1. Introduction
Global monitoring of wildfire emissions is supported by the network of geo-stationary weather satellites [
1]. Finer resolution polar orbiting systems provide further support by correcting for observational biases [
2,
3,
4], and are more commonplace in direct wildfire management applications. Even still, the use of satellite data in real-time emergency management decision-making remains rare, partially due to the latency of satellite wildfire data [
5,
6,
7]. Furthermore, at high latitudes the geostationary network is challenged by rapidly degrading spatial resolution and atmospheric transmission [
1,
4], leaving wildfire-prone northern boreal regions with limited wildfire monitoring capacity from polar-orbiting systems and a heavy reliance on monitoring from aircraft.
Efforts to enhance the uptake of satellite data in wildfire management have been pursued through satellite design (e.g., [
8,
9,
10]) and purpose-built information systems (e.g., Global Wildfire Information System [GWIS] [
11]); additionally, various commercial proposals have been proposed. However, to date no system has successfully delivered end-to-end operational support to address the specific needs of fire managers. In part this is a result of the broad range of wildfire management practices globally and the resulting variation in specific regional requirements for wildfire monitoring. Inability to accurately define the requirements of end-users presents a broad reaching barrier to operational implementation.
Responses to wildfires vary across Canada [
12] and range from a “Full Response” (immediate, aggressive initial/sustained attack), to “Monitored Response” (observation and periodic reassessment; [
13]), guided either by zonation, or wildfire specific conditions termed “appropriate response” [
14]. In situations with increased wildfire activity, the suppression capacity can be rapidly overwhelmed, resulting in escaped wildfires that may burn very large areas [
15,
16]. These larger wildfires represent only 3% of the number of wildfires in Canada, yet they account for 97% of the area burned [
12,
17], and require substantially more resources to manage [
18,
19].
In many higher risk locations, Canadian wildfire management agencies generally rely on the Initial Attack model where wildfires are detected early and suppressed small. The early detection and suppression of wildfires is critical to successful wildfire control resulting in fewer escaped wildfires, therefore reducing impacts and response costs [
20]. Wildfire remote sensing has been recognized for its capacity to detect wildfires (e.g., [
21]). However, there is a significant gap between what is required for “early” detection (e.g., identifying small sub-canopy wildfires) for wildfires that require suppression and what can be accomplished reliably with satellite remote sensing [
22]. Beyond detection, during periods of escalated wildfire activity it is useful to have current intelligence about all ongoing wildfires. Reliable wildfire intelligence is critical for situational awareness and informed decision-making including prioritization and strategic and tactical wildfire response.
Wildfires that are threatening communities and critical infrastructure are prioritized for suppression action over remote wildfires where there is more opportunity for the natural ecological role of wildfire on the landscape [
23]. Therefore, these wildfires more frequently grow larger and are generally managed through modified tactics (e.g., continuous mapping and monitoring; [
14]). Although satellites are not typically helpful for early detection [
22], there is an emerging requirement for large wildfire and regional intelligence gathering to maintain situational awareness during periods of escalated wildfire activity. This intelligence requirement can be met through the proper application of satellite technology, and is broadly described as wildfire “monitoring” here.
Wildfire monitoring is also an essential component of Canada’s efforts to track and predict smoke dispersion from active wildfires. In recent years, wildfire smoke has been the dominant cause of poor air quality for large portions of Canada [
24]. The impacts of smoke on communities can necessitate an evacuation, even without a direct threat from wildfire [
25,
26,
27]. This led to the development of methods to derive emissions from satellite-detected wildfires, and to use the emission estimates in smoke dispersion [
28,
29] and air quality [
30] forecast systems. These automated systems require the provision of timely and reliable wildfire activity data, with an emphasis on wildfire events that produce large long-range transportation of smoke, or wildfire events near communities. With additional development, these systems can evolve to incorporate Fire Radiative Power (FRP; MW) measurements as an additional source for the estimation of wildfire emissions (e.g., [
31,
32]), as is becoming increasingly common throughout the world (e.g., Global Fire Assimilation System (GFAS); [
33].
Under climate change a substantial increase in the frequency and intensity of wildfires is expected [
34,
35]. In particular, northern regions such as Canada are expected to see an increase of wildfire activity related to increases in conditions conducive to extreme wildfire weather [
19,
36,
37,
38,
39,
40,
41]. Consequently, frequency of extreme burning days where wildfires are able to escape is also projected to increase [
41,
42].
