Downloaded 10/07/21 01:35 PM UTC 2236 MONTHLY WEATHER REVIEW VOLUME 145

JUNE 2017 P E T E R S O N E T A L . 2235

A Conceptual Model for Development of Intense Pyrocumulonimbus in Western North America

DAVID A. PETERSON,EDWARD J. HYER, AND JAMES R. CAMPBELL Naval Research Laboratory, Monterey, California

JEREMY E. SOLBRIG Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, Colorado

MICHAEL D. FROMM Naval Research Laboratory, Washington, D.C.

(Manuscript received 17 June 2016, in ﬁnal form 19 November 2016)

ABSTRACT

The first observationally based conceptual model for intense pyrocumulonimbus (pyroCb) development is described by applying reanalyzed meteorological model output to an inventory of 26 intense pyroCb events from June to August 2013 and a control inventory of intense fire activity without pyroCb. Results are based on 88 intense wildfires observed within the western United States and Canada. While surface-based fire weather indices are a useful indicator of intense fire activity, they are not a skillful predictor of intense pyroCb. Development occurs when a layer of increased moisture content and instability is advected over a dry, deep, and unstable mixed layer, typically along the leading edge of an approaching disturbance or under the influence of a monsoonal anticyclone. Upper-tropospheric dynamics are conducive to rising motion and vertical convective development. Mid- and upper-tropospheric conditions therefore resemble those that produce traditional dry thunderstorms. The specific quantity of midlevel moisture and instability required is shown to be strongly dependent on the surface elevation of the contributing fire. Increased thermal buoyancy from large and intense wildfires can serve as a potential trigger, implying that pyroCb occasionally develop in the absence of traditional meteorological triggering mechanisms. This conceptual model suggests that meteorological conditions favorable for pyroCb are observed regularly in western North America. PyroCb and ensuing stratospheric smoke injection are therefore likely to be significant and endemic features of summer climate. Results from this study provide a major step toward improved detection, monitoring, and prediction of pyroCb, which will ultimately enable improved understanding of the role of this phenomenon in the climate system.

1. Introduction troposphere, drastically impacting the extent of down- wind transport. Intense pyroCb activity can also inject a Large smoke plumes from active fires often become significant quantity of aerosol mass into the upper tro- capped by cumulus clouds (pyroCu), which enhance the posphere and lower stratosphere (UTLS), occasionally vertical lofting of smoke through latent heat release. more than 7–10 km above the tropopause (Fromm et al. Under certain conditions, this moist ‘‘pyroconvection’’ 2005, 2008a,b). At this altitude, aerosol particles can mechanism allows pyroCu to develop into a larger fire- persist for long durations (Robock 2000), allowing for triggered thunderstorm, known as pyrocumulonimbus gradual spread over hemispheric scales. PyroCb are (or pyroCb; American Meteorological Society 2016). therefore likely to be a highly relevant and significant The deep convective column provides an efficient process governing the state of stratospheric aerosols and method for lofting smoke particles well into the upper affecting global climate. The current understanding of UTLS smoke injection Corresponding author: David Peterson, david.peterson@nrlmry. from pyroCb is based on individual fires observed in the navy.mil mid- and upper-latitude forests of North America,

DOI: 10.1175/MWR-D-16-0232.1 Ó 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2236 MONTHLY WEATHER REVIEW VOLUME 145

Australia, and Asia (e.g., Fromm et al. 2006, 2010, 2012). significant entrainment of ambient air in the midtropo- PyroCb have been identified as the source of several sphere (Trentmann et al. 2006). This suggests that UTLS aerosol layers previously presumed to be of vol- midtropospheric water vapor can be a primary driver of canic origin (Fromm et al. 2010). Remote sensing–based pyroCb development. studies reveal that pyroCb anvils have distinct, smoke- Through a detailed analysis of the 2013 Rim Fire in induced microphysical characteristics and longer life- California, Peterson et al. (2015) hypothesize that the times compared with traditional convection (e.g., meteorological environment commonly associated with Rosenfeld et al. 2007; Lindsey and Fromm 2008; Reutter high-based convection (dry thunderstorms) in the et al. 2014). The mechanisms driving pyroCb develop- southwestern United States is likely a general precursor ment and ensuing UTLS smoke injection can be differ- for pyroCb development. Fromm et al. (2006) also de- ent from those linked to rapid fire spread (Peterson et al. scribe pyroCb activity in the vicinity of traditional high- 2015). However, the specific meteorological conditions based convection in southeastern Australia. High-based driving pyroCb development remain highly uncertain, convection occurs when a layer with increased moisture especially over large spatial and temporal scales. and instability is advected over a dry, deep, and unstable All observed pyroCb to date are associated with large mixed layer, typically along the leading edge of an ap- and intense fires. It is therefore not surprising that a hot proaching disturbance (e.g., Rorig and Ferguson 1999; surface temperature, strong surface winds, low relative Nauslar et al. 2013). The upper-tropospheric dynamics humidity, and deep mixed layer are often associated and synoptic pattern must also be conducive for rising with pyroCb activity (Fromm et al. 2006, 2012; Cruz motion and vertical development of convection et al. 2012). A variety of fire weather indices, such as the (Wallmann et al. 2010; Nauslar et al. 2013). Mid- and Haines index (Haines 1988) or continuous Haines index upper-tropospheric conditions are therefore paramount (Mills and McCaw 2010) have been employed to char- for the development and maintenance of high-based acterize the pyroCb environment (Fromm et al. 2012). convection, presumably including pyroCb activity. In However, Peterson et al. (2015) show that these surface contrast to severe thunderstorm outbreaks, the triggering and near-surface-based variables and indices have lim- mechanism for high-based convection usually stems from ited utility for forecasting or diagnosing extreme wildfire orographic heating/lifting or a weak disturbance in the development, including formation of pyroCb. There are midtroposphere (Wallmann et al. 2010). Increased ther- also conflicting assertions on the importance of atmo- mal buoyancy from large and intense wildfires (e.g., Val spheric instability and the parameter or index used to Martin et al. 2012; Peterson et al. 2014) may also serve as a quantify its magnitude (Trentmann et al. 2006; Fromm potential trigger for pyroCb (e.g., Peterson et al. 2015; et al. 2010, 2012). A potential link to surface frontal Lareau and Clements 2016); especially during specific boundaries (Fromm et al. 2005; Trentmann et al. 2006; forms of surface fire spread (McRae et al. 2015). Cruz et al. 2012) and/or upper-level disturbances Many techniques have been employed to identify in- (Johnson et al. 2014; Peterson et al. 2015) has been dividual pyroCb events of varying magnitude, including identified, but the specific role these features have in visual observations, satellite remote sensing of cloud-top pyroCb development is currently unclear. brightness temperature near fires, and reverse trajecto- All convective cloud development requires a moisture ries from high-altitude smoke clouds observed by lidar source, usually from within the lower or midtropo- during the days following a potential pyroCb (e.g., sphere. However, the unique link between pyroCb and Rosenfeld et al. 2007; Fromm et al. 2005, 2008a,b, 2010). intense fire activity suggests that water vapor released Satellite observations reveal that North American as a by-product of combustion may also play a role. The pyroCb typically initiate during midafternoon and specific contribution of each potential moisture source terminate a few hours after sunset (Fromm et al. 2010). during pyroCb development remains a highly contro- Many pyroCb are characterized by relatively small anvil versial topic. Several studies suggest that latent heat clouds, persisting for less than one hour. In extreme production from combustion is a key driver of plume cases, pyroCb develop expansive anvils that persist for dynamics and subsequent pyroCb activity, with a negli- several hours, occasionally comprising multiple con- gible contribution from the ambient atmosphere (e.g., vective pulses (Rosenfeld et al. 2007). However, despite Potter 2005; Cunningham and Reeder 2009). In contrast, this growing body of literature, pyroCb detection at other studies show plume dynamics are primarily driven large regional or global scales has been unsystematic by sensible/radiant heat release, with the contribution of and limited, suggesting that pyroCb activity is likely water vapor from combustion decreasing rapidly with undersampled. height (Trentmann et al. 2006; Luderer et al. 2009). The U.S. Naval Research Laboratory (NRL) has Model simulations of pyroCb plumes show potential for developed a near-real-time pyroCb detection algorithm

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2237

FIG. 1. Study region map primarily based on fire-prone regions within the effective field-of- view of GOES-W. Red symbols represent 88 intense fires inventoried for pyroCb activity during June–August 2013 (Peterson et al. 2017), with symbol type indicating fire radiative power (FRP) characteristics. Blue triangles indicate the 11 fires that produced at least one intense pyroCb (IPCB) event. Thick orange curves indicate the viewing zenith angle (VZA) of GOES-W. Thin contours indicate surface elevation, with green shading representing regions of forest or mixed forest and chaparral vegetation. Fires that produced specific IPCB events highlighted in this study are denoted in light blue. based on geostationary satellite data (Peterson et al. stratospheric smoke injection on the climate system, as 2017). The algorithm takes advantage of the unique well as development of future predictive capabilities. microphysics of pyroCb, facilitating development of the first systematic event inventory across western North 2. Study region, data, and methods America during the primary fire season (June–August) of 2013. This companion study employs the combination This study incorporates an inventory of 26 intense of the pyroCb inventory and reanalysis data to examine pyroCb events developed by Peterson et al. (2017), the meteorological conditions during large and intense which is based on the lifetime of 88 intense wildfires pyroCb events, with the goal of building the first con- observed across western North America during June– ceptual model for their development. Periods of intense August 2013 (Fig. 1). The study domain is primarily fire activity (90th percentile of hourly fire radiative based on fire-prone regions within the effective field-of- power) that fail to produce pyroCb activity are also view for the Geostationary Operational Environmental examined. Results highlight the role of atmospheric Satellite observing western North America (GOES-W; dynamics, midlevel moisture content, and atmospheric currently GOES-15), with an equatorial subsatellite stability profiles in pyroCb development. Limitations of point of 1358W. This includes the western continental traditional fire weather indices and forecasting meth- United States and northern Mexico (MCONUS), as well odologies are also explored. This information is essen- as the remote boreal forest of western Canada (Fig. 1). tial for understanding the impact of pyroCb and ensuing All fires in MCONUS were located in regions with

