On July 1, 2025, the City of Berkeley finally published the evacuation study that had been planned since before the COVID times, but pushed back by actual events. The study was completed by a contracting group specializing in evacuation planning, KLD Associates. From what we have been able to see, they have a solid reputation in the business.
The study modeled both wildfire and tsunami evacuations, and a large number of evacuation scenarios. Our notes focus on the wildfire evacuation analysis, both from the North Berkeley Hills and from Panoramic Hill. For the sake of brevity we will discuss findings in this post, but will provide links to deeper analysis of every step.
The study was heavily dependent upon a Berkeley demographic survey, and upon a good number of mathematical assumptions-–the model’s capabilities also limited some of the conclusions. It discussed and listed the highest capacity roadways in Berkeley, but insisted on the fact that they should not be the exclusive paths for evacuation due to capacity issues.
Findings and conclusions
The key metric that the study predicts through the DYNEV II model is Evacuation Time Estimates, ETE.
Table 7-3 Legend: R02 = 1923 Fire, immediate evacuation; R04 - Panoramic Hill Fire
The study discusses in depth the way to derive the ETE. It makes clear that, while congestion and roadway capacity are important problems, the actual speed of mobilization for all evacuees is as much a determinant for ETE as is congestion—in particular when needing to deal with school children.
What we think are the more critical findings:
Evacuating the North Berkeley Hills—it can takes up to 4 hours 35 to evacuate the North Berkeley Hills, while the 1923 Fire reached Shattuck Avenue in 2 hours and 20 minutes. This result makes it crystal clear that limiting evacuation deaths requires most residents to evacuate upon an Extreme Fire Weather alert.
Evacuating Panoramic Hill—it can take more than 3 hours to evacuate Panoramic Hill, although much of that is not due to congestion but to the speed of mobilisation of the evacuees.
Phased vs. mass evacuation of the North Berkeley Hills—surprisingly, it is faster to evacuate the Berkeley Hills all at once, without distinction of evacuation zone.
Evacuate on Extreme Fire Weather alert—the study finds significant evacuation time advantage to evacuating on Extreme Fire Weather alert (up to 68%) and recommends it. This could save many lives, and it convinced me.
Parking restrictions in the Berkeley Hills—the study recommends against general red curbing as a whole, but recommends it in specific avenues of ingress. We find that the limits of the model do not allow it to evaluate 2-way traffic, which impairs the ability of the model to deliver a strong conclusion on this topic.
Carpooling—the study strongly recommends carpooling for evacuees, in order to decrease traffic congestion. We find it difficult to recommend to anyone that they should put their life into someone else’s hands and count on a neighbor being certainly able to evacuate them: it would have to be a choice of opportunity, but not one that may be relied upon as a sure way to evacuate.
Alternative routing—the study recommends that drivers look at alternative routes as much as the high-capacity axes. This is also a difficult recommendation to endorse if no red curbing is done, since it is exceedingly easy for many side streets in the Berkeley Hills to get blocked, even sometimes simply by a poorly parked car.
Evacuating on foot or by bike—the study recommends against foot or bike evacuations for wildfire evacuation in the Hills. There may come a time, however, when there are only bad choices for us in the Hills.
Traffic calming devices—the study finds traffic calming to slow down evacuation, and suggests some high-tech alternatives. Realistically, we do not think that the city’s budget will allow for it.
There are a number of other findings, that we find less germane to our tactical view of evacuation over the next three years. The full set of findings may be found here.
A longer discussion (which was already linked to earlier) of each topic may be found below. We intend to come back to several of these topics in the near future with further discussion and analysis. Do also look at the discussions provided in other local websites that are listed under “Notable articles.”
The study was based on a very large number of assumptions, that all appear to us to be reasonable, although we have one caveat that we will mention further down.
A source of many of these assumptions was "an online demographic survey [that] was conducted of the people living and working in Berkeley; 1,453 households responded to the survey, which corresponds to a sampling error of ±2.5% at the 95% confidence level based on the 2020 Census household data. The survey gathered demographic information (average household size, vehicle ownership, etc.), behavioral responses (would evacuees follow evacuation instructions issued by local officials), and time to complete mobilization activities.
Key takeaways include:
Community members could take as long as 3.5 hours to start their evacuation trip
(“mobilize”). Longer mobilization times lead to longer evacuation times overall (Section
7). Populations with the shortest mobilization times are employees and visitors. The population with the longest mobilization times is households with commuters who plan to pick up children from school in an emergency. See Section 5 (Table 5-9) and Appendix D.
