Vegetation Fires 2009 -2014: a GIS study on

adding value to Greater Manchester Fire and

Rescue Service (GMFRS) data – Report

Donlan, R. And McMorrow, J.

Geography, School of Environment, Education and Development,

University of Manchester, June 2015

Introduction

The purpose of the report was to review the Incident Reporting System (IRS) data

in relation to the fourteen categories of vegetation fire for four financial years, from

1st April 2009 to 31

st March 2014. The report is based on the work carried out

during an Applied Study course and my Dissertation for the MSc in Geographical

Information Science at the University of Manchester, 2014.The intention was to

use GIS techniques to differentiate the smaller vegetation fire incidents and the

factors which influence occurrence from the more resource intensive wildfires; then

compare the geography of wildfire in Greater Manchester using two alternative

definitions alongside that for all vegetation fires, looking at how the adoption of a

definition alters the picture. Then focus on the factors of landcover, the interface

between vegetation and urban areas, as well human influence to explain incident

occurrence and predict areas at greatest risk of ignition leading to a set of

reccommendations based on the findings.

Vegetation Fires

The density of all vegetation fires within the Greater Manchester Fire and Rescue

Service (GMFRS) region during the five year period, was highest around the larger

urban areas to the east and west of the city centre, along the Medlock and Irwell

valleys and around Leigh and Wigan. Figure 1 shows the pattern for spring (March-

May) in the driest three years of the five studied. The areas of lowest density were

suburban areas to the south of Greater Manchester and the suburbs of the

northern towns. The areas with little or no density were the rural margins,

especially the western agricultural areas and the upland areas of the Pennines and

Peak District to the north and east, this may relate to low or zero values on the GM

borders. Figure 2 identifies anomalous spring and summer weather over the five

year period. Temperatures anomalies were periods of temperatures exceeding two

degrees above the mean, while Rainfall deficits were periods when rainfall was

less than 50% below the average, which left vegetation more susceptible to risk of

ignition.

When fires occur

There are typically two fire seasons in the UK, spring (March to May) and summer

(June to August). Spring was the most significant season for the GMFRS area,

taking all outdoor fires together (Fig. 3). The relationship between rainfall

anomalies and the number of outdoor incidents was found to be significant,

Fig 1 – Density of outdoor vegetation spring fires for 2010, 2011 &

2013; the driest of the 5 years ( greatest negative rainfall

anomalies shown in Figure 2)

Fig 2 - Mean temperature and rainfall anomalies for North West England March

2009 to April 2014, based on averages between 1971 and 2000. Negative rainfall

anomalies (blue bars below the line) are dry periods

showing that spring had below 50% of average rainfall in all three months in 2010

while in 2011 and 2013 rainfall was below average in two months. The summer

only had one month (in 2010) below 50% of the average, with most months having

above average rainfall. Temperature anomalies were less of a factor, with fewer

months exceeding the average; the only two spring months were April 2011 and

March 2012, and July 2013 the only summer month which came close to

exceeding that figure.

How the pattern changes from year to year

The density maps (Fig 1) show the distribution of vegetation fires during the three

worst fire seasons in the IRS record. Spring 2010 had an exceptionally high fire

density, with low rainfall over the whole period. Spring 2011 had two drier months

but low fire density, whereas Spring 2013 was not as dry but had higher density

and shows that factors other than weather are involved. The timing of holiday

periods is a significant factor; an early Easter in the spring of 2010 and 2013

coincided with higher fire density. A later Easter in 2011, which based on weather

alone should have the highest density, but actually had lower density.

Defining wildfires

There are currently two proposed defintions to identify wildfire from other less FRS

resource-intensive vegaetation fires; one based on the Scottish Government’s

Operational Guidance for Wildfire manual of 2013, and another under

consideration by the CFOA Wildfire Group (Table 1). Both use the same IRS data

fields and both also include other criteria not operationally collected for IRS.

CFOA’s category 4 is equivalent to the Scottish manual in the IRS data fields it

uses. Category 5 requires two of these criteria to be met, so are the most

resource-intensive fires for GMFRS.