In 2019 the Canadian Space Agency (CSA) initiated the development of a dedicated wildfire monitoring satellite “WildFireSat” mission [
43]. WildFireSat (WFS) intends to leverage uncooled microbolometer technology developed by the CSA and Institut National d’Optique (INO). An earlier version of this technology called the New InfraRed Sensor Technology (NIRST) was the first mid-wave infrared (IR) microbolometer used in space-based wildfire remote sensing on the 2011 Aquarius SAC-D mission [
44]. Following the NIRST experiment, the same technology was used in a feasibility study (referred to as “Phase 0”) to demonstrate the technical feasibility of a cost-effective, dedicated Canadian Wildland Fire Monitoring System (CWFMS) [
45]. Since then, the detector technology has continued to evolve (e.g., [
46]), while new Low Earth Orbit (LEO) wildfire products (e.g., from the Visible Infrared Imaging Radiometer Suite (VIIRS) and Sea and Land Surface Temperature Radiometer (SLSTR); [
47,
48]) have filled some temporal coverage gaps (
Figure 1), which will improve the feasibility of a targeted wildfire monitoring mission.
Phase-A of the WFS mission is driven by the Mission Requirements [
49] which extend from the User Requirements. However, the User Requirements defined in CWFMS [
50] required substantial revisions to accommodate the new context of this mission. The aim of this study is to trace the process used to update and re-scope the CWFMS User Requirements for WFS through consideration of emerging science and ongoing end-user consultation (e.g., [
51]). This study presents the WFS User Requirements and provides cross reference to their heritage in CWFMS where applicable. We trace the two primary phases of this process: (1) Canadian wildfire management needs are assessed through direct engagement of wildfire management end-users, leading to a set of key Fire Management Functionalities (FMFs); (2) User Requirements for the WFS mission are refined through the integration of the wildfire management needs with the best available scientific techniques. The result of this study is the definition of User Requirements for the first dedicated operational wildfire monitoring satellite, forming the foundation for later stages of mission development.
2. Wildfire Management Needs Assessment
Wildfire management agencies generally employ a risk-based approach where the potential impact(s), likelihood, and resulting expected loss or benefit are assessed at the appropriate scale according to the complexity of the wildfire situation [
26,
52,
53]. Decisions often involve multiple decision-makers and stakeholders with varying perspectives concerning risk [
23,
26]. Decisions are not static and are frequently updated through an iterative process of determining and taking actions, monitoring outcomes, and revising actions until the situation is resolved [
53]. When assessing progress in the decision-making cycle, decision-makers require different types of intelligence.
We define wildfire intelligence as information which is collected to support wildfire management activities. This may include current or forecasted information such as: Wildfire behavior, location, size, shape, spatial context (e.g., fuels, topography, proximity to areas of concern, etc.), firefighting resource allocation and use, and wildfire effects and impacts (e.g., social, economic). The type, precision, accuracy, and timeliness of intelligence required varies depending on whether tactical or strategic planning is being conducted.
2.1. Wildfire Management Engagement
In the first step of defining the Mission Requirements, Canadian wildfire managers were surveyed to better understand the relative importance of the various wildfire monitoring products and the constraints for their relevance as a source of intelligence in both tactical (e.g., same-day/near-term operations) and strategic (e.g., longer-term preparedness, large wildfire planning) decision-making. Respondents were posed a series of questions regarding potential Earth Observation (EO) data products and asked to consider each in the context of both tactical and strategic decision-making. They were asked to provide the optimal and maximal data latencies (i.e., the time lag from collection to receipt of data when it has most value and the point at which it no longer has value), as well a rating of the importance of each product using a Likert scale (1–5; low–high); respondents were also invited to provide comments to aid in the interpretation of their responses.
Survey Results
In total, 55 senior staff from 7 of 13 wildfire management agencies in Canada responded to the survey (
Table 1). Most responses were completed individually while a few were coordinated efforts by groups. This sample size is consistent with comparable engagement efforts in similar communities (e.g., [
54]). The format of some responses required the data to be reformatted prior to analysis. Group responses were weighted according to the number of people contributing to ensure proportional representation.
Products identified as the most important in both tactical and strategic decision-making were active-fire products; these products are typically derived from thermal observations of active wildfire events, and contribute to assessments of location, spread rate, and intensity (
Table 2). Whereas post-fire mapping products (e.g., burned area and severity) were primarily valuable at the strategic level (as well as for non-response management activities, e.g., forest inventory). Tactical intelligence was generally required “as quickly as possible” for all products, with the median of responses indicating 30 min or less (
Table 2). For strategic uses the same general preference for active-fire intelligence is found (
Table 2); however, slightly longer latencies are acceptable. Data latencies of up to 2 h were indicated as the thresholds for tactical and strategic decision-making. In many cases the information continued to have some value to managers for several hours up to 24 h, but not necessarily for tactical or strategic decision-making. Notably, this consultation process also revealed that the timing of data delivery during daily operations was also a key factor in data utility, due to the cyclical timing of daily decision-making.
