Synoptic conditions associated with cool season post-fire debris flows in the Transverse Ranges of southern California

1.1 Post-fire debris flows

The Transverse Ranges of southern California feature a combination of steep and complex terrain, combustible fuels, a prolonged dry season, and strong wind events such as Santa Anas. These factors combined produce the most intense fire climate in the USA (Fig. 1; Wells 1981, 1987; Raphael 2003; Keeley et al. 2004). The Transverse Ranges are also prone to multi-year drought interspersed with wet weather, a combination conducive to growth and then desiccation of the region’s fire-prone chaparral vegetation.

Open image in new window

Wildfire has profound effects on storm runoff, erosion, and sedimentation in the Transverse Ranges. For several years following a fire, runoff rates can more than double due to alteration or removal of the vegetation and litter cover, soil-sealing translocation of minerals and ash, fire-induced degradation of soil and rock, and the development of water repellant soil conditions (DeBano 1981, 2000; Neary et al. 1999; Parise and Cannon 2012). Post-fire debris flows (PFDFs), the most severe runoff response to precipitation on burned watersheds, tend to occur in steep watershed areas burned at moderate to high severity, with the largest events often triggered by the first significant rainstorm (Cannon et al. 2008; Parise and Cannon 2012). PFDFs are a common threat to southern California communities (Eaton 1936; USGS 2005); since the early 1900s urbanization on alluvial fans and floodplains within and adjacent to the Transverse Ranges has resulted in loss of life and property associated with PFDFs (e.g., Chawner 1935; Eaton 1936; Troxell and Petersen 1937; Shuirman and Slosson 1992; Cannon et al. 2010).

Previous work has cited “intense convection” as the main cause of rainfall intensities sufficient for PFDFs (Slosson et al. 1991; Cannon et al. 2008, 2010; Moody et al. 2013), but provides few details as to broader scale conditions present when this intense convection occurs. One recent study provides an in-depth meteorological case study of an individual PFDF event in the western Santa Monica Mountains (Sukup et al. 2016). Absent, however, is a comprehensive examination of atmospheric conditions across multiple PFDF events.

Meteorological case studies were generated for each PFDF event date, and common features observed among events serve as an “ingredient list” for conditions conducive to PFDFs in the Transverse Ranges. While many of these ingredients may already be familiar to weather forecasters, they can utilize the analysis of conditions across a broad range of events to put forecast events in context and examine variability across events. This work facilitates non-meteorologist understanding of weather forecasts presented by NWS related to PFDFs, builds on past collaborative multidisciplinary work (NOAA-USGS Debris Flow Task Force 2005; Jorgensen et al. 2011), and provides a foundation for new research directions that cross the boundaries between meteorology, geology, and hydrology.

1.2 Meteorological conditions associated with intense precipitation in southern California

Southern California and the Transverse Ranges experience some of the highest storm precipitation totals in the nation, on par with totals seen in hurricanes in the southeastern United States and thunderstorms in the Midwest (Dettinger et al. 2011; Ralph and Dettinger 2012). The highest probable 1-h precipitation intensities in this region are on par with those seen in association with Midwest thunderstorms (NOAA HDSC 2017). At the synoptic (coarse) scale, mid-latitude cyclones are generally responsible for bringing cool season precipitation to California (Weaver 1962; Monteverdi 1995). These cyclones may vary in size, shape, intensity, and moisture transport; some may have associated atmospheric rivers or become closed lows, as described below. Additionally, many finer scale features and processes that go beyond the resolution of this synoptic-scale study are also at work to create convection “hotspots” that produce the short duration, high intensity precipitation conducive to PFDFs (Jorgensen et al. 2011; Moody et al. 2013). Some of the features considered in this study are atmospheric rivers, closed lows, orographic lift, and other types of lift.

