December 15th 2016 San Geronimo Creek Flood

12:30 pm 650 cfs

2:20 pm 1,250 cfs

4:23 pm: almost peak flow (the 3,000 cfs peak flow at the downstream gage was at 5pm)

Next day at 100 cfs: note high water marks on the fence

On December 15th, 2016, San Geronimo Creek flooded basements in San Geronimo and cars and houses in Forest Knolls. It was the largest flood since 2006.


San Geronimo Creek at the Lagunitas gage peaked at 5 pm at approximately 3,000 cubic feet per second (cfs)--roughly a ten-year flood. Average rain rates on Mt. Barnabe of 0.44 inches per hour from 10 am to 1 pm had raised the creek from about 200 cfs to about 1,000 cfs. And then there was a lull, an hour with only 0.26 inches of rain, and the creek rose more slowly to 1,200 to 1,300 cfs. But the radar image showed heavy rain on its way ashore from the Pacific Ocean, and sure enough it soon rained harder, with a peak hourly rate of 0.54 inches from 2-3 pm. The creek responded, rising rapidly again and kept rising rapidly as 0.44 inches of rain fell from 3-4 pm. As the creek reached about 2,200 cfs at 4pm, rain rates decreased, with only 0.13 inches falling from 4-5 pm, however the runoff from the previous hours was already on its way downhill, and perhaps spatially variable rain rates kept dropping higher amounts in other parts of the watershed. The creek crested at about 3,000 cfs at 5pm.

It is interesting to compare the December 2016 flood with the highest previous flood in the last few years, the December 11, 2014 flood. The rain rates in 2014 were close to 1/3 of an inch per hour around the time of the 9 am 1,500 cfs peak flow (slightly higher than a 2-year flood, with a 50% chance of recurrence each year), with the exception of an extraordinary hour from 7-8 am when 1.04" fell. Note that the rain gage used in 2014 was in Woodacre.

Assuming all the things we need to assume to make this a valid comparison (such as: different individual rain gages are representative of the average rainfall in the watershed upstream of the stream gage, uniform rainfall over the watershed, minimal sub-hourly variability in rainfall, accurate stream gage calibration, development of impervious surfaces in the watershed between 2014 and 2016 not increasing runoff, guesstimating approximate travel time between rain hitting the rain gage and flow hitting the flow gage, etc.), a few interesting conclusions can be proposed from this comparison:

• Rain rates of 1/3 of an inch per hour have trouble sustaining a creek flow that exceeds 1,200 cfs. We saw this rate allow a drop in flow after the peak in 2014, and we saw a rate less than this (0.26 in/hr) maintain the creek at about 1,250 cfs at around 2-3 pm on December 15th.
• Rain rates close to half an inch per hour, if sustained for a few hours, cause a tremendously fast rise to at least 3,000 cfs, and maybe higher if they kept up.
• Counter intuitively, the same amount of rain falling over a shorter period may be better (for avoiding flooding) than over a longer period. In 2016, 2.69" fell in the 7 hours before the peak flow. In 2014, 2.64" fell in the 5 hours before plus 1 hr after peak (2.31" was before the peak). Knowing only this, one might expect that the 2014 peak would have been larger, with the same amount of rain falling over a shorter period. But that was not the case.
• Wet antecedent conditions play a large role in filling up the watershed's soils and innumerable "reservoirs" (reservoirs meaning anywhere rainwater fills prior to "spilling" in overland flow, such as tree canopies, soils, depressions, etc.).In 2014, it took 1.4 inches of rain over 2 hours (0.7"/hr) to get the creek from 250 cfs to 1,500 cfs. In 2016, it took 2.1 inches of rain over 4 hours (0.53"/hr) to get the creek from 250 cfs to 1,500 cfs. It was a slower rate in 2016, but a larger total, and the larger volume filled up a lot of pore spaces that then "spilled" during the peak when their "reservoir" capacity ran out. 
• 3.94 inches fell during the 25 hours before the 2016 peak, and 3.09" fell during the 25 hours prior to the 2014 peak--not that different--so we should look at antecedent conditions even farther ahead: Four hours before the Dec. 11, 2014 peak, the creek was running at 33 cfs, and the rain started 13 hours before that when the creek was at 7 cfs (an inch of rain fell between 7 cfs and 33 cfs at 0.08"/hr). On Dec. 15, 2016, the creek was at 33 cfs 25 hours before the peak. The rain started 22 hours before that when the creek was at 13 cfs (an inch of rain fell between 13 cfs and 33 cfs at 0.05"/hr).

