Frequency of SAR Incidents

Frequency of SAR Incidents
SAR Incidents by Week and Type (2003 to 2010)

Most SAR teams maintain records of each incident independently from PEP, but even if they didn’t I’m sure they would all agree; there are more SAR incidents in the summer than in the winter. Further, calls come with greater frequency on the weekend, and even more frequently on long weekends.

Since every incident in the PEP Incident Summaries has a date associated with it, it’s relatively easy to chart what’s called a frequency distribution. Simply put, it’s a count of how many incidents occur in a given period of time. Choosing the right resolution on the time period can tell you a lot of information.

Incidents by Year and Type

First, the simplest time period: the year. Here are the number of incidents per year; I took the liberty of creating a stacked bar chart breaking down the incident types:

SAR Incidents by Year and Type 2003 to 2010
Figure 1: SAR Incidents by Year and Type 2003 to 2010

Figure 1 shows that the type of Incidents (Ocean, Land, Air, and Inland Water) seem to be about the same proportion year-to-year, and a clear increasing trend, although not dramatic. You can also see that he distribution of search types each year seems to be very regular.

Incidents by Region, Revisited

In my last post I showed a “doughnut chart” chart of incidents by region. Figure 2 charts the values as lines, by year. This example is easier to read the absolute values of each region (total number of searches per year), and you might get the impression (as I debunked) that the Southwest region has more searches.

SAR Incidents by Region (2003 to 2010)
Figure 2: SAR Incidents by Region (2003 to 2010)

Whereas the bar chart is good at showing the proportion of incidents from any region, and the pie chart is good at showing this for a small number of categories, the line chart is far better at comparing the absolute values of the different regions.

Incidents by Season

SAR Incidents by Season (2003 to 2010)
Figure 3: SAR Incidents by Season (2003 to 2010)

Figure 3 shows a slightly different view of incidents broken down by season. For this I used the definition of the seasons in the Northern Hemisphere; Spring (March, April, Map), Summer (June, July August), Fall (September, October, November) and Winter (December, January and February). This shows what most SAR members know by intuition, that there are more searches in the summer months. However, for planning it’s not very useful.

Incidents by Month and Type

A slightly more interesting question is how does the frequency vary by month? The first pass is to chart the percentage of all tasks by month.

SAR Incident Frequency by Month (2003 to 2010)
Figure 4: SAR Incident Frequency by Month (2003 to 2010)

Clearly July is the busiest month on average, followed by August. There is a deep drop off for November. However, it looks like the various incident types aren’t proportional for every month. A different view of the above shows what is happening.

SAR Incidents by Month (2003 to 2010)
Figure 5: SAR Incidents by Month (2003 to 2010)

It’s clear that the Inland Water searches are far more frequent in the summer than in the winter. This makes sense when you consider that most inland water in BC is frozen in the winter and the ice is breaking up in the spring, but this chart shows me that in the busiest month, inland water searches account for almost 1/3 of all searches in BC which is very significant. It’s also interesting to note that ground incidents are fairly stable from January through May, and increase to a peak in August – compared to the previous chart where the peak is in July. The peak is shifted by the prevalence of inland water incidents. The clear dip in search frequency is November is also notable.

SAR Incidents by Month (2003 to 2010)
Figure 6: SAR Incidents by Month (2003 to 2010)

In figure 6, I charted the frequency of searches for every month in the period from 2003 to 2010 to get an idea of how the pattern looks when it’s not averaged for every year. This shows the variability year to year. I also plotted the trend line in red to get an idea of how fast the number of incidents is increasing.

You can see that some years August is the busiest month, while in most of them it is July. You can also see that the pattern of Incident frequency we saw in figure 4 is repeated regularly every year, and that the busiest search month every year always has more incidents than any other month in the year. Furthermore, you can see that the difference in magnitude from the average to the maximum is about double – meaning the standard deviation is quite large.

Overall we can say that search frequencies are very regular, meaning that they repeat on a patter, but highly variable, meaning that the variability within any year is larger than 50%.

