The Butry-Thomas Analysis Discussion

Share on:

Wildfire Report Series: MODIS-Detected Fires in NFIRS

Discussion of approach

It is possible that some of the methods and assumptions used in this analysis are problematic.

Butry and Thomas make an implicit assumption that all fires detected by MODIS are wildfires, but MODIS can on rare occasions detect fires as small as 50 square meters (4-meter radius) and routinely detects fires 1000 square meters in size (18-meter radius). The criteria Butry and Thomas used to identify wildfires in the NFIRS data potentially omits many other kinds of incidents that could match the MODIS observation. Moreover, the criteria used omits other factors that could identify additional wildland fires. Notably, the criteria do not include other outdoor incident types:

140: Natural vegetation fire, other.
142: Brush or brush-and-grass mixture fire.
143: Grass fire.
170: Cultivated vegetation, crop fire, other.
171: Cultivated grain or crop fire.
172: Cultivated orchard or vineyard fire.
173: Cultivated trees or nursery stock fire.
561: Unauthorized burning.
631: Authorized controlled burning.
632: Prescribed fire.

The criteria do not consider property use:

655: Crops or orchard.
659: Livestock production.
669: Forest, timberland, woodland.
931: Open land or field.

The criteria omit certain areas of origin:

94: Open area, outside; included are farmland, field.

Last, the criteria do not include other possible suppression factors:

434: Poor or no access for fire department apparatus.
760: Unusual vegetation fuel loading.

Incidents defined under these categories would not be included in the analysis unless they also meet the criteria specified by Butry and Thomas.

The criteria Butry and Thomas use to identify potential wildfires rely upon values in the Basic, Fire and Wildland Fire modules. Of these, only the first Action Taken value is required and is populated with useable data nearly all of the time (Table 1). Area of Fire Origin, in the Fire Module, contains usable data in 81% of the reports where a Fire Module is completed (Table 2). Suppression Factors, also in the Fire Module, are optional fields and are sparsely populated (Table 3). Finally, Butry and Thomas use the presence of the Wildland Fire Module to infer a wildfire, but the Wildland Fire Module is only used for 23% of the incidents where it is allowed (Table 4).

Action Taken is recorded in the NFIRS Basic Incident Module. In this module, up to 3 values may be entered. The first one is required for all values of aid, and the others are optional.

Table 1: Usability of Primary Actions Taken (Basic Incident)

	Action Taken 1			Action Taken 2			Action Taken 3
	Primary¹	Aid given	Total	Primary	Aid given	Total	Primary	Aid given	Total
Total basic incidents	1,073,284	227,128	1,300,412	1,073,284	227,128	1,300,412	1,073,284	227,128	1,300,412
Invalid values²	11,975	7,538	19,513	2,337	854	3,191	1,080	154	1,234
Null values	1	0	1	730,701	157,477	888,178	944,352	199,607	1,143,959
Usable values	1,061,308	219,590	1,280,898	340,246	68,797	409,043	127,852	27,367	155,219
Percent of incidents	99	97	98	32	30	31	12	12	12

¹ Primary refers to incidents where AID = 1 (mutual aid received), 2 (automatic aid received), 5 (other aid given), or N (no aid). Aid Given refers to incidents where AID = 3 (mutual aid given) or 4 (automatic aid given).
² Invalid Values are entered values that are not listed in the NFIRS Complete Reference Guide, including "-", "00", "0", "6", "18", "25", "59", "68", "95" and "N."

Area of Fire Origin and Suppression Factors are recorded in the NFIRS Fire Module, which is only required for certain incident types where the reporting department (primary) is not giving aid. Area of Fire Origin is required. Suppression Factors are optional, and up to 3 values may be recorded.

Table 2: Usability of Area of Fire Origin (Fire Module)

	Area of origin
	Primary¹	Aid given	Total
Total Fire Module reports	598,828	5,803	604,631
Null values	448	274	722
Undetermined (UU)	94,464	2,020	96,484
Other (O)	18,751	192	18,943
Usable values	485,165	3,317	488,482
Percent of incidents with Fire Module	81	57	81

¹ Primary refers to incidents where AID = 1, 2, 5, or N. Aid Given refers to incidents where AID = 3 or 4.

Table 3: Usability of Suppression Factors (Fire Module)

	Suppression Factor 1			Suppression Factor 2			Suppression Factor 3
	Primary¹	Aid given	Total	Primary	Aid given	Total	Primary	Aid given	Total
Total Fire Module reports	598,828	5,803	604,631	598,828	5,803	604,631	598,828	5,803	604,631
Null values	426,033	3,900	429,933	595,107	5,709	600,816	597,467	5,765	603,232
Undetermined (UUU)	1	0	1	2	0	2	0	0	0
Other (0)	1,567	28	1,595	55	2	57	31	0	31
None (NNN)	154,590	1,536	156,126	0	0	0	0	0	0
Usable values	16,637	339	16,976	3,664	92	3,756	1,330	38	1,368
Percent of incidents with Fire Module	4	9	4	1	2	1	0	1	0

¹ Primary refers to incidents where AID = 1, 2, 5, or N. Aid Given refers to incidents where AID = 3 or 4.

The Fire Module is required for most incident types but may be substituted with the Wildland Fire Module for certain kinds of fires. The Fire Module is optional for certain incident types in the 100 series (fire incidents) but is not available for other kinds of incidents. Neither module is required if the department is assisting another department (where AID = 3 or 4).

Table 4: Presence of Fire and Wildland Fire Module by Module Requirement¹

	Primary²			Aid given			Total
	Incidents	% using Fire	% using Wildland	Incidents	% using Fire	% using Wildland	Incidents	% using Fire	% using Wildland
Fire or Wildland required	216,979	73	27	45,852	3	3	262,831	61	23
Fire required	365,130	100		156,146	3		521,276	71
Fire optional	328,130	23		21,459	1		349,589	21
Wildland optional	163,045		1	3,671		1	166,716		1
Total	1,073,284	56	6	227,128	3	1	1,300,412	46	5

¹ Either the Fire Module or the Wildland Fire Module is required for incident types 140-143, 160, and 170-173. The Fire Module is required for incident types 100, 111-112, 120-123, 130-138, and 161-164. The Fire Module is optional for all other fire incident types, and the Wildland Fire Module is optional for incident types 561, 631, and 632. Both modules are optional for Aid Given.
² Primary refers to incidents where AID = 1, 2, 5, or N. Aid Given refers to incidents where AID = 3 or 4.

Butry and Thomas did not have geocoded NFIRS incident data and matched selected NFIRS and MODIS incidents based on county and date. In practice, this means that any qualifying NFIRS incident in the county could be matched to any MODIS observation in the local direct protection area, regardless of the actual distance between the points. Also, matching incidents solely by date does not allow the researchers to identify possible matches where the NFIRS and MODIS reports occurred on adjacent dates.