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2238 MONTHLY WEATHER REVIEW VOLUME 145

FIG. 2. Example GOES-W IPCB detection imagery for (a) the Papoose fire in Colorado and (b) Pony/Elk fire in Idaho. Red and orange shading indicates BTD4–11 for pixels with BT11 below the IPCB threshold (BT11 ,2358C). White shading indicates BT11 for nonpyroCb clouds. The few pixels shaded in green represent deep convection (BT11 ,2208C) that passes the pyroCb microphysics and opacity tests, but remains warmer than the IPCB BT11 threshold. complex topography [mean elevation of 1800 m above and therefore have a high likelihood of injecting smoke sea level (MSL)], but the majority of the Canadian fires particles into the UTLS. occurred on relatively flat terrain, with a mean elevation Extreme aerosol loading within a pyroCb induces of 400 m MSL. Regardless of region, the North Ameri- a microphysical shift (from indirect aerosol effects) to- can Regional Reanalysis (NARR; Mesinger et al. 2006) ward smaller cloud droplets and ice particles (Rosenfeld vegetation layer indicates that all pyroCb were associated et al. 2007; Reutter et al. 2014; Chang et al. 2015), with fires located in dense forest or a mix of forest and producing a much larger daytime near-infrared (4 mm) chaparral vegetation (Fig. 1, green shading). This section reflectivity than traditional convection (e.g., Rosenfeld describes the pyroCb inventory, a control inventory of et al. 2007). Application of a 4- and 11-mm brightness intense fire activity in the absence of pyroCb, and the temperature difference (BTD4–11) therefore allows for meteorological data used in the analysis of these events. separation of IPCB from more pristine traditional convection (Peterson et al. 2017). However, the dependence a. PyroCb detection and inventory on 4-mm reflectivity currently limits pyroCb detection to

PyroCb detection for GOES-W begins with identiﬁ- daytime scenes (solar zenith angle , 808). Two BTD4–11 cation of deep convection near observations of active thresholds are employed based on regional variability in ﬁres (described in the next section). Intense pyroCb convective clouds, which is driven by meteorological (IPCB) must have large anvil ice clouds (Fig. 2), and conditions. Output imagery for the Papoose Fire in Col- therefore must exhibit a thermal infrared (11 mm) orado (Fig. 2a) highlights the ability of these thresholds brightness temperature (BT11) near an approxi- to separate the unique microphysics of IPCB activity mated homogeneous liquid-water freezing threshold from traditional convection. A cloud opacity test based

(BT11 ,2388C; Wallace and Hobbs 2006; Rosenfeld on a modified split window (11- and 13-mm brightness et al. 2007). However, to account for averaging of con- temperature difference) is also applied to reduce po- vection within the large pixel size of GOES-W (20– tential cloud edge noise and detection errors from thin 2 110 km ), the pyroCb algorithm employs an IPCB BT11 clouds passing over a hot fire. Imagery for the Pony/Elk threshold of 2358C(Peterson et al. 2017). All 26 events fire in Idaho (Fig. 2b) highlights the removal of thin in the pyroCb inventory reach this IPCB BT11 threshold, cloud-edge pixels by the opacity test in the IPCB anvil

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2239

TABLE 1. Intense pyroCb (IPCB) and high FRP (HFRP) observations by surface elevation and region during 2013.

Category All fires IPCB fires HFRP fires IPCB events HFRP events Fire surface elevation ,1000 m 44 2 (5%) 36 (82%) 3 68 1000–2000 m 20 3 (15%) 19 (95%) 7 74 2000–3000 m 21 4 (19%) 20 (95%) 8 55 $3000 m 3 2 (67%) 2 (67%) 8 5 Geographic location MCONUS 48 9 (19%) 44 (92%) 23 145 Canada 40 2 (5%) 33 (83%) 3 57 All 88a 11 (13%) 77 (88%) 26a 202 a Based on the pyroCb inventory of Peterson et al. (2017). region. (The entire suite of imagery products based on between observed IPCB events and a control inventory these methods is posted in near–real time at the NRL of intense burning without the associated presence of pyroCb website: http://www.nrlmry.navy.mil/pyrocb- deep convection. This is accomplished by employing the bin/pyrocb.cgi.) GOES Wildfire Automated Biomass Burning Algorithm Following Peterson et al. (2017), the time corre- (WF_ABBA; Prins and Menzel 1994; Prins et al. 1998), sponding with a specific pyroCb is defined as the first which retrieves fire pixels by taking advantage of the GOES-W scan meeting IPCB criteria. However, in this spectral contrast between a pixel containing fire and the study, multiple pulses of IPCB activity within 6 h are surrounding nonfire region using the same 4- and 11-mm considered a single event. The resulting IPCB inventory infrared channels as the pyroCb detection algorithm. is based on processing of 15-min GOES-W imagery for WF_ABBA also provides an estimate of instantaneous daylight hours (roughly 6000 scenes), including addi- fire radiative power (FRP; units of MW) for detected fire tional scenes available during Rapid Scan Mode (http:// pixels, which is a measure of radiant heat output and www.ospo.noaa.gov/Operations/GOES/west/rso.html). commonly used to approximate fire intensity (e.g., Table 1 provides details of IPCB observations for each Kaufman et al. 1998; Giglio et al. 2003; Wooster et al. region and surface elevation category. Of the 88 fires 2005; Schroeder et al. 2010; Peterson et al. 2013b). The examined, 11 (13%) produced IPCB activity (Fig. 1, true FRP value is occasionally higher than the GOES-W blue triangles), totaling 26 individual events. The ma- observation due to sensor saturation. jority of IPCB activity (23 events from 9 fires) was ob- In contrast with traditional calculations of linear ‘‘fire served in MCONUS, with only 3 IPCB events observed line intensity’’ (Byram 1959), satellite retrievals provide in Canada (from 2 fires). The details of each individual FRP released over a pixel area. Combined FRP in- IPCB event, including time, latitude, and longitude, are formation from WF_ABBA (;4-km resolution) and provided in Peterson et al. (2017). the Moderate Resolution Imaging Spectroradiometer As the viewing zenith angle (VZA) of GOES-W in- (MODIS) fire detection algorithm (MYD14/MOD14, creases away from nadir, pixel size also increases from 24 ;1-km resolution; Giglio et al. 2003; Giglio 2010) are to 39 km2 in MCONUS (VZA , 608) to 55–109 km2 in used to select clusters of active fire pixels, where a large the boreal regions of Canada (608,VZA , 758; Fig. 1, total FRP is retrieved over an area of 625 km2 during orange lines). Peterson et al. (2017) show that IPCB de- either a single day or a period of up to 5 days. A tection is possible in Canada, even at extreme VZAs threshold of 140 000 MW is employed to select the 88 (pixel size . 100 km2). However, a few IPCB events with fires shown in Fig. 1 (Peterson et al. 2017), from which small anvils may remain undetected, likely resulting in an the IPCB and control inventories are derived. underestimate of Canadian activity. Marginal pyroCb The control inventory used in this study is based on activity without an anvil (BT11 $ 2358C; failing to reach high FRP events (HFRP; Table 1) observed during the the UTLS) and small pyroCu will not be detected in time series of each individual fire. A consistent and MCONUS or Canada using current GOES-W imagery. representative hourly FRP time series is produced by normalizing fire detections from WF_ABBA (GOES-W) b. Control inventory of intense fire activity over a box marking the maximum fire boundaries As highlighted by Peterson et al. (2015), intense (Peterson et al. 2015). HFRP events are identified by burning and rapid fire spread do not always yield devel- the 90th percentile of hourly FRP, including hours only opment of pyroCb. Therefore, a detailed understanding where the observed minimum BT11 is warmer of pyroCb development also requires comparisons than 2208C. This BT11 threshold is warm enough to

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2240 MONTHLY WEATHER REVIEW VOLUME 145 exclude HFRP cases with deep convective clouds in the levels, allowing for a variety of thermodynamic param- vicinity of the fire, but cold enough to allow HFRP eters to be calculated (described in section 4). This events influenced by shallow convection, such as cap- combined information is the basis for a detailed analysis ping pyroCu (Peterson et al. 2015). Each HFRP event of dynamic and thermodynamic conditions during IPCB must also feature 12 h of separation between IPCB ac- development and HFRP events. tivity and any previous HFRP events. These criteria ensure there is only one HFRP event per day in the 3. General meteorology absence of pyroCb influence. Of the 88 fires displayed in Fig. 1, 77 (88%) produced Generally speaking, pyroCb can occur at any point at least one HFRP event (Table 1). A total of 202 indi- in a fire’s lifetime, occasionally as early as the first 48 h. vidual HFRP events were observed, with 145 observed In the study region (Fig. 1), the median and mean times in MCONUS (44 fires) and 57 observed in Canada (33 of IPCB development are 4 and 7 days from the first fires). Periods of high fire danger and vigorous burning FRP observation, respectively. While development can are shorter in high-latitude boreal regions (e.g., Potter occur during sunset and nighttime, pyroCb activity 2012), especially periods free of convective clouds. This peaks during the late afternoon and early evening hours results in fewer HFRP data points in Canada (Table 1). (e.g., Fromm et al. 2010) at the same time as the diurnal The 11 fires that failed to produce HFRP criteria were peak in fire activity (Giglio 2007; Mu et al. 2011). The either impacted by consistent deep cloud cover or GOES-W pyroCb detection algorithm is also most ef- exhibited a short FRP time series without sufficient time fective during this time period. The majority of observed between IPCB events (e.g., the Yarnell Hill Fire). IPCB events (85%) in Table 1 initiate between 1900 and While the HFRP inventory successfully captures in- 0000 UTC, when the solar zenith angle is less than 608 tense fire activity in the absence of deep convection, it (Peterson et al. 2017). The majority of these IPCB also does not account for rapid changes in subpixel fire develop in interior sections of the western MCONUS, characteristics prior to pyroCb development (McRae where traditional convection often fails to reach the et al. 2015). Satellite retrievals of instantaneous, sub- tropopause (Sassen and Campbell 2001). This feature pixel fire area and temperature are available, but require helps distinguish regional microphysical perturbations extensive processing and filtering (e.g., Peterson and in pyroCb using BTD4–11 compared with more vigorous Wang 2013; Peterson et al. 2013b, 2014). Potential re- convective storms. Since background ice particles at lationships between these subpixel retrievals and IPCB cloud top are otherwise relatively warm and large, development will therefore be explored in future smaller perturbed crystals in pyroCb anvils become research. more conspicuous (Peterson et al. 2017). As described sequentially within the following sec- c. Meteorological data tions, traditional means for assessing fire conditions, Meteorological data corresponding to each IPCB and which would presumably become proxies for IPCB de- HFRP event are obtained from the NARR (http:// velopment, are incomplete. A more thorough consid- nomads.ncdc.noaa.gov/data.php#narr_datasets), archived eration of regional and synoptic-scale meteorology is at the National Climatic Data Center (NCDC). Through necessary to understand IPCB development relative to data assimilation, the NARR blends a variety of obser- HFRP events. vational data into Eta model output containing 29 pres- a. Standard fire weather variables sure levels over a mesh across the North American continent of ;32-km grid spacing (Ebisuzaki 2004; Standard fire weather forecasts generally depend on Mesinger et al. 2006). In this study, 3-hourly NARR data three primary surface variables: relative humidity (RH), are linearly interpolated to match the observation times of temperature, and wind speed (Potter 2012). These ‘‘dry, each GOES-W scene. The value for each variable is either hot, and windy’’ criteria are used in many forecasting the mean of all NARR grid box centroids located within applications, and for building fire weather indices (e.g., the maximum fire perimeter boundaries or the closest Amiro et al. 2004; Peterson et al. 2013a; Field et al. centroid to the fire boundaries. 2015). Figure 3 shows the distributions of these three The standard NARR dataset is also used to derive variables in the study region for each event type. IPCB variables relevant to forecasting fire weather conditions develop with a surface-based mean (median) RH of (e.g., Haines index; Haines 1988) and high-based con- 21% (20%) and temperature of 268C (298C). While the vection (e.g., upper-tropospheric divergence and lapse IPCB events were statistically hotter and drier com- rates; Nauslar et al. 2013). Thermodynamic profiles for pared to HFRP events, there is very little separation each event are derived from the NARR’s 29 pressure between the distributions. Surface wind speed (based