Fig 3 – Wildfire occurrence 2009-2014 plotted against rainfall anomalies

The number of incidents using both definitions were plotted against below average

rainfall anomalies (Fig 3). Similarly to all outdoor vegetation fires, prolonged

periods of below average rainfall led to larger number of wildfires occurring in the

Spring of 2010 and 2011, then to a lesser extent in Spring 2013. March 2012 had

a peak in wildfires linked to both lower rainfall and higher temperatures, but was

not sustained over the whole spring period. This suggests that the same causal

factors underlie all vegetation fires taken together and just the more resource-

intensive wildfires (using either definition). It is the definition of a wildfire that

dictates the perception of the issue and its importance. It certainly dictates the

geography as Figures 4 and 5 show.

Fig 4 – Wildfire Density map for the Scottish Government wildfire manual definition

(IRS-based criteria only), 2009-2014

Density maps were produced using all incidents that met the Scottish Wildfire

definition (Fig 4) and the CFOA Category 5 Wildfire definition (Fig 5) to illustrate

the difference in extent and number for each definition. They can be compared

to the equivalent point density map for all vegetation fires regardless of size or

resources used (Fig 7).

The Scottish definition (Fig 4) shows wildfires concentrated on the edges of

rural uplands of the Pennines and Peak District and farmland in the west. It

also shows a high concentration in the deprived urban areas of east

Manchester, not usually associated with wildfires. The urban incidents are

linked to the number of vehicles and time spent at an incident, rather than the

size of the area burnt; this displays that resources and response to an incident

in urban areas is greater influence on the adoption of a definition, illustrating

that each region has a unique wildfire character and that a definition created for

a rural fire service region produces different results when applied to a more

urbanised region.

Fig 5 – Wildfire Density map using the CFOA Wildfire group category 5 (IRS-based

criteria only), 2009 -2014

The pattern of occurrence for the CFOA Category 5 definition indicates that very

high impact wildfires occur mainly in more rural areas associated with

continuous vegetation linked to agriculture or uplands associated with

agriculture and recreation. These incidents may be larger due to the time taken

to report and respond to these more remote incidents, or where there fewer

resources available.

Data on the over-the-border incidents attended by GMFRS (from the GMFRS

Information Management System) was added to the Scottish definition to produce

a map of the potential extent of wildfire risk beyond the GM Fire Authority boundary

(Fig 6). It shows that wildfire risk is crosses administrative borders and is

associated with urban populations visiting natural areas. The area around the

Lancashire border may provide as high a wildfire risk in the north of the Greater

Manchester Fire Authority as exists in the West Yorkshire and Derbyshire borders

in the east. GMFRS is therefore bounded on two sides by areas of high wildfire

risk. In fact in total there are over four times as many incidents over the border

compared to inside GM. of vegetation fires attended by GMFRS are over-the-

border incidents.

Fig 6 – Wildfire Density map for over-the-border Incidents and the Scottish

Government definition of wildfires (IRS-based criteria only), 2009-2014.

Type of land cover burned

A buffer zone of 500m was drawn around each wildfire point to allow for fire

spread. The land cover types in these combined zones are shown in Figure 8. It

provides an estimate of the type of vegetation burned by wildfires. If fire

perimeters were recorded, a more accurate breakdown of land cover and actual

burned area could be provided

For both definitions of wildfire, the largest three land cover types were improved

grassland, arable and horticulture, and rough grassland (Fig 8a & b). None of

these are ‘wildland’ vegetation, but they are all fine to moderate fuels with short

drying times and shorter spotting distances. The largest coarse fuel affected was

broadleaved woodland. The largest vegetation associated with ‘wildland’ areas

was acid grassland. These results reflect the land cover in the GMFRS region, with

no major discrepancy between proportion of land cover and fire occurrence.