Approximately 22% of households have children who attend school in Berkeley. In an
emergency, 88.2% of those households plan to pick up their children or send a trusted guardian to pick up their children at school. This behavior is expected to increase traffic congestion immediately around schools and in evacuating areas more broadly.
Automobile ownership: The average number of automobiles available per household is 1.55. 73.6% of households would use 1 vehicle for evacuation; 14.6% plan to use 2. Approximately 97% of households of 2 or more people have access to at least one vehicle. 7.5% of all households do not have access to a vehicle. See Appendix D.
Citywide, 12,870 residents do not have direct access to a vehicle and will require
transportation assistance. However, approximately eighty-seven percent (87%) of Berkeley’s transit-dependent population will rideshare with a neighbor or friend. This
leaves 1,686 (1.4%) of residents citywide who will rely on transportation assistance from the government. (See Table 3-10 and Appendix D.)
Six percent of households indicated that someone in the household would need help from someone outside the household (caretaker, personal attendant) to prepare to evacuate or to get to a vehicle. Some will also require specialized transportation assistance: citywide, approximately 1.6% of people will require a wheelchair accessible van, 0.2% require an ambulance and 0.4% require other modes of transportation. The remaining 3.8% do not require a specialized vehicle to evacuate.
Citywide 2.2% of Berkeley residents, employees, and college students plan to evacuate on foot and 1.4% plan to evacuate on bicycle. (Appendix D)
During an emergency evacuation, 1% of people would disregard evacuation orders and stay in place. (See Appendix D)
Emergency officials may notify people outside a hazard area that they are not in danger, and that they are requested to shelter in place (not evacuate). Ninety percent (90%) of households would comply with this request and the remaining 10% would choose to evacuate the area. (See Appendix D).
Eighty-nine percent (89%) of emergency evacuees plan to evacuate to a friend or relative’s home, a hotel, motel, campground, short-term rental (ex. AirBnB/VRBO), or second/seasonal home. Approximately 6% plan to evacuate to an evacuation shelter. See Figure D-12 for complete results.
Approximately 50% of households have pets. Approximately 38.4% of households have small pets/animals and 11.3% have large pets/animals. "
Assumptions and caveats II: mathematical modeling assumptions and limitations
The study conducted heavy mathematical modeling of possible evacuation scenarios, primarily using a mathematical model called DYNEV II. The mathematical model was built on many assumptions. Here are a few [many more are mentioned in the full length study]:
"The evacuation analysis procedures are based upon the need to:
Route traffic along paths of travel that will expedite vehicles’ travel from their respective points of origin to points outside the evacuated area.
Restrict movement toward the wildfire to the extent practicable and disperse traffic demand so as to avoid focusing demand on a limited number of roadways.
Move traffic in directions that are generally outbound relative to the location of the hazard.
DYNEV II provides a detailed description of traffic operations on the evacuation network. This description enables the analyst to identify bottlenecks and to develop countermeasures that are designed to represent the behavioral responses of evacuees. The effects of these countermeasures may then be tested with the model."
The model itself has significant limitations that must be understood in order to figure out where the study conclusions are more, and less, accurate:
“The focus of this study is to estimate times for evacuating vehicular traffic. It does not analyze ingress times, and as a result the interaction between inbound vehicles (such as responders and/or buses) with outbound evacuees is not considered. Additionally, the model is not multimodal and does not analyze evacuation times for cyclists or pedestrians or consider the interaction between drivers, cyclists and/or pedestrians. See Section 11 for Berkeley public safety recommendations related to evacuating on bike or on foot.” This caveat is important and will be again mentioned below.
As a straight quote from the study:
"Although the city has numerous exit routes, the following roadways possess the greatest capacity for vehicular traffic. Thus, they are likely to carry the greatest number of vehicles in an evacuation:
Roadways with direct exits
Roadways with direct exits
Interstate 580 (I-580)
San Pablo Ave (SR 123)
Sacramento St
Martin Luther King Jr Way
Adeline St/Shattuck Ave
Telegraph Ave
College Ave
Arlington Ave
Grizzly Peak Blvd
Ashby Ave
The following roadways do not provide direct exits out of the city, but connect into those that do and are likely to also carry the greatest number of vehicles in an evacuation:
These roadways are highlighted in Figure 9-1. While numerous roadways exit the city, these are the roadways that can process the most vehicles. These routes are not designated “Evacuation Routes” at the exclusion of other roadways. All available roadways should be considered evacuation routes[our emphasis]. These routes, along with others, will be used by the general population evacuating in private vehicles. Because of their high capacity, these routes are preferable for use by transit-dependent population evacuating on buses. The general population may evacuate to some alternate destination (e.g., lodging facilities, relative’s home, emergency shelter) outside the city or evacuated area. Transit-dependent evacuees will be routed to safety, outside of the evacuation area."