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2241

(Sassen and Campbell 2001). This ‘‘monsoonal anticyclone,’’ in combination with a low-level thermal trough, facilitates moisture advection from the tropical eastern Pacific, Gulf of California, and/or Gulf of Mexico (Higgins et al. 1997). The position and intensity of the anticyclone fluctuates during the summer, shifting the direction of moisture transport and the location of convective development (Johnson et al. 2007). The majority of IPCB activity (16 events triggered by 6 fires) was observed during the monsoon onset phase (1 June–5 July), before precipitation increases fuel moisture. Figure 4a displays the synoptic pattern for IPCB ac- FIG. 3. Distributions of standard surface fire weather variables vs event type. Blue and red colors are associated with IPCB and tivity observed during the Yarnell Hill (Arizona) and HFRP events, respectively. The boxes are bounded by the 25th and Silver (New Mexico) fires on 30 June. The monsoonal 75th percentiles, with the median indicated as a line bisecting each anticyclone was centered over southern Nevada, allow- box. The whiskers indicate the 10th and 90th percentiles of the ing for midlevel moisture advection and development of data, and mean values are displayed as triangles. The corre- traditional convection over the high terrain surrounding sponding number of data points is included at the top of each boxplot. each fire. IPCB activity observed on 27 June, during the Silver fire, as well as the Papoose and West Fork fires (Colorado) was associated with a similar synoptic envi- on a 3-hourly mean) shows the least distinction between ronment (Fig. 4b), with the center of the anticyclone event type and region, with mean and median values of shifted to the border between New Mexico and Colo- 2 about 5 m s 1 (9.7 kt) for all events. Standard dry, hot, rado. This orientation reduces midlevel moisture ad- and windy surface criteria are therefore unable to ef- vection, but is still suitable for development of fectively distinguish between IPCB and HFRP events. traditional high-based convection at the extreme surface While some regional variability is expected, the dis- elevation of the fires (.3000 m MSL in Colorado; tributions of the surface meteorology in Fig. 3 are very Table 1). similar to the values described in the literature for all In the Pacific Northwest and much of California, extreme characteristics of wildfires, including rapid fire IPCB activity generally develops over fires located spread, crowning, spotting, pyroCb, and even non- along the leading edge of an approaching disturbance. supercell, fire-generated tornados or ‘‘landspouts’’ (e.g., This ‘‘West Coast disturbance’’ pattern is described in Cunningham and Reeder 2009; Werth et al. 2011; Cruz relation to the 2013 Rim Fire (19 August) by Peterson et al. 2012; McRae et al. 2013; Peterson et al. 2015). This et al. (2015), and is illustrated in Fig. 4c. A cutoff area suggests that standard dry, hot, and windy surface cri- of low pressure near the coast allows for advection of teria are absolutely paramount for extreme wildfire subtropical midlevel moisture into central California, development and maintenance of rapid fire spread. providing a thermodynamic environment favorable However, this is where the correlative relationship be- for high-based convection, especially in the vicinity of tween these traditional indices and potential for con- elevated terrain. The synoptic pattern on 12 August vective initiation ends. for IPCB activity during the Pony/Elk and Beaver Creek fires (southern Idaho) is displayed in Fig. 4d. b. Synoptic pattern during intense pyroconvection This environment was very similar to the Rim Fire Inspection of the mid- and upper-level synoptic en- case, with a trough of low pressure located along the vironment during all IPCB activity from June to August Washington coast. Broad southerly flow was induced 2013 reveals that each event was associated with one of over ongoing fires, producing moisture advection in three distinct patterns: 1) monsoonal anticyclone, 2) West the midtroposphere and initiation of traditional high- Coast disturbance, and 3) Canadian ridge breakdown. based convection over elevated terrain. Both synoptic Examples are provided in Fig. 4.DuringMayand environments are also typical of rapid fire spread June, the southwestern MCONUS experiences a tran- (Werth and Ochoa 1993)andmajordrylightning sition from a midlatitude transient synoptic environ- outbreaks (Rorig and Ferguson 1999; Wallmann et al. ment to the summer monsoonal pattern (e.g., Higgins 2010) across much of the Pacific Northwest and et al. 1997; Johnson et al. 2007). Increased surface Northern California. A total of seven IPCB events heating causes a northeastward migration, broaden- triggered by three fires were associated with similar ing, and flattening of the subtropical Pacific ridge synoptic patterns in these regions.

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2242 MONTHLY WEATHER REVIEW VOLUME 145

FIG. 4. Example synoptic maps for IPCB events inﬂuenced by (a),(b) monsoonal anticyclone; (c),(d) West Coast disturbance; and (e),(f) Canadian ridge breakdown patterns. Black contours indicate NARR 500-hPa heights, plotted at 20-m intervals in (a)–(d) and 40-m intervals in (e),(f). Red contours indicate upper-tropospheric divergence, plotted 2 2 every 1 3 10 5 s 1. Shading indicates the mean midlevel relative humidity and orange triangles indicate the locations of ﬁres producing IPCB activity. The dashed white line in (e) indicates the axis of a short-wave trough.

In the boreal forests of western Canada, active ﬁre Macias Fauria and Johnson 2006; Peterson et al. 2010; seasons, periods of intense burning, and dry lightning Potter 2012). However, ﬁres tend to exhibit the most ex- strikes are commonly associated with the presence of a treme characteristics as the corresponding ridge begins 500-hPa anticyclone (e.g., Skinner et al. 1999, 2002; to break down, which is most commonly associated

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2243 with the arrival of a surface cold front and an upper- level trough (Nimchuk 1983; Westphal and Toon 1991). This pattern exhibits many similarities with the West Coast disturbance pattern. The primary differences in Canada are the relative lack of complex topography and closer proximity to the polar jet stream. This likely means that dynamic forcing mechanisms, including frontal boundaries, typically play a larger role in convective development. Figure 4e shows the synoptic pattern for IPCB activity during a large wildfire in Manitoba on 1 July. The fire was located within a strong 500-hPa height gradient associated with an approaching short-wave trough and surface cold front, highlighting the classic ‘‘ridge breakdown’’ pattern. Figure 4f shows the synoptic pattern during IPCB activity in the Northwest Territories. In this case, the fire was located within the ridge portion of a ‘‘Rex Block’’ pattern (Rex 1950a,b), ahead of an approaching long-wave trough along the west coast of FIG. 5. Distributions of dynamic forcing variables vs event type, Canada. All three Canadian IPCB events (two fires) using (a) upper-tropospheric divergence and (b) 500-hPa absolute developed under similar synoptic environments. While vorticity advection. Blue and red colors are associated with IPCB regional variability is expected, all synoptic patterns and HFRP events, respectively. The boxes are bounded by the 25th and 75th percentiles, with the median indicated as a line bisecting highlighted in Fig. 4 induce favorable fire weather con- each box. The whiskers indicate the 10th and 90th percentiles of the ditions near the surface, along with mid- and upper-level data, and mean values are displayed as triangles. The corresponding conditions favorable for convective sustenance. number of data points is included at the top of each boxplot. c. Dynamic forcing (Fig. 5a, positive values) during the majority of IPCB Development of deep convection requires an envi- activity observed across western North America (Fig. 1). ronment conducive to rising motion within the full depth In contrast, more than 50% of HFRP events, specifically of the troposphere. Calculations of divergence in the defined as devoid of deep convection, were generally upper troposphere are commonly used to infer upper- observed with weak convergence in the upper tropo- tropospheric jet structure and the magnitude of quasi- sphere (Fig. 5a, negative values). geostrophic forcing mechanisms influencing vertical While some upper-tropospheric divergence was typi- motion (e.g., Nauslar et al. 2013). In the southwestern cally present during IPCB development, its magnitude MCONUS, upper-tropospheric divergence during IPCB was much weaker in comparison with major severe development is typically a result of the summer mon- thunderstorm outbreaks observed in the central MCONUS soonal anticyclone (Figs. 4a,b), fed by intense diabatic (e.g., Knupp et al. 2014). Similarly, large-scale quasi- heating over elevated terrain. The Pacific Northwest and geostrophic forcing mechanisms, including strong ad- Northern California are located between the monsoonal vection of absolute vorticity (Fig. 5b), were typically pattern and the storm track associated with the polar jet. absent during IPCB development. Many IPCB events Upper-tropospheric divergence and convective initiation were therefore associated with weak rising motion in this region (Figs. 4c,d) typically results from a combi- (negative omega) in an environment only marginally nation of terrain effects and mesoscale jet structures (e.g., favorable for convective development (e.g., Wallmann Hamilton et al. 1998; Nauslar et al. 2013) embedded et al. 2010). Additional micro- and mesoscale influences, within the West Coast disturbance synoptic pattern such as topography and/or intense heating from a large (Peterson et al. 2015). Canadian IPCB events (Figs. 4e,f) fire (explored in section 7), may be required to trigger develop in the absence of complex topography. There- pyroconvective development. fore, upper-tropospheric divergence is primarily a result of transient disturbances along the polar jet. 4. Thermodynamic profiles Because of regional variation in tropopause height, the upper troposphere is defined as 250 hPa in MCONUS Several previous studies have attempted to describe the and 300 hPa in Canada. Each synoptic pattern dis- thermodynamic environment during pyroCb activity us- played in Fig. 4 produced upper-tropospheric divergence ing convective indices derived from nearby radiosondes

