy or beef farms. We suspected that a large proportion of these were bob calves sold by dairy farms. Only feeder calves sold by dairy farms were included in the analysis.

Dairy-type animals sold were classified into 1 of 3 categories: 1) adults (adult cows, bulls and heifers); 2) veal calves destined for slaughter; and 3) dairy calves destined for further rearing. The sales records for all dairy-type animals sold were then extracted from the database. In order to build networks, a unique node number was assigned to each individual seller and buyer.

Description of sales at the 2 markets

The number of transactions and the number of dairy-type animals sold by category at each market were plotted to determine if there was a seasonal trend at either market and if selling patterns by animal category sold differed between the 2 markets. Based on the observed pattern of sales in each of the 12, 1-week samples collected in 2004, we selected a subset of 4, 1-week samples in the months of January, May, August, and October to represent different seasons.

The number of individual buyers and sellers per wk at each market, the number of dairy-type animals purchased by type of buyer and the number of dairy-type animals sold by type of seller were averaged for the 4, 1-week samples to provide a single estimate for an average week of sales. The same analysis was repeated for each of the 3 categories of animals sold.

The province of Ontario was divided into 4 geographical regions (eastern, northern, central, and western) based on the first letter of the postal code in order to represent the geographical region of the buyers and sellers at the 2 livestock markets because actual locations of buyers and sellers were not available. Out-of-province movements were included in the analysis and were classified into a single group, except at Market 2 where movements into and from the province of Quebec were recorded separately because of the quantity of these interactions with that province. The numbers of animals sold per seller and animals purchased by buyer were obtained and averaged for each week of sales. The average median value and range of maximums were reported as the results were not normally distributed. The statistical software package Stata was used for all descriptive analyses (StataCorp 2005; Stata Statistical Software: Release 8.0; College Station, Texas, USA).

Estimating potential epidemic size

The networks of buyers and sellers were used to estimate the maximum number of buyers that could have purchased potentially exposed animals at a market following the sale of animals infected with a highly contagious disease. All sales taking place in the same week were grouped, even though they may have occurred on different days. Three measures were calculated at both markets: 1) the level of mixing, which represents the mixing of animals at the market through the process of sale, for example by using the same facilities and equipment, 2) the level of dispersal of sales, reflecting the dispersal of a shipment from a single source to multiple recipients, and 3) the level of assortment, to explore the assembly of animals from various sources for purchase by a single buyer.

Three assumptions were made to explore different levels of mixing and obtain estimates of potential epidemic size: 1) no mixing meant that if a seller sold infected cattle, only the cattle belonging to buyers purchasing animals from this seller could be exposed; 2) moderate mixing within category of cattle sold, meant that if infected adult dairy cattle were sold, for example, all cattle belonging to buyers of adult dairy cattle within the week were potentially exposed; and 3) complete mixing meant that cattle belonging to all buyers purchasing dairy-type animals within the week were considered exposed. If buyers or sellers were present at both markets during the same week, the potential epidemic size, depending on the assumption made in terms of mixing, could be affected by the number of individual buyers at both markets, either within category of animal sold or overall in the case of complete mixing. To truly estimate the number of potentially exposed buyers’ premises at a market, one would have to determine the effective contacts among animals sold at the market and then consider factors such as host susceptibility, infective dose, and pathogen virulence. The 3 simplified assumptions of level of mixing were used to explore 3 potential scenarios of spread. Within each level of mixing used, therefore, all exposed animals were considered, regardless of the factors listed above.

The level of dispersal was calculated by obtaining the number of individual buyers for each seller that sold animals at the market in a week, also known as the out-degree (6). For example, should a farmer have sold 4 cows that were purchased by 2 buyers, that farmer’s out-degree would have been 2. The level of assortment was obtained using the in-degree, which is the number of individual sellers per buyer (6). A buyer would have an in-degree of 10 should that buyer have assembled animals from 10 different individual sources at the market in a week. UCINet 6.186 for Windows: Software for social network analysis (7) and Pajek 1.23 (8, http://vlado.fmf.uni-lj.si/pub/networks/pajek/) were used for all network analysis measures and for visualizing the networks.

Results

Description of sales at the 2 markets

There was an average of 1289 [standard deviation (s) = 142.7, med = 1311, max = 1624] dairy-type animals sold at market 1, and 1118 (s = 199.3, med = 1178, max = 1388) dairy-type animals sold at Market 2 (Figure 1), including all feeder calves per week. The pattern of transactions and numbers of dairy-type animals sold were similar for both markets between the sample weeks in the months of January and August that year, but differed in the fall (September — December). Sales peaked at Market 2 while they were at their lowest at Market 1 during the fall sample weeks.