Atmospheric rivers (ARs; Figs. 1c, 3d) are narrow corridors of high water vapor transport typically found in the lowest 2.5 km of the atmosphere (Zhu and Newell 1998; Ralph et al. 2004, 2005). ARs are found ahead of the cold front in mid-latitude cyclones and source their moisture from the tropics and extratropics (Browning and Pardoe 1973; Ralph et al. 2004). They are typically <1000 km in width, >2000 km in length, and have integrated water vapor (IWV; specific humidity integrated over a vertical column) values exceeding 20 mm (Ralph et al. 2004; Neiman et al. 2008). Additionally, integrated water vapor transport (IVT; the product of specific humidity and wind integrated over a vertical column) exceeding 250 kg m⁻¹ s⁻¹ is a criteria of ARs that accounts for the importance of wind velocity in vapor transport, upslope vapor flux, and precipitation when the AR encounters terrain (Moore et al. 2012; Rutz et al. 2014). In southern California, ARs are most abundant in the cool season (November–April; Neiman et al. 2008; Dettinger et al. 2011) and account for roughly 40–50% of cool season precipitation (Dettinger et al. 2011; Ralph et al. 2013; Rutz et al. 2014). ARs are associated with most of the area’s extreme precipitation events (Dettinger et al. 2011; Ralph and Dettinger 2012) and have been found to produce, on average, twice the precipitation of winter storms without ARs (Neiman et al. 2008). ARs feature low-level jets (LLJs), strong winds in the lowest 2 km of the atmosphere (Browning and Pardoe 1973; Ralph et al. 2005). LLJs impacting coastal California vary in strength, from >12.5 m s⁻¹ in Ralph and Dettinger (2012) to >20 m s⁻¹ in Ralph et al. (2005). The presence and strength of a LLJ can help dictate precipitation intensity in complex terrain, with stronger LLJs producing grater upslope flux and enhanced precipitation (Ralph et al. 2006).

Closed lows (CLs) are a subset of mid-latitude cyclones that are frequently observed over California in the cool season months. They have closed height contours and complete cyclonic (counterclockwise) flow around their centers at mid- to upper levels of the atmosphere. These properties help to impede a closed low’s downstream motion such that CLs, often in concert with other features like ARs, can produce sustained precipitation (Oakley and Redmond 2014).

Tarleton and Kluck (1994) cite strong orographic forcing as one of the reasons a large concentration of major California precipitation events occur in the Transverse Ranges. Orographic precipitation occurs when moist air is forced to ascend a terrain barrier. As the moist air rises and cools, condensation and ultimately precipitation occur. Lin et al. (2001) define five common ingredients for intense orographic precipitation: (1) a conditionally or potentially unstable air mass—an air mass that, if forced to ascend, will continue to do so, (2) presence of a low-level wind speed maximum containing moist air oriented orthogonal to the mountain barrier, (3) presence of a steep mountain, (4) a slow-moving weather system, and (5) high precipitation efficiency, a quantity related to the vertical flux of moisture, horizontal length of storm, and propagation speed of storm. During an AR impinging on the Transverse Ranges and its associated LLJ, these conditions are often met, resulting in heavy precipitation (Neiman et al. 2002, 2004; Ralph et al. 2006).

Other types of forced ascent beyond orographic lift contribute to convective cells that trigger PFDFs in the Transverse Ranges as well. Along a cold front, air may be forced to ascend as an incoming cold, dense air mass forces it upward. This can result in the formation of a narrow cold frontal rain band, a line of intense convective cells parallel to the cold front (Hobbs 1978; Hobbs and Persson 1982). Upper level (above 300 hPa) jet structure may also contribute to the development of convective cells. There are locations in the jet structure that produce divergence at upper levels, favoring lift (Carlson 1998). Isolated thunderstorms in this region, while uncommon in the cool season, are occasionally observed during the boundary months of the season. These storms occur on a spatial scale of tens of kilometers, thus producing precipitation over a much smaller area than that affected by a mid-latitude cyclone.

2.1 Compilation of PFDF catalog

A catalog of post-fire debris flow events in the Transverse Ranges and relevant details was compiled for the period 1980–2014. This range is based on the availability and qualities of PFDF events as well as availability of moderate resolution meteorological information from the North American Regional Reanalysis (NARR; Mesinger et al. 2006) used in generating case studies. The catalog was compiled using a variety of publications, U.S. Geological Survey reports, and newspaper articles as noted in Table 1. The requirements for a PFDF event to be included in the database are:

• Occurred within two years of a fire

• Identified as a PFDF in scientific literature

• Time of event triggering rainfall was either known, or could be determined

• Occurred in cool season (November–April)

• Event is generally of note and familiar to the PFDF community

Based on these criteria, an original compilation of several hundred PFDF events was refined, with the resulting catalog containing 93 individual PFDFs occurring as part of 21 “events” (multiple PFDFs in a burned area) on 19 distinct dates. The events occurred in 12 burn areas in the Transverse Ranges, as shown in Fig. 2.