Fig 7 Point Density Map for all vegetation fires within Greater Manchester 2009-

2014

The risk of wildfire ignition is linked to the size and shape of the land cover and

how close it is to settlements or roads. Areas of large improved grasslands in

the GMFRS area intermix with urban areas, or agricultural areas intersect with

urban fringes, underlying the importance of the rural-urban interface in wildfire

risk. Wildland areas have much lower levels of human activity, so the risk of

wildfire ignition is lower. However, habitat is an important factor in wildfire

ignition and with a many wildland areas having continuous vegetation areas

exceeding 1,000ha, the risk of escalation of an incident is increased in these

areas from even limited human interaction.

An interface zone between urban and vegetated areas can be identified for Greater

Manchester to show vegetated areas at greatest risk of wildfire ignition. This rural-

urban interface (RUI) broadly falls into two types: peri-urban where settlements

begin to mix with rural areas, and a more rural interface towards the edges of the

region where agriculture and continuous wildland vegetation mix with human

influences and activities; the extent of this interface from the details within figure 9b

and 9c is approximately 500m within the Greater Manchester area.

Fig 8 - Vegetation land cover within the wildfire buffer zone (a) Scottish Wildfire definition (b) CFOA Wildfire definition (c) proportion of Land Cover Map 2007

classes inside Greater Manchester Fire and Rescue Service area.

Fires decrease away from urban areas and roads

The number of wildfires decreases with distance from the urban boundary for

both definitions. Large urban areas make up 43% of the GM area, they are

made up of the urban and suburban areas within Land Cover Map 2007 (fig 9c);

they appear to have a significant influence, with 60% of incidents occuring

within 202m and 80% within 425m (Fig 9a). The large urban areas influence is

due to both their extent and density of population in close proximity to areas of

vegetation. Most incidents occurred within a kilometre, so within easy walking

distance of urban areas and suggest human-caused ignitions.

For the Scottish wildfire definition (Fig 9b), 70% of incidents occurred within 500

metres of the urban boundary, with only 16% occurring over a kilometre away.

The distances were greater for the CFOA Category 5 definition (Fig 9c), but

75% of incidents occurred within 600 metres, with all other incidents occurring

within 1300 metres. These distances display the influence that human activity

and influence has for the majority of wildfire incidents and displays the areas of

greatest risk are situated, which in turn can influence policy on the

prevention of incidents. A-roads also have a strong influence; 60% of

incidents occur within 459m. They extend human access, and thus

ignitions, to more remote and larger wildland vegetated areas, increasing

the number of potential accidental and deliberate ignitions in those areas.

Wildfire Risk

In order to produce a map of potential wildfire risk (Fig 10), the three criteria

which were found to be most significant in explaining wildfire occurrence were

combined on one scale and weighted in turn; they were distance from the urban

vegetation interface, distance from A-roads and population density. It was found

that the weighting of the urban vegetation interface displayed the most

widespread areas of higher risk; for the Scottish wildfire definition 72% of

incidents occurred in the high risk areas and 65% of incidents were captured for

the CFOA category 5 wildfire definition. The results overall show several levels

of wildfire ignition risk. The highest risk is on the margins of Greater

Manchester, especially the Pennines and Peak District and shows the influence

of the A-roads that intersect them and where fuel supply is greater. The medium

risk for the agricultural areas and the edgesof urban areas, where fuel and

human activity are more balanced. The low risk areas are inside the urban

areas where fuel is sparse.

(a)

(b)

)

(c)

)

Fig 9 – Decrease in numbers of wildfires with distance from the urban boundary

for (a) all outdoor fires (b) Scottish wildfire definition (c) CFOA wildfire definition.

Fig 10 -Multi-Criteria Evaluation of the risk of wildfire ignition with vegetation

interface weighted

Which wildfire defintion for Greater Manchester?