Figure 1.

Number of dairy-type animals sold at (a) Market 1 and (b) Market 2 by type in 12 weekly samples in 2004.

Adult dairy cattle were sold in the greatest numbers (60% of dairy-type animals sold) at Market 1 while most dairy-type animals sold at Market 2 (62%) were calves (Figure 1). A number of feeder calves were indeed beef calves, or were sold at the market by beef farms. These were removed from the calculations as they were not the focus of the study: 26, 52, 128, and 140 beef calves were sold in January, May, August, and October, respectively. Up to 75% of sellers at both markets sold up to 3 to 4 animals in a week of sales. There were averages of 408 and 316 individual sellers per week at Markets 1 and 2, respectively. In both markets, farmers were the most frequent sellers of animals: 63% of sellers at Market 1 and 89% of sellers at Market 2, accounting for 60% and 67% of animals sold, respectively. Dealers played an important role at Market 2, selling an average of 33% of dairy-type animals.

Buyers purchased a median number of 4 animals per week of sales, although an important level of heterogeneity in these values was observed. The number of individual buyers per week at both markets was significantly lower than the number of sellers: 93 and 62 buyers per week on average at Markets 1 and 2, respectively. Farmers represented 45% of buyers at Market 1 while dealers represented 44% of buyers at Market 2, closely followed by farmers at 37% on average every week. Processing plants purchased an average of 63% of animals, mostly adult cattle, while farmers purchased 23% of animals at Market 1. Farmers purchased 49% of dairy-type animals at Market 2, while dealers purchased 27% of cattle on average per week at the same market. Three buyers at Market 2 were classified as feedlots and purchased 78% of calves on average every week. Dealers were the second most important purchasers of veal calves at both markets.

Most sellers (90%) and buyers (74%) at Market 1 were from southwestern Ontario and the rest were from central Ontario. There were shipments of animals to and from more distant areas: to eastern Ontario and the western part of the province of Quebec. In the case of Market 2, sellers (83%) and buyers (84%) were from eastern Ontario, while the rest were from the western part of the neighboring province of Quebec (east of Market 2).

Estimating potential epidemic size

The distribution representing the level of dispersal of sales of all dairy-type animals at each market is shown in Table 1. The distribution representing the level of assortment, shown in Table 2, was right-skewed showing a high level of heterogeneity in the number of sellers per buyer at both markets. This heterogeneity, represented by the small proportion of high values observed in the distribution, was a result of sales of adult cattle at Market 1 and calves at Market 2 (data not shown).

Table 1.

Out-degree distribution by seller, representing the level of dispersal of sales of all dairy-type animals at Markets 1 and 2 in 4 sample weeks in Ontario, 2004, assuming no-mixing at the market

	Percentiles
Sample week	Mean out-degreea	25th	50th	75th	95th	99th	Max
Market 1
January	2.0	1	2	2	5	8	13
May	1.7	1	1	2	4	5	8
August	1.9	1	1	2	4	7	14
October	2.1	1	2	2	5	8	19
Market 2
January	1.8	1	1	2	4	7	8
May	1.7	1	1	2	4	5	9
August	1.8	1	1	2	4	7	14
October	2	1	2	2	5	8	20

^aOut-degree — number of individual buyers per seller at the market. For example, a mean out-degree of 2 means that a seller sold animals to 2 buyers on average during that sample week of sales at the market.

Table 2.

In-degree distribution by buyers, representing the level of assortment of sales of all dairy-type animals at the 2 markets in 4 sample weeks in Ontario, 2004, assuming no-mixing at the market

	Percentiles
Sample week	Mean in-degreea	25th	50th	75th	95th	99th	Max
Market 1
January	8.9	1	2	7	40	209	209
May	9.5	1	3	7	24	201	201
August	7.6	1	3	8	31	132	132
October	7.1	1	2	7	24	109	109
Market 2
January	9	1	1	4	54	109	109
May	10.6	1	2	3	73	145	145
August	9.7	1	2	4	59	159	159
October	8.6	1	1	4	52	132	132

^aIn-degree — number of individual sellers per buyer at the market in each sample week. For example, a mean in-degree of 9 means that a buyer purchased from 9 different sellers, on average, in that week of sales at the market.