Interestingly, it seems that the study only considered roadways that are fully within Berkeley as means of evacuation, with or without a direct exit. We would have expected to see Centennial otherwise—but it is nowhere mentioned. This may limit somewhat the full impact of this discussion.
Evacuation Time Estimate for the North Berkeley Hills (the Ridge and Hillside)
The time to evacuate the North Berkeley Hills (i.e. the Ridge and Hillside) was shockingly high (depending upon the scenarios): “In a current-day repeat of the 1923 Berkeley Fire, known as Berkeley’s worst, the time to evacuate people out of the hazard area ranges from 1:35 (hh:mm) to 4:352 [depending upon scenarios].”
The evacuation study makes it 100% clear that there is no other solution to safe evacuation for the North Berkeley Hills (Ridge and Hillside) than to leave immediately upon an Extreme Fire weather alert from the City of Berkeley, since even if 75% leave early, it still may take up to 3 hours 35 to evacuate the Ridge and Hillside, when we know that it takes 2 hours 20 for a 1923-like wildfire to reach Shattuck Ave from the Ridge.
This situation creates a number of unusual problems, one of which is that there is right now no possibility of declaring a state of emergency or drawing upon special resources unless an emergency has already happened—but, if we wait until an emergency (wildfire) occurs, we will certainly incur a large number of evacuation deaths. Clearly, this will require many follow-up discussions.
To our surprise, it looks much faster to evacuate Panoramic Hill than to evacuate the North Berkeley Hills (The Ridge and Hillside): “In a fire affecting Panoramic Hill, an area shared with the City of Oakland with one way in and out, the time to evacuate people out of the hazard area ranges from 0:45 to 1:45.”
In reality, however, the study decided that it is appropriate to use a 90% percentile ETE for the ETE for Panoramic Hill, and a 100% percentile for the ETE for North Berkeley, Hills. In the table 7.3 above, where both cases are using 100% percentile ETE, The actual time to evacuate reaches 3 hours 10 for Panoramic Hill, although not because of traffic issues.
Comparing phased vs mass evacuation for the North Berkeley Hills
Another somewhat shocking result was that a phased evacuation of the North Berkeley Hills (Ridge and Hillside) was often slower than a single-phase evacuation, and never faster: “The results are shown in Table 7-3. They indicate that a phased evacuation approach can actually increase the ETE for those areas closest to the hazard, (phase a and phase b)”, and “Phasing the evacuation – evacuating areas at risk in stages, with the closest to the fire leaving first results in similar or worse evacuation times, both for the area overall and for the populations closest to the origin of the fire.”
We are assuming, then, that in the case of a wildfire driven onto the North Berkeley Hills, the city will likely give a single massive evacuation order for all zones in the North Berkeley Hills.
Leave Early Policy: evacuating on Extreme Fire Weather warning
The Berkeley Fire Department currently recommends that residents of the Fire Zones 2 and 3 (the majority of Ridge and Hillside) evacuate upon an Extreme Fire Weather warning. The study agrees: “The Berkeley Fire Department should maintain its Leave Early Policy for extreme fire weather. Study results indicate population-level improvements such as the leave early policy could improve evacuation times by as much as 68%.”
Red Curbing: parking restrictions in the Berkeley Hills
The study finds that “Implementing parking restrictions (red curbing) in the Berkeley hills is unlikely to substantially improve evacuation flow. The principal limitations to vehicle movement are not primarily due to reduced roadway widths caused by parked cars. Instead, bottlenecks occur at intersections, which have a finite capacity to manage the merging of traffic from multiple directions. Widening lanes by prohibiting parking would likely cause vehicles to arrive at these already congested intersections more rapidly, potentially worsening the congestion. Moreover, the inherent geometric characteristics of Berkeley’s hillside roadways, including their curvature and steepness, impose natural speed limitations, further reducing the impact of on-street parking on evacuation times. Additionally, during an evacuation, most privately owned vehicles currently parked along roadways will be in active use for egress, thereby diminishing the effect of side street parking on overall evacuation speed in the hills.”