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2244 MONTHLY WEATHER REVIEW VOLUME 145

FIG. 6. Low-elevation (,1000 m) thermodynamic profiles derived from the NARR for (a) IPCB activity and (b) HFRP events. Thin red and blue profiles, respectively, correspond to environmental temperature and dewpoint for individual events, with line styles varying in (a). Mean profiles are displayed in (b) as thick solid red and blue curves. The corresponding standard deviations are displayed as dashed curves. Mean MULCL height is displayed as a solid purple horizontal line, with the minimum and maximum displayed as dashed lines. Vertical bars on the right indicate important layers described in the text. Wind barbs represent the mean conditions in each layer. Relevant variables and indices derived from the mean profiles are provided at the bottom.

(e.g., Fromm et al. 2010). However, these data are only events from the same fires also shows a reduction in available twice daily, and are often located a long distance dewpoint depression in the mixed layer, the mean mid- from the fire in an environment with much different me- tropospheric layer (top of the inverted V) is drier, thus teorology. By employing the NARR’s 29 pressure levels, weakening the inverted-V signature (Fig. 6b). thermodynamic profiles are derived at the location of All 41 midelevation fires (1000–3000 m MSL) were each fire during each IPCB and HFRP event. located within the interior of the western MCONUS (Fig. 1, Table 1). Significant daytime heating produced a a. Significance of the ‘‘inverted V’’ profile much deeper dry mixed layer [.3000 m above ground Figure 6 displays individual NARR-derived profiles level (AGL)] during IPCB activity (Fig. 7a)compared for each IPCB and HFRP event observed at the 44 low with the lower elevation Canadian fires. The combination elevation fires (,1000 m; Fig. 1, Table 1). The three of these deep mixed layers and increased midtropo- IPCB profiles, all associated with Canadian fires, spheric moisture (lower dewpoint depressions) produced exhibited a relatively deep (mean of 1950 m) and near well-defined inverted-V profiles for the majority of mid- dry-adiabatic mixed layer (Fig. 6a). A large near-surface elevation IPCB activity. However, midelevation profiles dewpoint depression, defined as the difference between for HFRP events (Fig. 7b) were similar to their low- temperature (Fig. 6a, red curves) and dewpoint (Fig. 6a, elevation counterparts, with reduced moisture (larger blue curves), gradually decreased with altitude, and re- dewpoint depressions) in the midtroposphere. IPCB mained relatively low in the midtroposphere. This ele- events observed during the three high elevation fires vated midtropospheric moisture content above a dry ($3000 m MSL) exhibited larger midtropospheric dew- and unstable mixed layer is commonly referred to as an point depressions than those observed during low- and inverted-V thermodynamic profile (e.g., Wakimoto midelevation fires (Fig. 8a), thus producing a weaker 1985; Evans et al. 2012; Tory and Thurston 2015). While inverted-V signature. High-elevation HFRP profiles the mean of all profiles corresponding to 68 HFRP were dry at all levels (Fig. 8b). The potential impact from

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2245

FIG. 7. Midelevation (1000–2000 m) thermodynamic profiles derived from the NARR for (a) IPCB activity and (b) HFRP events. Thin red and blue profiles, respectively, correspond to environmental temperature and dewpoint for individual events. Mean profiles are displayed as thick solid red and blue curves. The corresponding standard deviations are displayed as dashed curves. Mean MULCL height is displayed as a solid purple horizontal line, with the minimum and maximum displayed as dashed lines. Vertical bars on the right indicate important layers described in the text. Wind barbs represent the mean conditions in each layer. Relevant variables and indices derived from the mean profiles are provided at the bottom. variation in surface elevation on midtropospheric mois- Van Wagner 1987; Haines 1988). These characteristics ture content is explored in greater detail in section 6. combine to promote intense burning and vertical smoke An inverted-V profile is significant because it high- lofting, which are typical of ‘‘blow-up’’ fire events (McRae lights the same thermodynamic environment observed and Sharples 2014). In contrast, supercell thunderstorms during traditional high-based convection and dry light- and associated tornado outbreaks typically coincide ning activity in western North America (e.g., Wallmann with moist conditions in the lower troposphere and drier et al. 2010). A variety of studies have shown similar conditions aloft (e.g., the ‘‘loaded gun’’ profile; http:// thermodynamic profiles for individual pyroCb events in www.theweatherprediction.com/thermo/soundings/gun/). MCONUS, Canada, and Australia, usually via nearby This is a key distinguishing characteristic that corroborates radiosondes (e.g., Trentmann et al. 2006; Rosenfeld why pyroCb have only been identified previously in spe- et al. 2007; Cunningham and Reeder 2009; deLaat et al. cific geographic areas and climate zones. 2012; Fromm et al. 2012; Johnson et al. 2014; Lareau and b. Convective available potential energy Clements 2016). However, with limited sample sizes in previous work, the connection between pyroCb and Several studies have employed convective available high-based convection was only recently hypothesized potential energy (CAPE) to estimate the buoyancy (Peterson et al. 2015). available for high-based convection (e.g., Peterson An inverted-V profile represents relatively dry and hot et al. 2010; Wallmann et al. 2010) and high-altitude surface conditions, as well as a deep mixed layer favor- smoke plumes (Potter 2005). The calculation of CAPE able for strong smoke column updrafts in the presence of is either based on lifting a surface parcel, the most limited vertical wind shear (described in section 5). The unstable parcel in the lower troposphere (e.g., lowest width of an inverted-V base (large dewpoint depression) 300 hPa), or a parcel with the mean properties of the can be related to the moisture content of surface fuels, lower troposphere (Bunkers et al. 2002; Wallace and and thus the intensity with which they will burn (e.g., Hobbs 2006). However, Fromm et al. (2010) show that

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2246 MONTHLY WEATHER REVIEW VOLUME 145

FIG.8.AsinFig. 7, but for high-elevation ($3000 m) IPCB and HFRP events. several pyroCb are associated with little or no CAPE increases the likelihood of convective development, it is (surface based). When an inverted-V sounding profile not an absolute indicator of pyroCb development. is present, air parcels originate from within a hot and c. PyroCb condensation level dry mixed layer and eventually encounter increased moisture near and mostly above the lifting condensa- Having now established that the development of tion level (LCL), but no variant of CAPE considers high-based convection and pyroCb is dependent on entrainment (Holton 2004). CAPE calculations also varying thermodynamic processes within the lower and fail to account for an increase in buoyancy produced by midtroposphere, it is desirable to examine differences the intense heat of the fire (e.g., Potter 2005; Lareau at each level relative to event characteristics. However, and Clements 2016). variation in surface elevation and latitude across the Because of the unknown magnitude of sensible heat- study region (Fig. 1) greatly affects each layer’s relative ing from each individual fire, CAPE values for each depth, thus preventing the use of a fixed standard event in this study are determined by lifting the most pressure level to mark the boundary between layers. unstable (MU) parcel below 400 hPa (MUCAPE). The The most logical means to separate lower- and mid- height of the MU parcel ranged from the surface to tropospheric processes during pyroCb is the dynamic 600 m AGL for more than 75% of observed IPCB events cloud base altitude, which is commonly inferred using (mean of 350 m). The mean MU parcel height was only calculations of the LCL. 200 m AGL for HFRP events. Figures 6–8 provide the Similar to the calculation of MUCAPE, the dynamic mean MUCAPE values for each event type, showing LCL for pyroCb activity is determined by lifting the that IPCB events exhibited higher MUCAPE than MU parcel below 400 hPa. The resulting most unstable HFRP events, especially for low-elevation fires. This LCL(MULCL)canthenbeappliedtoseparatelower- suggests that MUCAPE can be useful for identifying and midtropospheric analysis. However, an intense fire IPCB potential. However, all IPCB events occurred often increases the temperature of air parcels entering with less MUCAPE than traditional severe thunder- its plume (Potter 2005), producing higher saturation storms (e.g., Knupp et al. 2014), and several IPCB events vapor pressures compared with the ambient back- from high-elevation fires developed with no MUCAPE. ground. This delays condensation thereby increasing Therefore, while the presence of MUCAPE presumably the LCL height and associated cloud base (Luderer