The level of mixing assumption had an important impact on the number of buyers’ premises potentially exposed through the market. When assuming complete mixing at the market among all animals sold, the resulting networks presented a “hub and spokes” appearance showing the market at the center of all buyers and sellers (Figure 2a); the market acts as a hub, connecting all buyers and sellers. Figure 2a shows the important role that a few buyers and sellers have in connecting the buyers and sellers at the 2 markets that week. When no mixing was assumed at the market, the network topology changed significantly (Figure 2b); the markets were removed so that buyers and sellers were linked directly, giving the impression that the market itself is not necessary in linking the buyers and sellers. This network diagram (Figure 2b) represents a direct farm-to-farm potential transmission network. Moderate mixing is shown in Figure 2c and represents the same type of topology as Figure 2b; however, the number of nodes in the network is reduced because the network is now compartmentalized by animal type. In both Figures 2b and 2c, we again see the important bridging role that a few buyers and sellers have in linking the networks of Markets 1 and 2.

Figure 2.

Network diagrams representing the movement of dairy-type animals from sellers to buyers using 2 markets in Ontario, under different mixing assumptions in the week of October 11, 2004. (a) Assumes complete mixing at the market and therefore links cattle from all sellers to the market and cattle from all buyers at the market for that week. (b) Assumes no mixing at the market and therefore links cattle from the seller directly to cattle from the buyer. (c) Assumes moderate mixing, which links cattle from sellers and buyers of dairy cattle within age category. The location of the nodes does not represent actual geographical location.

The predicted number of potentially exposed livestock holdings following the sale of infected animals at the 2 markets in Ontario is shown in Table 3. This demonstrates the impact of the mixing assumption on the predictions. As presented above, in the weekly sample from the month of October, dealers linked the 2 markets together by buying and/or selling at both markets. Therefore, the potential epidemic size could include cattle from buyers at the 2 markets if we considered complete mixing or partial mixing (within animal category). The potential epidemic size when assuming no mixing was not affected that week as it represented only cattle from the individual buyers of a load of animals from a seller.

Table 3.

Number of buyers of potentially exposed cattle at Markets 1 and 2 averaged across the 4 weekly samples under the 3 mixing assumptions in Ontario, 2004

	No mixinga	Moderate mixing (within age category)b	Complete mixingc
	Maximal number of buyers of potentially exposed cattle at Market 1 based on the mixing assumption used
Range from 4 weekly samples in 2004	8–19	Adults 59–93 Veal 20–26 Calves 20–63	85–171
	Maximal number of buyers of potentially exposed cattle at Market 2 based on the mixing assumption used
Range from 4 weekly samples in 2004	8–20	Adults 30–93 Veal 9–23 Calves 21–63	51–171

^aWhen no mixing is assumed, only the maximal number of individual buyers per seller in a week of sales is used. The values in this table show the range of these maximal values over the 4 weekly samples.

^bWhen moderate mixing is assumed, the maximal number of individual buyers of animals by type (adult dairy cattle, veal, or calves) in a week of sales is used. The values in this table show the range of these maximal values over the 4 weekly samples, broken down by type of animals sold.

^cWhen complete mixing is assumed, all buyers of dairy cattle at the market during a week of sales represent the maximal number of buyers of exposed cattle. The values in this table show the range of these maximal values over the 4 weekly samples.

Discussion

The movements of adult milking cows among DHI farms in the province were studied to estimate the number of livestock holdings that could be exposed following the introduction of a disease such as FMD in Ontario (5). These estimates represented the low-end of the potential epidemic size because they only included the movement of adult milking cows from farm to farm captured within the DHI database. They did not include the movements of cattle through markets, nor did they include the movement of heifers and calves.

In this study, we show the important role that the 2 selected markets could play in spreading infectious disease agents to a high number of livestock holdings within and outside Ontario. Because of their role in selling adult cows for slaughter and bob calves, these markets are critical in the dairy cattle movement network. Three concepts are important to consider when studying the movements through markets: 1) the level of mixing at the market as a result of direct contacts among infected and susceptible animals and indirect contacts among animals through sharing rings, pens and equipment; 2) the level of dispersal, meaning that infected animals from a single seller may be purchased by > 1 buyer, leading to a dispersion of infected animals in the population; and 3) the level of assortment that shows how buyers will purchase animals from a number of sources. To our knowledge there are no published results of the amount of mixing that occurs at livestock markets, although mixing of animals at markets has been suggested as being important for disease dissemination (1,9).