However, we do find a problem in the model capability that impairs the quality of this finding. We found early that the model “does not analyze ingress times, and as a result the interaction between inbound vehicles (such as responders and/or buses) with outbound evacuees is not considered.” We find two very significant issues with this:
Evacuation traffic in our small streets will not be one-way but two-way traffic, depending upon how individual drivers decide they want to evacuate. The model does not account for that.
Neither does the model account for firefighting resources ingress.
It seems to us that these two issues may be enough to impeach this specific conclusion of the study. We should mention, however, that the study does raise one of these points itself: “It’s possible that parking restrictions could improve the ingress of emergency responders both in day-to-day response and during evacuations. The City could consider the strategic implementation of parking restrictions along arterial roadways in the hills, with a focus on those roadways predominantly utilized by emergency response vehicles for ingress during emergency response.”
Since road congestion is a major evacuation issue for Berkeley Hills residents, the study recommends encouraging carpooling to reduce the number of vehicles on the roads: "Evacuees with vehicles should carpool (i.e., evacuate as a household in a single vehicle) to the extent possible to reduce the number of vehicles on the road, which will lessen traffic congestion and could reduce evacuation time. Evacuees with vehicle access should also be encouraged to include transit-dependent neighbors in their evacuation plans. Similarly, transit-dependent evacuees should be encouraged to improve their mobilization times by connecting with neighbors and making plans to carpool in an emergency evacuation. "
While we are wholeheartedly in favor of carpooling, outside of households evacuating together we find it difficult to encourage people to depend fully upon a third party for evacuation: it is a life-or-death situation, not a Berkeley-to-SF commute. This choice would need to be a choice of opportunity, and everyone needs a surefire way to evacuate.
The study mentions drivers using 2nd and 3rd routing choices: “Evacuees should understand and practice using primary, secondary, and tertiary evacuation routes. An ability to react to changing conditions will improve a household’s evacuation travel time and will contribute to a reduction in overall evacuation times by distributing traffic on available roadways.”
We are unsure that many would be willing to make the choice of driving through the small side streets of the Berkeley Hills at evacuation time if nothing is done about red curbing. The likelihood of a permanent jam in these streets is very high in most people’s minds.
The study recommends against evacuating from the Hills in that manner. “Berkeley’s public safety officials do not currently recommend biking or walking as a primary evacuation plan for wildfire. While, under some circumstances for some people, pedestrian or bicycle evacuation could potentially provide a faster means of egress than using a car, there are significant safety risks associated with this approach. Ultimately, the decisions about when and how to evacuate belong to individuals, who will decide based on their own risk tolerance. All evacuations involve potential for injury or death, if only from collisions and crashes. The lowest-risk option is to pre-emptively leave the Berkeley Hills on days when the Berkeley Fire Department has declared extreme fire weather.”
There is no doubt that preemptive evacuations are the right way to go. In the case of a fast-moving wildfire occurring outside of Extreme Fire Weather, the judgment call for many of us having to leave at the last minute may be difficult. Many died in the 1991 Oakland fire after abandoning their cars.
The study finds that, as they stand today, they slow down evacuation: "This analysis indicates that the Traffic Calming Devices (TCDs) throughout the city are performing as designed […] this intended function conflicts with the goal of evacuation, which is to expedite the egress of individuals from the affected area. "
The study follows up by providing possible high-tech alternatives for existing traffic calming devices. We do not see any such options in the present budget climate. It may have to be a choice for or against traffic calming by the city council, with a price to pay on both sides.
The study recommends against it: “these analyses therefore illustrate an upper bound of potential long-term impacts from ADU/JADU development in the Hillside Overlay. Based on the results of the study, it is recommended that the City institute separate, more restrictive ADU/JADU development provisions in the Hillside Overlay. In addition, these studies identified impacts to evacuation times from a projection of likely development that could result from implementation of “Middle Housing” zoning changes. The results of this study indicated that in a repeat of the 1923 Fire (R02), Middle Housing could add between 5 and 10 minutes overall to ETEs. It is recommended that the City examine these increases in the context of fire spread scenarios.”
So far, the city has followed this recommendation. Anyone who has driven through the side streets of the Berkeley Hills understands exactly why increased housing density in the Hills can only lead to more evacuation deaths. We hope the city keeps on listening to the voice of reason.