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2247 et al. 2009; Lareau and Clements 2016). The MULCL thereforeservesasalowerboundaryforcloud-base altitude based on ambient lower-tropospheric conditions. Figures 6–8 display the mean MULCL (solid purple), as well as minimum and maximum values (dashed purple). While MULCL height is similar for IPCB and HFRP events, it is highly variable, especially for midelevation fires. The dynamic MULCL altitude indicates the top of the lower-tropospheric layer (Figs. 6–8, green vertical bars) and the bottom of the midtropospheric layer (Figs. 6–8, pink vertical bars) for each individual IPCB and HFRP event. Any overlap shown in Figs. 6–8 results from variation in MULCL altitude within the bulk dataset. Be- cause of variation in the depth of the troposphere produced by the large latitudinal extent of the study region (Fig. 1), the upper boundary of the midtropospheric layer is fixed at 300 hPa in MCONUS and 400 hPa in FIG. 9. Distributions of lower-tropospheric stability indices vs Canada. These levels are selected because they are lo- event type using (a) the Haines index and (b) lower-tropospheric cated just below the mean height of upper-tropospheric lapse rate. Blue and red colors are associated with IPCB and HFRP events, respectively. The boxes are bounded by the 25th and 75th features (e.g., jet stream) in each region. The following percentiles, with the median indicated as a line bisecting each box. sections describe the relevant meteorology in each layer, The whiskers indicate the 10th and 90th percentiles of the data, and progressing from lower to midtroposphere. mean values are displayed as triangles. The corresponding number of data points is included at the top of each boxplot. Horizontal lines in (a) indicate potential for plume-dominated behavior as 5. Lower troposphere described by the Haines index. In addition to surface conditions favorable for intense fire activity (section 3a), lower-tropospheric conditions the green vertical bar denoting the lower troposphere favorable for pyroCb development must support a ro- based on the dynamic MULCL definition. This display bust convective smoke column capable of reaching be- indicates that the HI layer is often too narrow to ade- yond the MULCL. This situation is commonly referred quately measure the lower-tropospheric lapse rate. This is to as a ‘‘plume-dominated fire,’’ where the fire is creat- especially true for midelevation thermodynamic profiles ing and sustaining its own favorable weather conditions (Fig. 7), where the midelevation HI variant often starts (e.g., Rothermel 1991; Werth and Ochoa 1993). Plume- below the first profile layer, and does not account for the dominated behavior is primarily driven by increased large depth of the mixed layer. Applying the high- lower-tropospheric instability (e.g., Potter et al. 2008) elevation variant to these profiles increases the charac- and reduced wind shear (e.g., Byram 1954). terization of mixed layer depth, but would exclude the lowest levels of the profile. High-elevation fires (Fig. 8) a. Instability are the only category where the HI layer compares well to In western North America, regional potential for the MULCL-based lower-tropospheric layer. plume-dominated behavior is most commonly quantified The HI has a fixed upper bound of 6, corresponding to using the Haines index (HI): an integer scale (1–6) based the most extreme potential for plume-dominated be- on two equally weighted ingredients for moisture and havior (Haines 1988; Potter et al. 2008). Figure 9a shows stability, respectively derived from a lower-level (850 or that the majority of IPCB and HFRP events in MCONUS 700 hPa) dewpoint depression and lower-atmospheric exhibited at least moderate potential for plume- lapse rate (Haines 1988; Potter et al. 2008). The specific dominated activity (HI $ 5). An example of high- and pressure levels used for HI calculations vary, depending midelevation HI variants is provided in Fig. 10a for IPCB on local topography. Originally based on mandatory activity during two large fires in the complex topography pressure levels from soundings obtained in the late 1980s, of Idaho on 12 August (as in Fig. 4d). Both variants the ‘‘low,’’ ‘‘mid,’’ and ‘‘high’’ elevation variants of the HI yield at least moderate potential for plume-dominated and their application are very subjective. behavior, but the depth of the lower-tropospheric layer Figures 6–8 include respective HI layers for each fire and magnitude of instability are unclear. It is therefore elevation category as an orange vertical bar adjacent to challenging to decipher which HI variant is most

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2248 MONTHLY WEATHER REVIEW VOLUME 145

FIG. 10. Maps quantifying lower-tropospheric instability for IPCB activity observed during the Pony/Elk and Beaver Creek fires on 12 Aug using (a) the Haines index and (b) lower-tropospheric lapse rate. Green and blue contours in (a) represent the high-elevation variant of the Haines index. Green and blue contours in (b) represent MULCL height. Brown contours indicate surface elevation. Black symbols indicate the locations of fires producing IPCB activity. applicable. The HI upper limit can be extended for in- events were both associated with low bulk shear values 2 creased sensitivity to the most extreme fire danger (Mills averaging 3–8 m s 1 (6–16 kt). As shown by the mean wind and McCaw 2010). However, this ‘‘continuous Haines profiles in Figs. 6–8, IPCB and HFRP events also de- index’’ may also lack the fidelity to separate pyroCb veloped with very little low-level directional shear. How- from other extreme wildfire characteristics (Peterson ever, IPCB events generally exhibited the largest LTLR et al. 2015). values (Fig. 9b), and therefore more lower-tropospheric An improved measure of lower-tropospheric in- instability, which presumably maximized potential for stability, free from subjective pressure levels, can be plume-dominated behavior. While the combination of a produced by calculating the lapse rate from the surface large LTLR and reduced low-level shear is important for to the dynamic MULCL (Figs. 6–8, green vertical bar). plume-dominated behavior, it is unrepresentative of se- Plume-dominated potential increases as this lower- vere thunderstorm and tornado outbreaks east of the tropospheric lapse rate (LTLR) approaches dry adia- Rocky Mountains, which often coincide with larger low- 2 2 batic (29.88Ckm 1), especially when the mixed layer is level bulk shear values (occasionally . 20 m s 1), larger deep (Jenkins 2004). The example in Fig. 10b (1800 UTC) directional shear, and increased low-level moisture (Evans shows that both Idaho fires were located within a deep and Doswell 2001; Knupp et al. 2014). This is a primary (3000 m) lower-tropospheric layer with a LTLR less characteristic distinguishing IPCB phenomenology rela- 2 than 28.08Ckm 1, at least 3 h before IPCB initiation tive to traditional vigorous convection. (2100 UTC). Across the study region, the majority of IPCB development occurred with LTLR values less 2 6. Midtropospheric moisture and instability than 28.58Ckm 1,withHFRPeventsoftenlinkedto 2 LTLR values greater than 28.08Ckm 1 (Fig. 9b). A robust convective smoke column in the lower troposphere is not likely to produce IPCB activity unless b. Vertical wind shear conditions in the midtroposphere are also favorable for Intense fires burning in an environment with large ver- development of high-based convection (Peterson et al. tical wind shear in the lower troposphere are likely to 2015). The inverted-V thermodynamic profiles in exhibit tilted and less concentrated smoke plumes, Figs. 6–8 show that ambient midlevel moisture likely decreasing the likelihood of deep pyroconvection (e.g., plays an important role in IPCB development, which Lareau and Clements 2016). Figures 6–8 provide the wind supports the results of several modeling studies (e.g., vector difference (or bulk shear vector) from the surface to Trentmann et al. 2006; Luderer et al. 2009). However, the MULCL altitude for each event type. IPCB and HFRP the amount of midtropospheric moisture required has

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2249 never been quantiﬁed nor clearly correlated with event occurrence. By employing the dynamic MULCL-based deﬁnition of midtroposphere (Figs. 6–8, pink vertical line), mean midtropospheric RH (RHMT) and total precipitable water (TPWMT) can be calculated, thus providing relative and absolute measures of moisture content.

Figure 11a provides distributions of RHMT values for IPCB and HFRP events as a function of fire surface elevation. These are the same elevation categories as the profiles in Figs. 6–8, with the midelevation category split into two 1000-m intervals. IPCB events corresponded with RHMT values ranging from 40% to 60% within the two lowest-elevation categories, but decreasing to 20%– 40% for the highest-elevation category. Figure 4 highlights the spatial variability of RHMT found for several fires at varying elevations. As expected, Canadian IPCB events (fire elevation , 500 m) were associated with the largest RHMT values (Figs. 4e,f). The inverse relationship with surface elevation is even more pronounced when using TPWMT (Fig. 11b). Low- elevation IPCB events corresponded with TPWMT values ranging from 7 to 14 mm, decreasing to 1–2 mm for high-elevation events. While few comparisons exist in the literature, Johnson et al. (2014) report TPWMT values (above 600 hPa) of 5.0 mm for pyroCb activity observed during a 2012 wildfire at an approximate elevation of 2000 m MSL in Colorado, which compares well to the midelevation TPWMT distributions. As fire surface elevation increases above sea level, a shallower and more elevated portion of the vertical profile is used to calculate TPWMT. In addition, less latent heating is required to develop convection capable of reaching the UTLS, especially with the deep lower-tropospheric mixed layers typical of IPCB events (Figs. 6–8). It is therefore difficult to highlight a single threshold of FIG. 11. Distributions of midtropospheric moisture and in- RHMT or TPWMT representative of IPCB development stability variables vs event type and surface elevation, using mid- across all of western North America. tropospheric (a) relative humidity (RHMT), (b) total precipitable HFRP events also exhibited an inverse relationship water (TPWMT), and (c) potential instability (PI). Blue and red between the two midlevel moisture metrics and surface colors are associated with IPCB and HFRP events, respectively. The boxes are bounded by the 25th and 75th percentiles, with the elevation. However, the mean and median values for median indicated as a line bisecting each box. The whiskers in- HFRP events were lower than those for IPCB events in dicate the minimum and maximum values of the data, with mean all but the highest-elevation category (Figs. 11a,b). values displayed as triangles. The corresponding number of data These results, along with the thermodynamic profiles in points is included at the top of each boxplot. Figs. 6–8, reinforce the importance of ambient midlevel moisture during IPCB development, especially for fires at low surface elevations. height (due/dz , 0). In this study, PI is approximated In addition to a moisture source, high-based con- via the mean due/dz for NARR vertical levels convection requires unstable conditions in the midtropo- tained within the midtropospheric layer (Figs. 6–8,pink sphere, which can be quantified using several metrics. vertical line). Similar to RHMT and TPWMT, an inverse Equivalent potential temperature ue, a function of both relationship exists between PI and surface elevation moisture and temperature, is used to identify potential (Fig. 11c). IPCB events originating from fires below instability (PI), which is present when ue decreases with 2000 m MSL coincided with PI values below or near

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2250 MONTHLY WEATHER REVIEW VOLUME 145