Mixing at the market results not only from animals sharing pens and rings or having close contacts through fences, but might result from being in the same area where localized airborne spread is possible. Indirect contacts such as handlers moving within the market and the use of equipment among the facilities are also part of mixing at the market. A simulation study in California (10) considered mixing at the California State Fair to be complete, meaning that animals from all livestock holdings represented at the fair could be in contact with each other. Only a contact rate determined the level of connections among animals of different holdings. For this reason, we explored 3 assumptions about mixing at the market.

A complete mixing assumption represented the upper bound of potential number of infections, as suggested in Robinson and Christley (4). A more conservative assumption was used when mixing was assumed to have occurred only within type of animals sold as it was felt that certain groups of animals could be sold at the same time and have more chances of coming into contact with each other. This assumption provided moderate epidemic size estimates compared with the other 2 assumptions. Finally, the most optimistic and probably unrealistic assumption did not consider any spread at the market and led to the lowest estimates of potential epidemic size. The considerable differences in potential epidemic size across different mixing assumptions suggest that it should be a priority to investigate the level of mixing occurring in these livestock markets. Accounting for this assumption will be critical when modelling the spread of FMD both within markets and within the general population of farms. In addition, factors such as host susceptibility, pathogen virulence and the infective dose should be considered in modelling studies in order to improve the realism of spread predictions. As shown in this study, not accounting for this mixing or transmission at the market will underestimate the potential spread and impact of markets in an outbreak. It might be particularly useful to make note of changes in these levels over time in order to keep abreast of possible increases in the risk of larger potential epidemics so that disease control personnel are not taken by surprise at the time of a real incursion.

The level of dispersal observed in the 2 markets was lower than observed in the study by Robinson and Christley (4). In that study, 41% of movements resulted in cattle being dispersed to between 2 and 4 animal holdings, but extremes occurred where cattle originating on 1 holding were dispersed among 62 holdings. This difference can be explained by the fact that the UK study included beef cattle in the number of animals sold per seller. In the 2 markets of the present study, most buyers sold only a few animals, which automatically restricted the level of dispersal possible at the markets. This probably resulted from the way the dairy industry functions in Ontario. Calves are born at regular intervals throughout the year and most farmers will produce their own replacement heifers but often have some excess calves to sell. Slaughter cows will also be sold throughout the year with a low occurring in the summer and early fall as a result of the demand for milk in fall for which producers tend to keep their milking cows, and a peak occurring in late fall and early winter. The pattern of sales of the different animal categories at the 2 markets in this study was suggestive of these trends; however, as our study only included 12 wk in 1 year, we could not draw conclusions. This hypothesis should be explored in other studies through markets in other years.

The level of assortment at the 2 markets was highly heterogeneous and reflected the presence of processing plants and feedlots as buyers of dairy-type animals. In Market 1, processing plants purchased slaughter adult cattle from various individual sources and feedlots purchased calves from various sellers at Market 2. In such situations, there is a potential for infection of a large number of animals, although in both cases purchased cattle are then located (assembled) at 1 site, and in the case of processing plants, only for a short time. One other study evaluated the level of assortment of shipments of steers purchased at livestock markets by a feedlot in Alberta (2). This study showed a significant level of assortment in which each truckload of 60 calves, on average, purchased by the feedlot combined animals from as many as 20 to 30 farms, or a median of 2 calves sold per seller per truckload.

To facilitate our analysis, we grouped the sales that occurred in a single week and assumed that all buyers’ animals could have been exposed, regardless of when the buyers actually purchased animals during the week. For example, at 1 of the markets, veal calves are sold on Mondays, then again on Thursdays along with dairy cattle. Grouping the sales for the week together in this way could have affected our level of dispersal, assortment, and mixing at the market and we may have overestimated the potential epidemic size. In some cases, animals will arrive and leave the market on the day of their sale. But it is also possible to have animals that stay overnight at the market prior to their sale and after their purchase. It was therefore possible to assume that the 2 days of sales were not completely independent of each other and could reasonably be analyzed together. The possibility of indirect transmission through contamination of equipment and facilities added further support to our assumption: if infected animals were sold on the first day of sale of the week, the market could be considered a source of contamination for the remaining sales unless adequate cleaning and disinfection of premises was performed. For this same reason, in a disease outbreak investigation, all sales of livestock having occurred in a period of 2 to 3 wk would be investigated, as all would be assumed to have been potentially exposed, regardless of the day of sale. Therefore, we felt it justifiable to study the week of sales as one group of sales.