So if you want to predict a big flood on San Geronimo Creek--whether the creek exceeds about 1,500 cfs--look for an inch of rain in less than a day, followed by a day with over 3 inches of rain. On that second day, look for rain rates exceeding 1/3 of an inch per hour and approaching 1/2 inch per hour for at least 2 hours. These conditions will make it rise really fast between 250 cfs and 1,500 cfs. Once it is over 1,500 cfs, anything over 1/3 of an inch per hour is bad news.

12/18/16 Update: After looking back a bit farther, I see the December 2012 floods were even bigger than the one in 2014. On December 2, 2012, from 6-8 am, San Geronimo Creek went from 200 cfs to 2,200 cfs in 2 hours! This was roughly a 5-year flood (20% chance of recurring each year). At Mt. Barnabe, over 3" fell on Nov 30, over 1" fell on Dec 1, and 2.24" fell on Dec 2nd, with a peak hourly rate of 1.09"! Then three weeks later, on December 23, 2012, it peaked at 1,730 cfs at 12:45 pm, about 5 hours after exceeding 200 cfs. Over 1.5" fell on Dec 21, 1.2" on Dec 22, and 2.33" on Dec 23. Peak hourly rain rates of 0.37" were sufficient to get the creek higher than it got in 2014.

12/19/16 Update: I haven't been able to get out and see much, but initial impressions:
1. Significant aggradation of point bars and sediment deposition on floodplains (the few places where they still exist).
2. Significant thinning of alders less than four inches in diameter on point bars. The remaining trees may get large enough to withstand future floods of this magnitude.
3. Cobble-sized bedload was moving (deposited on the Creamery Creek confluence point bar).
4. Significant movement of large woody debris, including toppling of trees on banks, both large trees on steep banks and small trees on flatter areas.
5. I spoke to the gentleman that lives next to the Creamer Creek culvert, and he said that the culvert was plugged (the road flooded) and the county removed most of the larger material. You can see the high water marks on the upstream fence and in his yard. He was climbing down to the culvert with a measuring tape.

You can imagine that when people build in floodplains, or inherit or purchase already-existing poorly-sited development, they want to avoid redesigning or moving their own infrastructure in major ways. In order to protect it from future flooding, they will be asking the county to clear large wood and build larger culverts and armor banks. These apparently "minor" modifications are not comparable to the "minor" modifications property owners can make to protect their infrastructure. Building larger culverts is wise--especially where fish passage is impaired--but clearing wood and armoring banks is not always a good idea. Clearing wood and armoring banks may be cheap and seem to be minor actions, but cumulatively these can result in major impacts to the streams. Both actions speed up water, which can result in less flooding, but also can generate more bank erosion because the water is moving faster. The key is to slow down the water wherever possible. A key tenet of proper watershed management is to keep water and nutrients high in the watershed for as long as possible--don't just try to channelize everything and flush everything out to the ocean. Watersheds that hold onto water and nutrients have more complex habitats and healthy functions, which benefit us all. Instead of individuals throwing rocks on the banks and pulling trees out of the creek, they would be wise to come together (even as the county) to develop a creek management plan, but with a different purpose than the salmon enhancement plan. The creek management plan would devise desired channel characteristics for each reach of stream that would both protect salmon and property, to the greatest extent possible, where those goals overlap. Where they don't, certainly at times the habitat would have to compromise, but most of the time the property owners could attempt to reduce their impacts on the creek by moving infrastructure out of the way, or waterproofing it, or redesigning it in some way. A plan could be designed that could meet both objectives, reach by reach, with priority habitat and infrastructure for each reach identified.

We have a functioning-but-impaired stream system running through our valley. It is not just a flood control channel--yet. Incremental actions could result in more channelization and a poorer stream system. With the current high property values, owners are going to be responding to this flood with their own funds in a piecemeal manner, and will cause great degradation if they are not careful. They will also be asking the county to make changes that might benefit individuals, however may not always be in the public interest. I hope the existing plans and regulations, the careful thoughtful execution of those by county staff, and the care, love, and wisdom of individual property owners are all able to maintain as much of the naturally-functioning stream ecosystem as possible.

12/31/16 Update: The December 2012 flood was higher in San Geronimo than this one, even though this one was bigger at the Lagunitas gage downstream. That means that this flood had a greater proportion of rain runoff entering the creek downstream of San Geronimo than in December 2012. That seems consistent with the considerable flooding this time in Forest Knolls. Also interesting are reports that the Platform Bridge Road was a foot underwater at 5 pm and 18 inches at 8:50 pm. At the SPTSP gage, the 9 pm flow was about half of the 5 pm flow; at Point Reyes Station the peak was around 9 pm, almost double the 5 pm flow. About 4/5 of the peak flow at 9 pm at PRS passed through SPTSP at about 6 pm, revealing roughly a 3-hour travel time between the two gages.