FIG. 12. Maps quantifying midtropospheric moisture and instability for IPCB activity observed during (a) a large ﬁre in Manitoba on 1 Jul and (b) the Pony/Elk and Beaver Creek ﬁres on 12 Aug. Shading indicates the mean

midtropospheric PI. Blue contours represent RHMT and brown contours indicate surface elevation. Black symbols indicate the locations of ﬁres producing IPCB activity. zero, suggesting a strong and/or deep midtropospheric 7. PyroCb conceptual model summary layer where due/dz was less than zero. While PI values became positive for high-elevation IPCB, they remained The detailed meteorological analysis provided in the lower than their HFRP counterparts in every elevation preceding sections shows that a conceptual model for category (Fig. 11c). IPCB development (Fig. 13) must be based on a synoptic The collocation of PI with increased moisture content environment that is at least marginally favorable for in the midtroposphere is paramount for development of high-based convection, which is characterized by an high-based convection (Nauslar et al. 2013). Figure 12a inverted-V thermodynamic proﬁle near the surface. highlights a region of substantial PI found coincident with Every IPCB event begins in the lower troposphere

RHMT values greater than 60% approaching a Canadian (Fig. 13, yellow box) with a large and/or intense fire fire (surface elevation 5 186 m) 1.5 h before IPCB initi- capable of developing a deep convective smoke column. ation. These conditions, along with an approaching short- Intense burning is maintained by dry, hot, and somewhat wave trough and dynamic forcing (Fig. 4e), allow for windy conditions at the surface, corresponding with a rapid convective development, including IPCB activity. high or extreme fire danger rating from the suite of In contrast, two Idaho fires influenced by the West Coast surface-based fire weather indices (e.g., Van Wagner disturbance synoptic pattern (Fig. 4d) experienced much 1987; Haines 1988; Mills and McCaw 2010). Fuel loads, weaker PI, but with similar RHMT values (Fig. 12b). localized fire dynamics, terrain factors, and their com- IPCB activity was likely able to develop over both fires bined interaction may also play a role (e.g., McRae due to the higher surface elevation (1500–2500 m). et al. 2015). The lower-tropospheric layer (surface to Several additional methods have been developed to MULCL) is usually deep (typically 2000–4000 m), dry, 2 quantify mid- and upper-tropospheric instability. Pri- and unstable (LTLR ,28.58Ckm 1), with minimal mary examples include the upper-tropospheric lapse wind shear (both speed and directional). This increases rate (UTLR; Wallmann et al. 2010) and high-level total the potential for plume-dominated behavior, which in totals (HLTT; Milne 2004). However, similar to the HI, turn increases the likelihood of the plume reaching be- these methods rely on fixed pressure levels, and there- yond the MULCL. fore have limited utility for analysis and forecasting of Near and above the MULCL, a midtropospheric high-based convection and IPCB across large regional moisture source is paramount, often coincident with a domains with varying topography. UTLR values, de- layer of PI (due/dz , 0). This increases the potential for fined here as the lapse rate between the MULCL and the moisture entrainment, condensation, and latent heat top of the midtropospheric layer (400 or 300 hPa), are release within the rising smoke plume, driving it higher provided in Figs. 6–8. HLTT calculations are not con- into the mid- to upper troposphere, and initiating sidered in this analysis. pyroCb development (Fig. 13). However, the quantity of

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2251

than traditional convection. IPCB activity observed during the Pony/Elk fire in Idaho (10 August) was associated with a mean surface elevation of 1593 m MSL (Fig. 2b), while nearby traditional convection occurred over terrain with elevations of 2000–3000 m MSL. Similarly, both IPCB events observed during the Cal- ifornia Rim Fire (19 and 21 August) developed over lower terrain than traditional convection (Peterson et al. 2015). Intense FRP may also provide the extra convective heating and buoyancy needed to break a capping inversion, which would be most likely within the warm sector of an approaching disturbance during Canadian ridge breakdown. IPCB events may therefore occasionally develop in the absence of traditional triggering mechanisms when an otherwise favorable thermodynamic and synoptic environment is in place. This potential relationship between FRP, convective heating, thermal buoyancy, and IPCB development remains an important topic for future research. In western North America, meteorology favorable for high-based convection is frequently observed during the entire primary fire season (May–September). This sug- FIG. 13. Schematic of a meteorology-based conceptual model for development of IPCB activity in western North America. The gests that IPCB activity can also be expected every fire yellow box indicates processes occurring near the surface and season. Several fire-prone regions in Asia (e.g., Siberia) within the mixed layer. Background photo (via NWS Sacramento and Australia also exhibit conditions favorable for high- and S. Ellsworth) shows development of deep pyroconvection based convection during the fire season, which likely near Lake Tahoe during the 2014 King Fire (available online explains the perennial occurrence of IPCB activity in at http://wildfiretoday.com/2014/09/14/california-king-fire-near- pollock-pines/). these regions. Any region with a low frequency of high- based convection (e.g., the Gulf Coast) therefore has a reduced likelihood of IPCB activity. The meteorology midlevel moisture and magnitude of PI required for highlighted in the conceptual model (e.g., inverted-V development are strongly dependent on the surface profile) is uncommon in tropical regions, which likely elevation of the contributing fire, with lower-elevation explains why IPCB events are rarely (if ever) observed fires requiring greater PI and relative moisture. Mid- at low latitudes. and upper-tropospheric dynamic forcing is usually weak, but can play an important role for fires with 8. Conclusions minimal orographic lift, especially in close proximity to the polar jet (e.g., Canadian fires). Many IPCB events This study describes a conceptual model for intense have therefore been linked to at least some upper- pyroCb (IPCB) development using the combination of tropospheric divergence. meteorological reanalysis data and a remote sensing– This conceptual model explains why IPCB events based inventory of 26 events (Peterson et al. 2017) ob- are often observed in close proximity to or embedded served in western North America during the primary fire within traditional high-based convection (e.g., Fromm season (June–August) of 2013. One of the key compo- et al. 2006; Peterson et al. 2015, 2017). For example, nents of this model is a deep, dry, and unstable lower- IPCB activity during the Papoose fire in Colorado tropospheric mixed layer surmounted by a moist and (28 June) was nearly surrounded by traditional con- unstable midtropospheric layer, which is typical of tra- vection (Fig. 2a), with additional IPCB events ob- ditional high-based convection (dry thunderstorms) served in the same region (New Mexico Silver fire; found over elevated terrain in semiarid regions. Varia- Peterson et al. 2017). However, in a few cases, in- tion in topography and thermal buoyancy produced by creased thermal buoyancy associated with intense FRP intense fire activity are also key factors in IPCB devel- will act as its own triggering mechanism. The most opment. Traditional surface fire weather variables and common manifestation of this effect is an IPCB event lower-tropospheric indices (e.g., Haines index) do not developing earlier in the day and/or over lower terrain provide enough information to quantify IPCB potential

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2252 MONTHLY WEATHER REVIEW VOLUME 145 by themselves. Periods of intense burning [high fire ra- information is paramount for improved fire suppression diative power (HFRP)] that fail to produce IPCB ac- efforts. In addition, the next generation of geostationary tivity usually lack a midtropospheric moisture source, satellites (e.g., GOES-R; see online at http://www.goes-r. which is the strongest selector for IPCB activity in the gov/) will greatly improve IPCB and fire detection ca- conceptual model. IPCB events have other distinctions pabilities, especially in regions with large satellite viewing from HFRP events, including greater instability in the angles (e.g., Canada). Therefore, the combination of lower troposphere and more favorable dynamics (di- meteorology from numerical forecast models and im- vergence) in the upper troposphere. proved exploitation of satellite observations can be Conditions driving IPCB development are shown to employed to develop future IPCB prediction and be fundamentally different from other severe weather monitoring capabilities. events, including supercell thunderstorms and associated tornado outbreaks. This suggests that the IPCB Acknowledgments. This research was performed conceptual model does not describe the most favorable while David Peterson held a National Research Council environment for deep overshooting convection. How- Research Associateship Award at the Naval Research ever, it is the only meteorological environment that will Laboratory in Monterey, California. Support was pro- support intense fire activity occurring at the same time vided by the Naval Research Laboratory Base Funding as deep, moist convection. The IPCB conceptual model Program. Additional support was provided by the NASA therefore reconciles the most efficient mechanism for Air Quality Applied Science Team under Award injecting relatively large quantities of aerosol particles NNH09ZDA001N-AQAST. We are grateful to Melinda into the upper troposphere and lower stratosphere, Surratt at the Naval Research Laboratory for GOES-W aside from volcanic eruptions. data processing. High-based convection is frequently observed across western North America every fire season. Conditions REFERENCES favorable for high-based convection often coincide with ongoing large wildfires. This implies that IPCB, far from American Meteorological Society, 2016: ‘‘Pyrocumulonimbus.’’ Glossary of Meteorology. [Available online at http://glossary. the niche phenomenon they initially appeared to be ametsoc.org/wiki/Pyrocumulonimbus.] (e.g., Fromm et al. 2000; Fromm and Servranckx 2003), Amiro, B. D., K. A. Logan, B. M. Wotton, M. D. Flannigan, J. B. are very likely significant and endemic features of the Todd, B. J. Stocks, and D. L. Martell, 2004: Fire weather index regional summer climate. Meteorological conditions system components for large fires in the Canadian boreal forest. favorable for high-based convection and IPCB are also Int. J. Wildland Fire, 13, 391–400, doi:10.1071/WF03066. Bunkers, M. J., B. A. Klimowski, and J. W. Zeitler, 2002: The observed in other highly fire-prone regions of the world. importance of parcel choice and the measure of vertical wind Therefore, annually recurring IPCB events may very shear in evaluating the convective environment. Preprints, well represent a new and potentially significant modu- 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. lator of the lower stratospheric aerosol layer, with Meteor. Soc., P8.2. [Available online at https://ams.confex. a potential impact on global climate and circulation. A com/ams/SLS_WAF_NWP/techprogram/paper_47319.htm.] Byram, G. M., 1954: Atmospheric conditions related to blowup fires. companion paper (Peterson et al. 2017) describes a USDA Forest Service, Southeastern Forest Experiment Sta- practical methodology for identifying IPCB operation- tion, Asheville, NC, Station Paper SE-SP-35, 36 pp. [Available ally, which will allow for routine and retrospective online at http://www.treesearch.fs.fed.us/pubs/45778.] monitoring and inventorying of such events to help ——, 1959: Combustion of forest fuels. Forest Fire: Control and constrain this source in other fire-prone regions Use, K. P. Davis, Ed., McGraw-Hill, 61–89. Chang, D., and Coauthors, 2015: Comprehensive mapping and worldwide. characteristic regimes of aerosol effects on the formation and The IPCB conceptual model is also highly relevant to evolution of pyro-convective clouds. Atmos. Chem. Phys., 15, other hazards from high-based convection, including dry 10 325–10 348, doi:10.5194/acp-15-10325-2015. lightning strikes, microbursts, and dust storms (ha- Cruz, M. G., A. L. Sullivan, J. S. Gould, N. C. Sims, A. J. Bannister, boobs). For example, the deep and dry lower- J. J. Hollis, and R. J. Hurley, 2012: Anatomy of a catastrophic wildfire: The Black Saturday Kilmore East fire in Victoria, tropospheric layer (high MULCL) observed during Australia. For. Ecol. Manage., 284, 269–285, doi:10.1016/ IPCB activity is supportive of strong downdrafts and j.foreco.2012.02.035. microbursts during traditional high-based convection Cunningham, P., and M. J. Reeder, 2009: Severe convective storms (e.g., Wakimoto 1985). When considering that intense initiated by intense wildfires: Numerical simulations of pyro- fire activity and IPCB are often located near traditional convection and pyro-tornadogenesis. Geophys. Res. Lett., 36, L12812, doi:10.1029/2009GL039262. convection, potential for erratic fire behavior, such as a deLaat, A. T. J., D. C. S. Zweers, R. Boers, and O. N. E. Tuinder, rapid shift in spread direction, greatly increases (e.g., 2012: A solar escalator: Observational evidence of the self- Johnson et al. 2014; Hardy and Comfort 2015). This lifting of smoke and aerosols by absorption of solar radiation

