Evaluating FWI for Personal Weather Station

In An effort for a Simpler Fire Weather Index I described my new FWI and the theory behind it. In short, this pwsFWI (as I have baptised it) is meant to be a generic FWI, valid everywhere and independent of geology and vegetation.

The pwsFWI is a (not too) complex measure of local meteorology, an indicator composed of humidity, wind speed and temperature. It fluctuates under ‘normal’ conditions and if it becomes dryer (a longer period without rain, the number becomes higher. As soon as it starts raining, the value starts dropping.

Evaluation method and data

I made a first implementation for evaluation purposes, but since I do not live in an area where forest and nature fires are abundant, I needed some extra data for evaluation from elsewhere.

Calibration or testing would be other qualifications of this phase.

Cumulus has the advantage of having data easily accessible, more easily than other programs. So I wrote some software to use my own data for first testing and then asked the community for additional data. I had some quick reactions and after some discussion it was made clear, that many sites have their dayfile online for website purposes.  Not all data are good as you can read below. I made a selection of some sites and started using those dayfiles for evaluating my Fire Weather Index.

Those sites are:

  1. Inverellit (AUS): http://weather.inverellit.com/index.htm ; Latitude  S 29° 46′ 26″; Longitude  E 151° 07′ 14″; Elevation 586 m.
  2. Vilppula (FI): http://www.saahiisi.com/ ; Latitude: N 62 01 00 ; Longitude: E 024 29 21 ; Elevation: 111 m.
  3. Wairoa (NZ): http://www.wairoa.net/ ; Latitude: S 39° 02′ 42″ ; Longitude: E 177° 25′ 35″ ; Elevation: 55 m.
  4. Oinofyta (GR): http://oinofytaweather.gr/ ; Latitude: 38.308 N ; Longitude: 23.642 E ; Altitude: 110 m.
  5. Niesiolowice (PL): http://www.pogoda-niesiolowice.kaszuby.pl/ ; Latitude: N 54° 12′ 24″; Longitude: E 17° 51′ 01″; Altitude: 186m.
  6. Apeldoorn (NL): http://www.apeldoorn.tk/ ; Latitude:  52.21986 N ; Longitude: 6.00046 E ; Altitude: 11m.
  7. Komoka (CAN): http://www.komokaweather.com/ ; Latitude: N 42° 57′ 15″; Longitude: W 81° 26′ 06″; Elevation: 244 m;

In total I selected 7 sites, most in areas vulnerable to forest fires, some (FI, CAN) were not. I used data from 1 januari 2016 for northern hemispere and 21 june for southern hemisphere to have four seasons in the testset. The site in NL was badly chosen, which I will explain below. All data have been stored in a sheet which you can download here to study. Comments are welcome.

Some additional datasheets are added to compare previously discussed indices (Chandlers Burning Index, Angstrom Index and FMI and The Canadian FWI). All those indices are either too simple, lineair, too complex or not up to their task.


In the sheet there are a lot of data and graphs. I advice you to start with the graphs, to get an overview. The graphs of the sites above contain a value for pwsFWI for every date, together with the constituting measurements. Next it has a Smoothed pwsFWI and that is the value you must look at, it is the value to be used as a warning indicator. Currently it runs from 0 to 600. The interpretation at the moment is:

0 – 150: Danger level: LOW: The fuel is wet, no drying has taken place.

150 – 250: Danger level MODERATE: The combination of wind, humidity and temperature favour a drying process which has started.

250  – 350: Danger level: HIGH: The drying process has continued and is basically complete. Little humidity left in litter and surface of wood.

450 – 550: Danger level: VERY HIGH: Last stage of the drying process, fuel has become easily inflammable. Ponds are falling dry.

550 and up: Danger level: EXTREME: Drying has completed, no moisture left in fuel. Fuel easily inflammable, a spark can ignite it. Almost explosive.

I asked the website owners to check the values against their experience and memory.


  1. The sheet pwsFWI provides insight in the interaction between temperature, humidity and VPD (vapour pressure deficit). It is just there for clarification.
  2. The sheets pwsFWI2 and pwsFWI3 show the pwsFWI over a relative short period – data from my own station – and show the effect of what I call ‘Rain Quenching‘ and Delay. Rain Quenching is the effect of rapid stopping and reversing the drying process by rain. As an effect the fire danger index (pwsFWI) will drop very quickly if enough rain will fall. You see a period of end of July, one of two heatwaves in the Netherlands in 2019 which ended by rain and a stark drop in temperature, just after the pwsFWI reached a peak and fire danger was becoming real. in pwsFWI2 you see the danger level without the Rain Quenching method, in pwsFWI3 you see the effect clearly demonstrated. Delay is the effect in nature, that nothing happens immediately. If temperature rises and wind start blowing, drying a forest is not immediate. It takes some days. I made an effort to implement this.
The test datasets:
  1. Inverell: Three fire seasons show up with some peaks at a value of 600 (extreme fire weather). In comparison with other locations this gives me the impression that the maximum value for pwsFWI might be 700 or even 800. Feedback from Phil, the siteowner, is described here.
  2. Vilppula: High up in Finland, this example illustrates the fact that real fireweather is not an issue up north (or south). This does not mean forests over there can or will not burn but they will be not be easily lit.
  3. Wairoa: With four seasons it can easily be seen that fire danger is often imminent although not as extreme as in Inverell. The season 2016/17 was definitely more dangerous (and probably had more fires) than the other years.
  4. Oinofyta: These four seasons show the true effect of the strong Mediterranean winds en the effect of drying the woods. All four seasons had periods of extreme danger and no doubt there were live fires during all seasons. It is this graph which illustrates why I think the maximum value of pwsFWI should be higher, probably in the range of 700 – 800 (and the other danger ratings adjusted accordingly).
  5. Niesiolowice: This Polish village is 55 km from Gdansk in a forested area. Fire weather is an issue there and the weather site carries the famous FWIcalc program and dashboard. The pwsFWI is coming close to 600 but reaches never an extreme level. 2018 has been the driest season of all four shown.
  6. Apeldoorn: Apeldoorn is close to the largest forested area’s of the Netherlands so that might have been an interesting pws. However, all values stay low and hardly ever surpass the value of 150. That is strange. Of course I know the Netherlands and I compared it with my own station over the last three months. It does not match. Looking at the data, It can be seen that values of humidity are high, wind is too low and rain also seems extreme (but not impossible). My guess is, that either the calibration of the station is off or it’s location is not good enough (no open field, drip of of trees or elsewhere sheltered. I will check.
  7. Komoka: Located in a forested area, but takes a lot of rain and is influenced by the great lakes. Although the last seasons seems to have values of pwsFWI surpassing 200, this will never be a fire dangerous spot. Which is already confirmed by PaulMy .


This is the current status. I will continue developing, especially from a theoretical point of view, and implement corrections for altitude and season. I will blog on this when I have some result. Comments are welcome and if necessary will induce change to the pwsFWI.

Making this tool publicly available is a wish, but problems with the development environment may put this back a little. It also depends on interest of course. And I will try to make it available on all platforms CumulusMX runs on. This means I will try to develop under the same environment as CumulusMX is developed. That is a drawback, because it is new to me. At the same time : challenging.

Geef een reactie