Open image in new window

2.2 Timing of PFDF events

To assess meteorological conditions associated with the occurrence of a PFDF, it was necessary to assign a trigger time to each event. This posed a challenge as time of occurrence is rarely provided in the PFDF literature due to lack of observations. Indeed, Guzzetti et al. (2008) found that globally, only 5.1% of the 2626 published landslide and debris flow events had timing data accurate to ±12 h (Staley et al. 2013).

2.3 Connecting PFDF events to meteorology

3.1 Synoptic scale features

Consistent with the dominance of AR events, compositing of IWV shows the upper three quartiles of grid points at the time of event exceed the 20 mm AR threshold (Fig. 4). Averaged across events, IWV values were in the 92nd percentile with respect to climatology (“Appendix”). For the IVT variable, the 250 kg m⁻¹ s⁻¹ AR threshold fell in the lower third of the lower middle quartile of the distribution at the time of event (Fig. 4). All six non-AR events had a majority of their 12 grid points at the time of event below IWV and IVT thresholds for ARs. Averaged across events, IVT values were in the 95th percentile with respect to climatology (“Appendix”). In some cases, IVT or IWV at event-3 h or event-6 h were higher than at event time (Fig. 4). An AR is typically located in the warm sector of a storm, preceding the cold front. Since many of the events exhibit lift associated with the cold front, it is possible to see convection capable of initiating a PFDF occur following the maximum values of IVT or IWV in a storm. Averaged across events, 75% of the moisture flux (product of specific humidity and wind speed) was located below 600 hPa (~4 km; Fig. 6b). This is higher in the atmosphere than in previous studies of ARs off the California coast, where 75% of moisture flux was observed below 2.25 km (Ralph et al. 2005).

Open image in new window

Most events featured a neutral or negatively tilted trough (as seen in Fig. 3a, b; “Appendix”). In the case of a negatively tilted trough, instability and convection are favorable, as cold air advection occurs at upper levels above relatively warm air at low levels (MacDonald 1976). Instability can still occur within a neutral and positively tilted trough as well.

3.2 Jet position, structure, and winds

At the NARR event time, the dominant direction of the 300 hPa upper level flow over the composited study areas was southwest to west-southwest (Fig. 5, top; “Appendix”). All observations fell between 185° and 285°, with 90% of observations falling between 215° and 275°. In a majority of observations (67%), the average speed of the 300 hPa flow over the area of interest was ≥40 m s⁻¹, indicative of a weak of a weak to moderate flow aloft. Several of the events examined see 300 hPa wind speeds in excess of 50 m s⁻¹, indicating moderate to strong flow. The velocity of the upper level winds indicates the strength of the upper level divergence, which promotes upward vertical motions and potential for precipitation (Clark et al. 2009; O’Hara et al. 2009).

Open image in new window

The position of the upper level jet was to the south of the region of interest in a majority of events (68%; Table 2), typically placing the Transverse Ranges in the curved jet exit and/or the left exit of a jet streak, an area associated with strong upper level divergence and lift, as shown in Fig. 3a. This is consistent with the “southerly displaced jet stream” cited by Tarleton and Kluck (1994) as a typical feature in extreme precipitation events in California. In several cases (26%), the upper level jet was directly over the study area, and finer scale splitting within the jet developed entrance or exit regions over the study area favorable for upper level divergence. In only one case, the upper level jet was located to the north of the region, though was positioned such that the right entrance of the jet was over the Transverse Ranges.

At mid-levels (700 hPa), winds were predominantly from the southwest at the time of the event, with 69% of observations falling between 205° and 245° (Fig. 5, middle row). The predominant speed was 15–20 m s⁻¹ (38%); and 37% of observations were greater than 20 m s⁻¹. From NARR time-6 h through NARR time of event, wind speeds increased and direction became more uniformly from the southwest.