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2253

in the February 2009 Australian Black Saturday plume. to the 1994 Palm Sunday tornado outbreak. Mon. Wea. J. Geophys. Res., 117, D04204, doi:10.1029/2011JD017016. Rev., 126, 2061–2089, doi:10.1175/1520-0493(1998)126,2061: Ebisuzaki, W., 2004: National Climatic Data Center data docu- JFFDFW.2.0.CO;2. mentation for NOAA Operational Model Archive and Dis- Hardy, K., and L. K. Comfort, 2015: Dynamic decision processes in tribution System (NOMADS) North American Regional complex, high-risk operations: The Yarnell Hill Fire, June 30, Reanalysis (NARR) ‘‘Merge’’ data set. NOAA Tech. Doc. 2013. Saf. Sci., 71, 39–47, doi:10.1016/j.ssci.2014.04.019. DSI-6175, 11 pp. [Available online at https://nomads.ncdc. Higgins, R. W., Y. Yao, and X. L. Wang, 1997: Influence of the noaa.gov/docs/ncdc-narrdsi-6175-final.pdf.] North American monsoon system on the U.S. summer pre- Evans, J. S., and C. A. Doswell, 2001: Examination of derecho cipitation regime. J. Climate, 10, 2600–2622, doi:10.1175/ environments using proximity soundings. Wea. Fore- 1520-0442(1997)010,2600:IOTNAM.2.0.CO;2. casting, 16,329–342,doi:10.1175/1520-0434(2001)016,0329: Holton, J. R., 2004: An Introduction to Dynamic Meteorology. 4th EODEUP.2.0.CO;2. ed. Vol. 88. Academic Press, 535 pp. Evans, M., J. Constantino, B. Lambert, and R. Grumm, 2012: A pre- Jenkins, M. A., 2004: Investigating the Haines Index using parcel liminary study of inverted-V soundings and downstream severe model theory. Int. J. Wildland Fire, 13, 297–309, doi:10.1071/ weather in New York and Pennsylvania. Natl. Wea. Dig., 36, 1–20. WF03055. Field, R. D., and Coauthors, 2015: Development of a Global Fire Johnson, R. H., P. E. Ciesielski, B. D. McNoldy, P. J. Rogers, and Weather Database. Nat. Hazards Earth Syst. Sci., 15, 1407– R. K. Taft, 2007: Multiscale variability of the flow during the 1423, doi:10.5194/nhess-15-1407-2015. North American Monsoon Experiment. J. Climate, 20, 1628– Fromm, M., and R. Servranckx, 2003: Transport of forest fire 1648, doi:10.1175/JCLI4087.1. smoke above the tropopause by supercell convection. Geo- ——, R. S. Schumacher, J. H. Ruppert Jr., D. T. Lindsey, J. E. phys. Res. Lett., 30, 1542, doi:10.1029/2002GL016820. Ruthford, and L. Kriederman, 2014: The role of convective ——, and Coauthors, 2000: Observations of boreal forest fire outflow in the Waldo Canyon Fire. Mon. Wea. Rev., 142, 3061– smoke in the stratosphere by POAM III, SAGE II, and lidar 3080, doi:10.1175/MWR-D-13-00361.1. in 1998. Geophys. Res. Lett., 27, 1407–1410, doi:10.1029/ Kaufman, Y. J., and Coauthors, 1998: Potential global fire moni- 1999GL011200. toring from EOS-MODIS. J. Geophys. Res., 103, 32 215– ——, R. Bevilacqua, R. Servranckx, J. Rosen, J. P. Thayer, 32 238, doi:10.1029/98JD01644. J. Herman, and D. Larko, 2005: Pyro-cumulonimbus injection Knupp, K. R., and Coauthors, 2014: Meteorological overview of the of smoke to the stratosphere: Observations and impact of a devastating 27 April 2011 tornado outbreak. Bull. Amer. Me- super blowup in northwestern Canada on 3–4 August 1998. teor. Soc., 95, 1041–1062, doi:10.1175/BAMS-D-11-00229.1. J. Geophys. Res., 110, D08205, doi:10.1029/2004JD005350. Lareau, N. P., and C. B. Clements, 2016: Environmental controls ——, A. Tupper, D. Rosenfeld, R. Servranckx, and R. McRae, on pyrocumulus and pyrocumulonimbus initiation and devel- 2006: Violent pyro-convective storm devastates Australia’s opment. Atmos. Chem. Phys., 16, 4005–4022, doi:10.5194/ capital and pollutes the stratosphere. Geophys. Res. Lett., 33, acp-16-4005-2016. L05815, doi:10.1029/2005GL025161. Lindsey, D. T., and M. Fromm, 2008: Evidence of the cloud lifetime ——, and Coauthors, 2008a: Stratospheric impact of the Chis- effect from wildfire-induced thunderstorms. Geophys. Res. holm pyrocumulonimbus eruption: 1. Earth-viewing satel- Lett., 35, L22809, doi:10.1029/2008GL035680. lite perspective. J. Geophys. Res., 113, D08202, doi:10.1029/ Luderer, G., J. Trentmann, and M. O. Andreae, 2009: A new look 2007JD009153. at the role of fire-released moisture on the dynamics of at- ——, and Coauthors, 2008b: Stratospheric impact of the Chisholm mospheric pyro-convection. Int. J. Wildland Fire, 18, 554–562, pyrocumulonimbus eruption: 2. Vertical profile perspective. doi:10.1071/WF07035. J. Geophys. Res., 113, D08203, doi:10.1029/2007JD009147. Macias Fauria, M., and E. A. Johnson, 2006: Large-scale climatic ——, D. T. Lindsey, R. Servranckx, G. Yue, T. Trickl, R. Sica, patterns control large lightning fire occurrence in Canada and P. Doucet, and S. Godin-Beekmann, 2010: The untold story of Alaska forest regions. J. Geophys. Res., 111, G04008, pyrocumulonimbus. Bull. Amer. Meteor. Soc., 91, 1193–1209, doi:10.1029/2006JG000181. doi:10.1175/2010BAMS3004.1. McRae, R. H. D., and J. J. Sharples, 2014: Forecasting conditions ——, R. H. D. McRae, J. J. Sharples, and G. P. Kablick III, 2012: conducive to blow-up fire events. CAWCR Research Letters,No. Pyrocumulonimbus pair in Wollemi and Blue Mountains 11, Center for Australian Weather and Climate Research, Mel- National Parks, 22 November 2006. Aust. Meteor. Oceanogr. bourne, Australia, 14–19. [Available online at http://www.cawcr. J., 62, 117–126, doi:10.22499/2.6203.001. gov.au/researchletters/CAWCR_Research_Letters_11.pdf.] Giglio, L., 2007: Characterization of the tropical diurnal fire cycle ——, ——, S. R. Wilkes, and A. Walker, 2013: An Australian pyro- using VIRS and MODIS observations. Remote Sens. Environ., tornadogenesis event. Nat. Hazards, 65, 1801–1811, doi:10.1007/ 108, 407–421, doi:10.1016/j.rse.2006.11.018. s11069-012-0443-7. ——, 2010: MODIS collection 5 active fire product user’s guide ——, ——, and M. Fromm, 2015: Linking local wildfire dynamics to version 2.4. Department of Geography, University of Mary- pyroCb development. Nat. Hazards Earth Syst. Sci., 15, 417– land, 61 pp. 428, doi:10.5194/nhess-15-417-2015. ——, J. Descloitres, C. O. Justice, and Y. J. Kaufman, 2003: Mesinger, F., and Coauthors, 2006: North American Regional An enhanced contextual fire detection algorithm for Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/ MODIS. Remote Sens. Environ., 87, 273–282, doi:10.1016/ BAMS-87-3-343. S0034-4257(03)00184-6. Mills, G. A., and L. McCaw, 2010: Atmospheric stability environ- Haines, D. A., 1988: A lower atmosphere severity index for wild- ments and fire weather in Australia—extending the Haines land fires. Natl. Wea. Dig., 13 (3), 23–27. Index. CAWCR Tech. Rep. 20, CAWCR, 151 pp. [Available Hamilton, D. W., Y. L. Lin, R. P. Weglarz, and M. L. Kaplan, online at http://www.cawcr.gov.au/technical-reports/CTR_ 1998: Jetlet formation from diabatic forcing with applications 020.pdf.]

