In the third post looking at the 2014 incident records published by the London Fire Brigade, we explore if there are any temporal patterns
The image above is a plot of the coordinates recorded for every incident from 1st January to 30th November 2014 within the Greater London area, coloured by the type of property attended – residential dwellings (red and orange), non-residential (purple), outdoor structures (green) and vehicles (blue and grey). See if you can spot the cluster of grey dots representing aircraft. Shouldn’t be difficult if you read the last post…
In this post, we’re going to look at the London Fire Brigade (LFB) incident records for 2014 from a time perspective. As with the last post, we’ll first look at differences between incident categories, and then between properties.
First up, is there any variation across the seasons?
The vagaries of the British weather may be highlighted by flooding incidents rising from June to August… but the most interesting pattern is how automatic false alarms increase from June through to October whilst Fire incidents decrease. It will be interesting to see if this is an annual pattern. We’ll be looking at the five-year trends in another post.
Looking at the categories as a stacked bar chart for both months of the year and days of the week doesn’t show many surprises:
In the months, February is the lowest but there is a significant increase in incidents under the ‘Make Safe/Evacuation/Rescue’ category. We had a lot of flooding in London and the surrounding areas last February. It’s possible that some incidents were categorised as ‘Evacuation’ and ‘Rescue from water’.
The most interesting breakdown to explore is hourly. What happens to the volume and type of incidents throughout a 24-hour day?
Plotting each category individually, it’s clear to see that false alarms caused by automatic fire alarms rise rapidly during the morning. They decline in the afternoon whilst incidents of actual fires continue to rise.
Plotting as a stacked bar chart shows how the total volume of incidents varies throughout the day and highlights the increase in fires in the afternoon as false alarms decrease:
And finally, visualising the percentages for each hour of the day:
Plotting as percentages more clearly emphasises the shift from actual fires and manual false alarms or no action required to false alarms triggered by automatic fire alarms from 6am to 11am.
Let’s have a look at the time spread by location type to see if there is any suggestion of correlation:
In the morning it’s a competition between residential dwellings and non-residential premises but by midday, incidents at dwellings continue to rise while those at non-residential tail off. Outdoor incidents rise steadily from 6am to 6pm before tailing off, with a jump at 5pm that coincides with post-work ‘Happy Hour’. Will have to dig deeper to test that assumption…
Plotting as a stacked bar chart again helps show how all incidents are divided up amongst the property categories:
There does seem to be a correlation between the number of automatic fire alarms causing false alarms and the rise in visits to non-residential locations during the morning.
Visualising as percentages helps highlight the pattern. Incidents involving non-residential properties clearly rise during the morning before tailing off and flattening out in the afternoon, which is a similar pattern to false alarms triggered by automatic fire alarms.
Just by plotting the basic information, some patterns have already begun to emerge. In the next analysis, we’ll be exploring the mobilisation data, distance travelled and local variations in types of incident.