We should be able, with a little math, to confidently predict how many searches the province will have in any given month. An analysis like this can also help determine what the rate of increase in searches is. I’ll probably attempt this in another post.

Incidents by Week

Charting incidents by week might seem like a complete waste of time, but it does give you insight into the continuous volume of incidents.

SAR Incidents by Week and Type (2003 to 2010)
Figure 7: SAR Incidents by Week and Type (2003 to 2010)

Figure 7 shows each search type as a separate line, the total searches is included as it’s own line, and the mean of all searches in the province over the time period is drawn as the red line. This shows that there is an average of average 21 frequency of searches per week in BC.

SAR Incidents by Week (2003 to 2010)
Figure 8: SAR Incidents by Week (2003 to 2010)

In figure 8 I removed the different search types in order to see the total better. I included the mean in red, and the trend line in orange. In this chart you see the same pattern as expressed in figure 5 (Incidents per Month) with more “noise”; You can also see that while the average incident frequency (per week) is 20 to 21, the maximum in July is approximately 45 to 50 – more than double. The jaggedness of the line can tell you the variability week-to-week, but the overall pattern is still visible.

The July peaks on this chart seem to coincide with the July long weekend. In 2004 it looks like there was a double peak. In a future article I want to examine the coincidence of long weekends and search volume.


Simple statistics on the time-based series of SAR Incidents in BC can give you a very good idea of search frequency and distribution over time. Using this information, planners could allocate resources for peak periods if necessary. An increasing trend in incident frequency was noted, although it is not large. The seasonal contribution of inland water searches was noted; this pattern is different from the rest of the data, enough that it may be considered as a separate trend.

3 Comments on “Frequency of SAR Incidents

  1. I find this quite interesting. To an individual, an emergency or SAR event might be significant life changing/defining occurrence with a feeling of being "random" and "unexpected". I have always been surprised how regular and predictable emergencies are. The daily fluctuation of call volume in my day job is shockingly narrow.

    As for your thoughts on SAR forecasting, I think that a predicting tool or algorithm would be more accurate if you were to use historical data and compare it to external factors. primarily weather, but also wildfire activity, avalanche hazards, day of the week, long weekends etc.

    For example, as weather improves coming in to Summer, Inland SAR activities will increase, until it becomes too hot or areas are closed due to fire bans etc. Activity might then shift to water activities etc. Taken to the next level, you could predict specific search types on a given day. Such as backcountry winter searches might be more likely during mild/sunny weekends unless the avi conditions become high and scare everyone away. (or it might predict the opposite, even though there are less users during poor weather, the increased risk might increase the likelihood of a SAR response)

    I realize that this would take an enormous amount of data mining to plot historical regional weather forecasts, avi conditions, etc against historical SAR responses, but you seem to be up for a challenge….

    At the very least, I think that any SAR response algorithm has to consider weather as a variable.

  2. Wow Tom, huge comment.

    My first response is that I believe that incident rates are purely a function of the number of people recreating. The primary reason I think this is the case was outlined in my flowchart (which admittedly need work)

    Another reason why I do not think weather is a dominant factor is the extreme regularity of the accident rates.

    Weather may be a factor in accidents, but without people it's just weather. Really bad weather reduces the number of people in the backcountry, so climbers and mountaineers will stay home when the forecast is really bad. The reduced numbers mean reduced accidents.

    For example, there is a "double spike" in the stats for July 2004. Sylvia's (my partner's) first response when I showed her this was "might have been bad weather on the long weekend so people putt off their trips." Which I thought was as good an idea as any.

    Which brings me to a way to test your hypothesis against mine; if we can find a source for historical weather data we could try and see if there is a correlation between high accident weekends and bad weather.

    I'm leaning toward a correlation between holiday long weekends though.

  3. Sorry, I probably wasn't clear. I was thinking more that weather would predict the numbers of people going out and the types of activities they might be doing. I agree that number of people recreating is the key factor. Otherwise there wouldn't be spikes over weekends.

    But that the unexpected surprise in the data might be that bad weather results in more tasks.

Leave a Reply

Your email address will not be published. Required fields are marked *