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC 2254 MONTHLY WEATHER REVIEW VOLUME 145

Milne, R., 2004: A modified total totals index for thunderstorm 1991. J. Geophys. Res., 99, 16 719–16 735, doi:10.1029/ potential over the Intermountain West. NOAA/NWS West- 94JD01208. ern Region Tech. Attach. 04-04, 15 pp. [Available online ——, J. M. Feltz, W. P. Menzel, and D. E. Ward, 1998: An overview at http://www.wrh.noaa.gov/media/wrh/online_publications/ of GOES-8 diurnal fire and smoke results for SCAR-B and TAs/ta0404.pdf.] 1995 fire season in South America. J. Geophys. Res., 103, Mu, M., and Coauthors, 2011: Daily and 3-hourly variability in 31 821–31 835, doi:10.1029/98JD01720. global fire emissions and consequences for atmospheric model Reutter, P., and Coauthors, 2014: 3-D model simulations of dy- predictions of carbon monoxide. J. Geophys. Res., 116, namical and microphysical interactions in pyroconvective D24303, doi:10.1029/2011JD016245. clouds under idealized conditions. Atmos. Chem. Phys., 14, Nauslar, N. J., M. L. Kaplan, J. Wallmann, and T. J. Brown, 2013: A 7573–7583, doi:10.5194/acp-14-7573-2014. forecast procedure for dry thunderstorms. J. Oper. Meteor., 1, Rex, D. F., 1950a: Blocking action in the middle troposphere and its 200–214, doi:10.15191/nwajom.2013.0117. effect upon regional climate I. An aerological study of blocking Nimchuk, N., 1983: Wildfire behavior associated with upper ridge action. Tellus, 2, 196–211, doi:10.3402/tellusa.v2i3.8546. breakdown. Alberta Energy and Natural Resources Forest ——, 1950b: Blocking action in the middle troposphere and its effect Service, Rep. T/50, 45 pp. upon regional climate II. The climatology of blocking action. Peterson, D., and J. Wang, 2013: A sub-pixel-based calculation of Tellus, 2,275–301,doi:10.1111/j.2153-3490.1950.tb00339.x. fire radiative power from MODIS observations: 2. Sensitivity Robock, A., 2000: Volcanic eruptions and climate. Rev. Geophys., analysis and potential fire weather application. Remote Sens. 38, 191–219, doi:10.1029/1998RG000054. Environ., 129, 231–249, doi:10.1016/j.rse.2012.10.020. Rorig, M. L., and S. A. Ferguson, 1999: Characteristics of lightning ——, ——, C. Ichoku, and L. A. Remer, 2010: Effects of lightning and wildland fire ignition in the Pacific Northwest. J. Appl. and other meteorological factors on fire activity in the North Meteor., 38, 1565–1575, doi:10.1175/1520-0450(1999)038,1565: American boreal forest: Implications for fire weather fore- COLAWF.2.0.CO;2. casting. Atmos. Chem. Phys., 10, 6873–6888, doi:10.5194/ Rosenfeld, D., M. Fromm, J. Trentmann, G. Luderer, M. O. acp-10-6873-2010. Andreae, and R. Servranckx, 2007: The Chisholm firestorm: ——, E. Hyer, and J. Wang, 2013a: A short-term predictor of Observed microstructure, precipitation and lightning activity satellite-observed fire activity in the North American bo- of a pyro-cumulonimbus. Atmos. Chem. Phys., 7, 645–659, real forest: Toward improving the prediction of smoke doi:10.5194/acp-7-645-2007. emissions. Atmos. Environ., 71, 304–310, doi:10.1016/ Rothermel, R. C., 1991: Predicting behavior and size of crown fires j.atmosenv.2013.01.052. in the northern Rocky Mountains. USDA Forest Service, In- ——, J. Wang, C. Ichoku, E. Hyer, and V. Ambrosia, 2013b: A sub- termountain Research Station, Research Paper INT-438, pixel-based calculation of fire radiative power from MODIS 46 pp. [Available online at http://digitalcommons.usu.edu/ observations: 1 Algorithm development and initial assess- cgi/viewcontent.cgi?article51064&context5barkbeetles.] ment. Remote Sens. Environ., 129, 262–279, doi:10.1016/ Sassen, K., and J. R. Campbell, 2001: A midlatitude cirrus cloud j.rse.2012.10.036. climatology from the facility for atmospheric remote sens- ——, E. Hyer, and J. Wang, 2014: Quantifying the potential for ing. Part I: Macrophysical and synoptic properties. J. Atmos. high-altitude smoke injection in the North American boreal Sci., 58, 481–496, doi:10.1175/1520-0469(2001)058,0481: forest using the standard MODIS fire products and subpixel- AMCCCF.2.0.CO;2. based methods. J. Geophys. Res. Atmos., 119, 3401–3419, Schroeder, W., I. Csiszar, L. Giglio, and C. C. Schmidt, 2010: On doi:10.1002/2013JD021067. the use of fire radiative power, area, and temperature esti- ——, E. J. Hyer, J. R. Campbell, M. D. Fromm, J. W. Hair, C. F. mates to characterize biomass burning via moderate to Butler, and M. A. Fenn, 2015: The 2013 Rim Fire: Implications coarse spatial resolution remote sensing data in the Brazil- for predicting extreme fire spread, pyroconvection, and smoke ian Amazon. J. Geophys. Res., 115, D21121, doi:10.1029/ emissions. Bull. Amer. Meteor. Soc., 96, 229–247, doi:10.1175/ 2009JD013769. BAMS-D-14-00060.1. Skinner, W. R., B. J. Stocks, D. L. Martell, B. Bonsal, and ——, M. D. Fromm, J. E. Solbrig, E. J. Hyer, M. L. Surratt, and A. Shabbar, 1999: The association between circulation J. R. Campbell, 2017: Detection and inventory of intense py- anomalies in the mid-troposphere and area burned by wild- roconvection in western North America using GOES-15 land fire in Canada. Theor. Appl. Climatol., 63, 89–105, daytime infrared data. J. Appl. Meteor. Climatol., 56, 471–493, doi:10.1007/s007040050095. doi:10.1175/JAMC-D-16-0226.1. ——, and Coauthors, 2002: A 500 hPa synoptic wildland fire cli- Potter, B. E., 2005: The role of released moisture in the atmo- matology for large Canadian forest fires, 1959-1996. Theor. spheric dynamics associated with wildland fires. Int. Appl. Climatol., 71, 157–169, doi:10.1007/s007040200002. J. Wildland Fire, 14, 77–84, doi:10.1071/WF04045. Tory, K. J., and W. Thurston, 2015: Pyrocumulonimbus: A litera- ——, 2012: Atmospheric interactions with wildland fire behaviour— ture review. Australian Bureau of Meteorology, Bushfire and I. Basic surface interactions, vertical profiles and synoptic Natural Hazards CRC, Rep. 2015.067, 13 pp. [Available online structures. Int. J. Wildland Fire, 21,779–801,doi:10.1071/ at http://www.bnhcrc.com.au/publications/biblio/bnh-1882.] WF11128. Trentmann, J., and Coauthors, 2006: Modeling of biomass smoke ——, J. A. Winkler, D. F. Wilhelm, and R. P. Shadbolt, 2008: injection into the lower stratosphere by a large forest fire (Part Computing the low-elevation variant of the Haines index for I): Reference simulation. Atmos. Chem. Phys., 6, 5247–5260, fire weather forecasts. Wea. Forecasting, 23, 159–167, doi:10.5194/acp-6-5247-2006. doi:10.1175/2007WAF2007025.1. Val Martin, M., R. A. Kahn, J. A. Logan, R. Paugam, M. Wooster, Prins, E. M., and W. P. Menzel, 1994: Trends in South-American and C. Ichoku, 2012: Space-based observational constraints biomass burning detected with the GOES Visible Infrared for 1-D fire smoke plume-rise models. J. Geophys. Res., 117, Spin Scan Radiometer Atmospheric Sounder from 1983 to D22204, doi:10.1029/2012JD018370.

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC JUNE 2017 P E T E R S O N E T A L . 2255

Van Wagner, C. E., 1987: Development and structure of the Ca- ——, and Coauthors, 2011: Synthesis of knowledge of extreme fire nadian Forest Fire Weather Index System. Canadian Forestry behavior: Volume I for fire managers. USDA Forest Service, Service, Forestry Tech. Rep. 35, 35 pp. Pacific Northwest Research Station, General Tech. Rep. Wakimoto, R. M., 1985: Forecasting dry microburst activity over PNW-GTR-854, 144 pp. [Available online at http://www.fs. the High Plains. Mon. Wea. Rev., 113, 1131–1143, doi:10.1175/ fed.us/pnw/pubs/pnw_gtr854.pdf.] 1520-0493(1985)113,1131:FDMAOT.2.0.CO;2. Westphal, D. L., and O. B. Toon, 1991: Simulations of micro- Wallace, J. M., and P. V. Hobbs, 2006: Atmospheric Science: physical, radiative, and dynamic processes in a continental-scale An Introductory Survey. 2nd ed. Vol. 92. Academic Press, 483 pp. forest-fire smoke plume. J. Geophys. Res., 96, 22 379–22 400, Wallmann, J., R. Milne, C. Smallcomb, and M. Mehle, 2010: Using doi:10.1029/91JD01956. the 21 June 2008 California lightning outbreak to improve dry Wooster, M. J., G. Roberts, G. L. W. Perry, and Y. J. Kaufman, lightning forecast procedures. Wea. Forecasting, 25, 1447– 2005: Retrieval of biomass combustion rates and totals from 1462, doi:10.1175/2010WAF2222393.1. fire radiative power observations: FRP derivation and cali- Werth, P., and R. Ochoa, 1993: The evaluation of Idaho wildfire bration relationships between biomass consumption and fire growth using the Haines Index. Wea. Forecasting, 8, 223–234, radiative energy release. J. Geophys. Res., 110, D24311, doi:10.1175/1520-0434(1993)008,0223:TEOIWG.2.0.CO;2. doi:10.1029/2005JD006318.

Unauthenticated | Downloaded 10/07/21 01:35 PM UTC