An important finding was that most buyers and sellers at these 2 markets were located in the same geographic region as the markets. This was also observed in the analysis of the 2001 FMD UK outbreak where most farmers would sell and purchase at local markets (1). In that study, livestock dealers selling at different markets linked distant areas. Long distance movements for Market 1 could be considered as those going to eastern Ontario and Quebec, or out of province. For Market 2, long distance movements were those going to southwestern Ontario and provinces other than Quebec. Although a small proportion of these movements took place at Market 1, their consequence in potentially seeding infection in distant areas should not be overlooked. Precise distance analysis of movements would have been facilitated by the use of actual location of buyers and sellers. These data were unfortunately not available.

McLaws and Ribble (11) characterized the impact that markets, acting as hubs in a network of livestock movements, can have on the spread of FMD. Through the study of 24 FMD epidemics that occurred during 1992–2003 they found that 1 factor, representing if FMD-infected animals had passed through a market prior to detection of the first case or the implementation of movement restrictions, was likely very influential in leading to the large FMD epidemics in the UK in 2001, and in Taiwan 1997. They concluded that the movement of infected animals through a market was the activity with the highest risk of leading to very large FMD outbreaks. For this reason, it would be important should an infectious disease like FMD be diagnosed in Canada to quickly stop movement activities and identify if infected animals had moved through markets.

Markets can also be an important drain on the resources directed at controlling the spread of diseases. The quality of the sales recording system found at these markets will greatly influence the level of resources necessary to track potentially exposed animals. The 2 markets we studied had computerized records of sales occurring on a weekly basis. However, at the time the data were collected for the study, electronic records were kept in memory for only the past week of sales. As a result, the records used in this study were all paper-based and data collection for 12 wk of sales required a month of full-time work for each market. If an outbreak of a highly infectious livestock disease was to be detected in Ontario a significant period of time could be spent collecting data. We recommend that markets keep a history of sales, in electronic format for at least 3 wk, and preferably 12 mo after the sales.

Although markets could have a great impact in disseminating an infectious disease agent, they are also extremely important to the dairy industry. They represent an important price discovery system in the Canadian dairy market which allows the price of dairy cattle to be dependent upon the supply and demand. All slaughter cattle and most calves are sold at these markets. The closure of these markets would cause great disruption in the industry which, if temporary for the control of a disease outbreak, can subsist. Permanent closure could lead to unintended consequences for which disease controllers are unprepared, such as the emergence of dealer-type activities to compensate for the lack of markets. These types of activities could potentially be much more difficult to control during an outbreak.

Dealers represented an important proportion of buyers of dairy-type animals at the 2 selected markets, especially Market 2, which could lead to a higher number of farms being exposed to a disease should infected animals have been sold at a market. Livestock dealers typically will purchase a number of animals from different farms and sell them at the market. They will in turn purchase animals from various sources at the market and sell them to other markets or farms, sometimes at long distances. As a result, dealers act as smaller markets in the network of animal movements and pose a high level of risk for disease spread because of their activities and their mobility. Because we did not pursue the activities of dealers in this study, the potential epidemic sizes are therefore probably underestimated. For this reason the activities of dealers should be investigated further. Further spread could also occur if all markets in the province were included as we showed that in 1 out of the 4 1-week samples in this study, sellers and buyers could link different markets through sales and purchases. There are 33 registered livestock markets in Ontario based on information from the Ontario Ministry of Agriculture, Food and Rural Affairs (http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/lscsa_list.htm). Because Canada lacks an animal movement database, it is not possible to obtain the required sales information from all livestock markets in order to develop movement networks. Therefore, a complete picture of dairy-type animal movements in the province is not possible at this time.

Due to their importance in the network of dairy-type animal movements, markets in Canada have the potential to play a large role in the silent spread phase of an FMD outbreak as the one that occurred in the UK in 2001 (1). Prevention measures, the communication of principles of biosecurity, and information on the signs of foreign animal diseases in livestock should be made available at markets. In addition, discussions with market owners about the importance of having quickly searchable electronic records of sales should take place on a routine basis. Finally, when faced with an outbreak of a foreign, highly contagious disease, regulatory authorities should place priority upon stopping all sales at markets and determining if any potentially infected livestock had passed through a market.