2.2. Summary of Wildfire Manager Needs
In order to maximize the value of a satellite system for wildfire management, certain features were highlighted through additional comments provided during the end-user consultation. These features included: Fast and consistent data delivery, mapping of active and inactive wildfire areas, smoke and air quality information, wildfire behavior, and threat estimates, as well as detection in remote regions.
2.2.1. Fast and Consistent Data Delivery
Daily wildfire management activities follow planning cycles which depend on the scale of management occurring. For example, an incident command team responsible for planning and carrying out wildfire operations on a large wildfire may have different needs for the frequency and timeliness of information than those planning strategic response at a regional, provincial, or national scale. Generally speaking, in order for intelligence to be incorporated into daily planning activities data must reflect the current situation (i.e., low/short latency), but it is also important to receive the information at a consistent time of day to facilitate routine integration.
2.2.2. Mapping of Active and Inactive Wildfire Areas
Although there is a definite interest in the actively spreading portion, intelligence is required for the entire wildfire. The full perimeter of the burned area as well as the active and previously burned area are valuable in wildfire operations. Previously burned areas may still be smoldering and require prolonged suppression, while unburned “islands” in these areas pose a threat for re-burning. Managers also indicated that they were satisfied with the 375 m spatial resolution of the VIIRS I-band wildfire products [
47,
55], for general applications (though for high complexity incidents fine resolution airborne mapping may also be necessary).
2.2.3. Wildfire Behavior and Threat Estimates
The proximity and threat to interface zones was identified as critical intelligence in the survey (
Table 2). Proximity to these zones is achievable through accurate detection and mapping in conjunction with national interface maps [
54].
Wildfire behavior observations were considered to have both tactical and strategic value, particularly in terms of estimating the potential threat of a wildfire. Information relating to the rate and direction of wildfire spread (ROS (m s
−1) and DIR (deg); [
56,
57]) as well as the Fire Intensity (FI, (kW m
−1)), are essential to characterizing the behavior of actively spreading wildfires. Johnston et al. (2017) [
58] demonstrated that FI can be estimated directly from IR measurements of FRP. Wildfire behavior is of particular interest during the late afternoon “peak burn” period (
Figure 1). This information should ideally be paired with the spatial context (e.g., adjacent fuels, topography, and proximity to areas of concern; [
59]) to provide estimates of proximity and threat to these areas.
2.2.4. Detection in Remote Regions
Wildfire management practices in Canada vary dramatically across the landscape, and generally in relation to population distributions (e.g., [
9,
10]). In vast remote areas, wildfire managers do not typically conduct dedicated detection activities due to the decreased likelihood for negative impacts from wildfire, and the higher cost and operational complexity of these patrols (e.g., [
60]). Space-based EO is particularly suitable for gathering intelligence in these situations [
22]. The value of EO-derived detection of wildfires identified in the survey (
Table 2) was highlighted as particularly valuable in these regions in the comments.
2.2.5. Smoke and Air Quality Information
Although smoke management tools were not identified as critical to tactical decision-making in the survey responses, this information is critical for other emergency management operations. Non-fire management users require smoke-related intelligence for critical operations such as evacuation planning [
61,
62,
63], public health forecasting [
64], and aviation visibility [
65]. Smoke forecasting using tools such as FireWork [
30,
66] and BlueSky [
29,
67] are dependent on wildfire size and location information. Other smoke monitoring applications (e.g., Global Fire Assimilation System (GFAS); [
33]) require FRP [
31,
32] as a primary input.
The delivery of operational smoke and air quality forecast is a highly automated process. As of 2020, Environment and Climate Change Canada will launch a new air quality forecast twice a day (initiated at 00 and 12 UTC), and smoke forecast every 6 h (00, 06, 12, 18 UTC). Each execution is updated with the latest wildfire information available. The scheduling of forecast executions is tied to the availability of new weather and wildfire emission data, and computing resources. Due to this scheduling, the data latency requirements are less stringent than for wildfire management applications, but the requirement to focus on the peak burn overpass period still persists. Additionally, smoke and air quality applications emphasize a strong interest in smaller wildfire detection, an interest in thermodynamic parameters controlling plume rise and height (e.g., [
68,
69]), and a larger coverage (e.g., North America).
2.3. Fire Management Functionalities:
The needs identified above were translated into a set of key Fire Management Functionalities (FMF) necessary to define the User Requirements:
- (1)
The Area of Interest (AoI) is defined as the whole vegetated Canadian landmass (
Figure 2);
- (2)
Daily (or better) coverage of the AoI at a specific and consistent time of day including peak burn (1600–2000 local time), with data delivery (to end-users) before the start of the operational response period (~0700 local) for overnight observations and before the end of day planning period (~1900 local) for peak burn observations;
- (3)
Detection and mapping of wildfires and their plumes, specifically:
- i.