At low levels (925 hPa), wind direction was predominantly from 155° to 215° (60% of observations; Fig. 5 bottom row and “Appendix”). The dominant speed was 5–10 m s⁻¹ (49%), with 8% of observations exceeding 15 m s⁻¹. This is significant in that the low-level moderate intensity southerly winds are orthogonal to the east-to-west oriented Transverse Ranges, providing one of the necessary conditions for heavy orographic precipitation (Lin et al. 2001).

Wind profiler data (NOAA ESRL 2016; as in Fig. 3e) were used to diagnose the presence of LLJs for events when data were available. Profiler data were available for the ten post-2005 events, though available locations were inconsistent. Profiler data confirmed the presence of LLJs in the seven of these events; six of which were AR events. No LLJ was detected in three of the four non-AR events in this period (“Appendix”).

3.3 Atmospheric stability

Stability profiles below ~3 km (700 hPa) as observed in vertical profiles could be broadly divided into weakly unstable, moist neutral, or unstable near surface becoming moist neutral with height (Table 2; “Appendix”). A large majority of the events showed moist-neutral stability either in all levels below 3 km or in the 1–2+ km layer (Fig. 6a). Moist-neutral stability is recognized as little or no change in θ _e* with height \(\left({{\partial \theta_{e*} /\partial {{z}}}} = 0\right)\). The significance of moist neutrality is that if the parcel is displaced upward, it will maintain its new position. Air parcels in a neutral setting can be forced to ascend relatively easily in the presence of a lifting mechanism such as a cold front or mountain barrier, resulting in convection. Moist-neutral stability is a common feature of ARs (Ralph et al. 2005). Several events saw instability at low levels (\({{\partial \theta_{e*} /\partial {{z}}}} \, <\, 0\); Fig. 6a) transitioning to moist neutral near 1 km, distinct from the moist-neutral layer observed from the surface to 2.8 km in a composite of AR events presented by Ralph et al. (2005).

Open image in new window

The scale for CAPE begins at 0, and higher values of CAPE indicate greater instability and severe weather. In the CAPE climatologies created for periods relative to the PFDF events studied here, a CAPE value of 100 J kg⁻¹ was on average the 85th percentile for the maximum values in the study area (“Appendix”). At the NARR event time, median NARR CAPE was 20 J kg⁻¹ and ranged from 0 to 1330 J kg⁻¹, with the values exceeding 500 J kg⁻¹ all coming from grid points associated with the two Springs Fire cases in the Santa Monica Mountains. At the time closest to the event, median CAPE from the VBG soundings was 40 J kg⁻¹ and ranged from 0 to 463 J kg⁻¹ (Fig. 7; “Appendix”). A general trend of increasing CAPE was observed in NARR in the 3-h and 6-h time steps leading up to the event (Fig. 7).

Open image in new window

3.4 Analysis of radar imagery

Radar imagery was available for 14 unique PFDF event dates through NCEI’s Radar Data Map (https://gis.ncdc.noaa.gov/maps/ncei/radar). All events had radar returns of at a minimum 50 dBZ (approximately >48 mm h⁻¹; as in Fig. 3f),¹ indicative of heavy precipitation. Six of the 14 events had very strong returns over 60 dBZ (>200 mm h⁻¹), indicative of very intense precipitation. Narrow cold frontal rain bands (NCFRs) were identified in five of the 14 events (Table 2; “Appendix”; example in Fig. 8). NCFRs show up in radar imagery as a narrow band of high radar reflectivity on the order of 200 km long with breaks and gaps along their length (Fig. 8; Jorgensen et al. 2003). The five NCFRs in this study all followed a similar west-to-east path across the southern California Bight with the northern half of the NCFR situated over land, while the southern half was over water and the feature’s long axis perpendicular to the coast. This preferential orientation likely occurs due to blocking and modification of the low-level front by coastal terrain (Neiman et al. 2004; Hughes et al. 2009).

Open image in new window

Radar imagery associated with one PFDF event on 12 November 2009 resulted from the development of a very isolated convective cell in the San Gabriel Mountains, reminiscent of a warm season thunderstorm event. The remaining eight of the 14 PFDF events for which radar data were available featured other types of convective activity such as orographic forcing and mesoscale rain bands (as in Fig. 3f) that are not discussed in detail within this paper.

