One of the goals of the current implementation of pwsFWI is to see its behaviour under all conditions. It is therefore very interesting to see that behaviour in the two stations in semi-arid zones (in Australia and in Spain) where long dry periods may suddenly alternate with rain after which the drought returns.
One station in the Netherlands which was added recently, (Reeshof, Tilburg – link will be made final after revision of the site) however had a situation which required some explanation and investigation. On 26 july 2019 the warning level spiked from blue to orange and after a day it fell back. That triggered Phil (from the Australian testsite) to make a remark: 26/07/2019 jumps out.
Right. It does. And that may have effect when long dry periods suddenly receive a big shower. So I thought it over and replied:
[…] at the moment I think it is a correct calculation and it has to do with the strange weather we had at that time (in the whole of the Netherlands, actually in the whole of Western Europe).
It was the second heatwave (out of three total this year which is uncommon in itself) which started around 22/7 and culminated at 26/7 with the highest temperature (officially) ever recorded in the Netherlands on the 25th of 40.7 °C. What we see in the pwsFWI record is that it starts to rise very quickly, starting 22/7 and reaching 539 on the 26th. That day ended with a lot of rain (and probably thunder) so after that day the pwsFWI value starts to fall. I have been analysing this, trying to find a cause. So far I think it is a true and valid calculation of an extreme anomalous weather situation.
I did ask pdw112 for his pwsFWIanalyse.csv (and maybe his dayfile) to analyse it a bit more in detail and I will continue to check the algorithm for error, but so far I am pretty confident of the numbers. The fact that it skipped the yellow warning phase is in line with what the Dutch felt during those days: extreme unusual weather. […]
I have no reason to change anything because of this observation, but I may come back on this. So far I would say it is weird but can be explained.
I think I have to change my opinion expressed in the last sentence.
As anyone can see in the pwsFWIanalyse.csv, the actual value on the 26th of july spiked because the dayFWI spiked. After the calculation of the dayFWI – the actual value governed by the science equations as described here – that value needs to be treated.
A treatment required to implement the slowness of reaction by nature or as I say: Rain contradicts drying but not immediately; Drying timber in a forest is not a single day event. To continue that argument: bringing moisture back into a forest isn’t a single day event either. A forest has a dampening effect on moisture (within the forest and it’s direct surroundings). There is so much literature about this effect, that in this blog, I can only point to a google search (which worked for me). If you really want to go in depth on the subject, study forest ecology. Many books and articles have been written on the subject.
So when suddenly temperature, humidity or wind change, the pwsFWI may and probably will not follow directly.
And that is where the problem lies. After the calculation of the driving force of drying the fuel (which determines for roughly 70% the pwsFWI) there is an heuristic algorithm to implement the effect of rain and the slowness of the reaction by nature. I call this Smoothing (dampening the effect of peaks and troughs and sudden changes in measurements) and Quenching (stopping the drying process as a result of the addition of water such as rain). Quenching is ultimately represented by an increase in relative humidity.
It is in the Smoothing and Quenching algorithm where the first rain is actually ignored. The day with that first rain, as such is overrated when it occurs and underrated the days after (as is the case in Reeshof on 26/7/19). I will need to review that.
Possible changes in that algorithm will have no effect on the minimum or maximum values of pwsFWI, when the Smoothing and Quenching becomes irrelevant (e.g. in a prolonged wet or dry period). It only affects the relatively short period (of roughly five days) around first rain after a dry period.
If it is only one small shower, the effect will be minor and disappear quickly. A nice example of this is the small shower which hit Phil’s Backyard on 22 sep 2019.
A heavy shower (as in the Reeshof-case) will have a longer effect. If there will be even more rain will bring the vegetation back in the wet state and drying has to start all over again. But in any case, the day the first shower falls will have to count in some way.
I will review the Smoothing and Quenching algorithm and when runs with known dayfiles have satisfactory results, a new version of the software will be published.