The ability to detect wildfires with comparable or improved sensitivity to existing satellite systems, and to serve as an early-detection system for remote access wildfires;
- ii.
There must be sufficient spatial resolution and geolocation accuracy for locating and mapping wildfires in relation to their previous position and other landscape features;
- (4)
Estimation of wildfire behavior, specifically:
- i.
The ability to collect FRP measurements;
- ii.
The ability to characterize sub-pixel wildfire features (e.g., temperature and area);
- (5)
Compatibility with other available EO data sources and formats;
- (6)
Near-real-time data, with tactical products to be delivered within 30 min, and a 2-h latency for all end-user products as threshold for utility.
The interconnectivity of the FMFs with the requirements laid out in the User Requirements Document [
49] and the Mission Requirements Document [
50] is summarized in
Table 3. The development and rationale for these requirements is described in the following sections.
4. Discussion
The aim of this study was to define requirements for a wildfire monitoring satellite system explicitly to support wildfire management. To do so we surveyed experienced wildfire managers with various specialties in order to better capture their needs in the form of the six FMFs. These functionalities were then used to guide the definition of the WFS User Requirements based on current scientific and technical capabilities.
This study does not describe all factors or analyses considered throughout the process of defining the User Requirements for WFS. Some decisions were made in defining the scope and thus impacted the entire process. For example, the assumption of a LEO satellite mission was taken as the baseline. However, it could be argued that Highly Elliptical Orbiting (HEO) satellite constellations are better suited for the Canadian AoI (e.g., [
89,
90]), although they are far more costly.
Further, many active- and post-fire products exploit the availability of Short-Wave infrared (SWIR) spectral bands for overnight active wildfire detection and burn severity mapping. Although this spectral band has been considered it was not deemed essential to deliver the FMFs so it was not included as an essential requirement for the mission, though it does remain a goal.
The greatest challenge throughout this process has been defining threshold requirements. It is a simple task to identify the ideal system, payload, and detector. However, determining the limit beyond which a system will no longer be functional is challenging. There is little precedence for marginal systems which can be drawn from. Definition of end-user “usefulness” happens gradually and there is rarely a single threshold of usefulness, particularly when introducing new capabilities. Ultimately, numerous trade-off analyses must be carried out by the Space Team and the User and Science Team where prioritization of the various goal and threshold requirements is necessary for mission development.
The inclusion of trade-off criteria that underpin these requirements is essential as budgetary, technical, and practical limitations will inevitably limit the ability to achieve all of the goal requirements. For example, finer spatial resolution will either reduce the swath and therefore coverage or increase the number of pixels and data volume, challenging the latency requirements. Balancing the consequences of competing requirements is difficult to prescribe a priori. Generally, the trade-off criteria hold coverage of the AoI and IR payload performance as the highest priorities, although overall ability to meet the FMFs is the guiding need. As such, the requirements presented here are a documentation of process at this point in time, but are expected to evolve throughout the mission. The intent is for these requirements to be interpreted in close coordination with the User and Science Team throughout the full mission development.
5. Conclusions
In this study we provided an overview of the approach taken to understand Canadian wildfire management EO needs and transcribe User Requirements which can be used to develop a purpose-built wildfire monitoring satellite to meet their needs. The User Requirements presented here originate in the requirements for CWFMS [
49], which were refined through consultation of scientific (e.g., [
51]) and wildfire management users to produce the User Requirements for the WFS mission.
The translation of the Wildfire Management Needs into User Requirements for WFS is a foundational step in Phase-A of the mission (
Figure 4). Through this process we developed qualitative FMFs based on operational wildfire management needs. Considering technical capabilities and limitations allows User Requirements to be defined for a non-specific space system to address the FMFs. When financial and scope considerations are applied to the User Requirements, the WFS Mission Requirements [
50] can be specified. Ultimately, Phase-A of WFS culminates by extending the Mission Requirements into detailed technical specifications for the satellite in the System Requirements. As
Figure 4 illustrates, although each stage of Phase-A becomes more technically specific, all of the System Requirements are traceable to their origins in wildfire management needs.
WildFireSat aims to deliver a purpose-built operational wildfire monitoring satellite to support wildfire managers as the primary users. To that end, despite the technical and scientific challenges of the mission, the key to operational success remains in the hands of the wildfire management community. In order to achieve meaningful impact in wildfire management operations, the end-user engagement described in this study must continue for the duration of the mission to ensure that wildfire management needs continue to be heard and that wildfire managers develop a sense of ownership in the mission.