3.5 Tools and applications

The prominence of ARs and their related features in the occurrence of PFDFs in the Transverse Ranges suggests that those concerned with PFDFs will benefit from incorporating the use of online AR forecast and diagnostic tools into their decision-making. One such tool is the US West Coast AR Landfall Tool available through the Center for Western Weather and Water Extremes (CW3E) at http://cw3e.ucsd.edu/?page_id=491. This tool provides a 16-day forecast where the user can see the probability, magnitude, location, and timing of AR conditions arriving along the West Coast as well as how AR conditions vary in a forecast model through time. Proper use of this tool among others can generate awareness of the potential for a PFDF and support planning and decision-making in both research and emergency response. Additionally, the IVT variable assessed in the aforementioned AR forecasting tool has been shown to be more successful in long-range forecasts than precipitation (Lavers et al. 2016). Thus, forecasts of IVT can be used to provide forecasts of likelihood for heavy rainfall with greater certainty further ahead than the traditional precipitation forecasts.

3.6 Limitations and future work

Limitations of this study lie in NARR’s 32 km resolution and thus its inability to resolve fine-scale processes important to the development of convective cells such as blocking, barrier jet features, and low-level convergence along terrain barriers that are common in the region (Small 1999; Neiman et al. 2004; Hughes et al. 2009). Additionally, Hughes et al. (2012) have noted challenges in how NARR represents low-level winds and winds at the land-sea boundary, which may impact results for winds at these levels. The coarseness of the NARR data may also impact the accuracy of estimates of stability. CAPE, for example, can vary greatly in the course of a storm event. Sukup et al. (2016) show a significant increase in CAPE following a frontal passage after the PFDF had already occurred during the 12 December 2014 PFDF event. Thus, it is possible that with the spatial and temporal coarseness of both NARR and radiosonde data stability variables assessed are not representative of the true event time and may be biased high or low based on times available.

While AR conditions make up a majority of cool season PFDF events in this study, there are, on occasion, isolated thunderstorm events such as the 12 November 2009 PFDF event included herein. Thus, we advise that all storm types should be considered in emergency preparedness; however, advantage should be taken of recent advancements in AR detection and prediction given the dominance of ARs among the PFDF cases explored in this study.

This work is, in essence, an analysis of cases of intense precipitation in the Transverse Ranges, subset by PFDF occurrence. A broader approach would be to look at all precipitation events over a particular threshold in the region. However, PFDF thresholds have been noted to vary in space (Staley et al. 2016) thus choosing one representative threshold may not suffice. Focusing on events known to produce impactful PFDFs ensures precipitation was indeed sufficient. Additionally, applying meteorological analyses to verified impactful PFDF events allows us to make a direct connection with the experiences and concerns of our target audience in a way that a more abstract approach of exploring precipitation over a particular intensity may not.

The lack of observations of both post-fire debris flow activity and precipitation limits the assessment of null events. Without a high-resolution network of gauges and instruments that can record debris flow response and triggering rainfall within the burn area, it is difficult to determine whether precipitation of sufficient intensity and duration for PFDF activity did or did not fall on a burn area. In cases outside of research efforts that utilize instrumentation, PFDFs are often only noted if they impact human infrastructure. Thus, if a PFDF occurred in an inaccessible remote area of a watershed, it may not be documented. In the null case most relevant to this work (all favorable synoptic-scale conditions present, but intense precipitation does not occur, field observations made, and no PFDFs present), analysis using mesoscale modeling would be needed to assess why intense cells did not develop, which is beyond the scope of this work. Thus, this work focuses on well-documented damaging debris flows that affected structures or infrastructure downstream of the burned watershed.

A major challenge that remains in this research topic is identifying the exact timing and location where intense convective cells might develop (isolated or within a larger storm system). Similar to the suggestions of Moody et al. (2013) and Shakesby et al. (2016), we propose future work should focus on high-resolution (≤1 km) modeling of the region to identify favored areas for intense convection under a variety of flow regimes. For modeling efforts to be successful, precise timing of a greater number of PFDF events is necessary. There are many challenges to overcome in instrumenting basins, as described in Kean et al. (2011) and Staley et al. (2013), but, where present, we have found this timing data essential to assessment of the meteorological component of PFDF events.