Figures 4 and 5 show that definitions of wildfire dictate the geography of ignition

risk. This study was based on two definitions (Table 1): the Scottish wildfire

manual, or Category 4 in the proposed categorisation and the CFOA Wildfire

Group definition; and CFOA’s proposed Category 5. Results based on the IRS

data for Greater Manchester display the differences between the definitions and

reinforce the idea that each region has a unique wildfire picture. The Scottish

manual definition identified more incidents but captured many urban vegetation

fires, which it was not designed to do, but did highlight wildfire risk in

surrounding rural areas. The CFOA Wildfire Group category 5 requires a

combination of two factors, so captures fewer incidents, but highlights the areas

at greatest risk from more resource-intensive fires. A combination of these two

definitions appears more appropriate for the GMFRS region, because both were

required to identify key locations of potential wildfire risk where GMFRS

resources are deployed, whether within Greater Manchester or over-the-border.

In practice Category 4 incidents require inclusion with regard to wildfires with

(b)

)

(a)

)

(c)

)

regard to size, however the number of vehicles and duration would be need to

be viewed on the effect one had on the other, namely if less than four

appliances led to a longer duration or a less than six hour duration was due to a

large number of appliances responding. The location of an incident and the

resources available are the determining factors as to whether an incident is or

becomes a wildfire as the weather and land cover.

Table 1 –Proposed categorisation of wildfire using IRS, as discussed at CFOA wildfire

Group, June 2014.

Recommendations

1. An agreed definition of wildfire should be adopted so that consistent data can

be collected. It would provide a better understanding of where the risk is

currently highest and how this may change over time. A rolling report could be

run to extract and monitor trends in wildfire incidents.

2. Prevention and preparedness activty should be targetted in the highest risk

areas (Fig 10). Prevention includes existing strategies such as public education

and work with Local Fire Operations Groups, especially with land owners and

land management agencies to manage fuel load and continuity in at-risk

vegetated areas. Preparedness includes siting of wildfire-specifc resources at

Fire Stations closest to seasonal wildfire hotspots, e.g. priority should be given

to Mossley and Littleborough in the east, Ramsbottom and Horwich in the north

and Irlam in the west, as they are closest to the areas of highest risk.

3. Standardise the geolocated point recorded in IRS to improve the accuracy of

GIS analysis. It should be within the fire footprint and preferably at the

estimated point of ignition.

4. A process should ideally be put in place to collect additional data for the

resource-intensive category 5 fires, most critically:

Fire perimeter to provide, for example, a more accurate estimate of burned

area, types of vegetation burned, and how often fires recurr at a given location

(the current geolocation only gives information only at one point)

Fire investigation to determine cause and ignition point

Weather data from the nearest weather station during and in the run up to the

incident to help understand fire behaviour and implications for fire-fighter

safety

The 0-5 Met OfficeFire Severity Index, or sub-indices such as Fine Fuel

Moisture Code or Initial Spread Index, which are available from the European

Firest Fire Information System (EFFIS,

http://forest.jrc.ec.europa.eu/effis/applications/current-situation/)

This extra data would help build a better understanding of the types of wildfire that

occur in Great Manchester, and the underlying causal factors. It can be used to

inform future prevention and safety strategies.

5. Cross border data sharing on wildfire conditions and fuel availability is crucial.

GMFRS attends over-the-border incidents in large continuous vegetated areas which

cross Fire Authorities, especially in Lancashire, West Yorkshire and Derbyshire, all of

which are areas of high wildfire risk. It is recommended that over-the-border

incidents should be recorded by GMFRS using the IRS data fields, rather than the

present system, irrespective of another Fire Service ‘owning’ the incident and

reporting formallyit to DCLG. It would provide a true picture of the wildfires actually

attended by GMFRS. It would provide a clearer evidence base of wildfire risk for the

wider region to inform resourcing and strategies to manage wildfire risk.

References

Donlan, R. and McMorrow, J. (2015) Vegetation Fires in Greater Manchester: a GIS

study using Incident Recording System data, 2009 -2014 -- Report for Greater

Manchester Fire and Rescue Service. Geography, School of Environment, Education

and Development, University of Manchester.

Available from: http://www.kfwf.org.uk/resources/briefingpapers/ Contacts;

richarddonlan@hotmail.com Julia.mcmorrow@manchester.ac.uk