Active Fire Data¶
Let us compare the current data which we can find on the websites of INPE Brazil with the NASA Active Fire Data on firms.modaps.eosdis.nasa.gov.
How does a Brazilian prediction of the fire risk compare to an actual observation by a satellite?
Brazilian data¶
Fire risk index map for 16-03-2020 (1 day prediction)¶
check the predicted risk of the max. FRP : POINT (-60.68428 4.21427)¶
Spots160320 =brazil[brazil.geometry.str.contains("-60.68428")] OK 0.07
There are only 52 observations that contain frp-values. That makes 24,3% for that day of data.¶
At row 305 we find an outlier with a value of 777,7. I don't know yet if this is a mistype or a technical bug. I'll turn this value into 0.77 for better compliance with the other data.
standardize diasemchuv precipitac riscofogo frp¶
Fire risk index map for 16-03-2020 (actual observations)¶
Finding out the radiated power for a spot and the predicted fire risk index/value of an area.¶
In order to verify the (usefulness and) effectiveness of the predicted fire risk Risco de Fogo (RF), we need to link the predicted fire risk (area) to the actual observations of wild fires (point within the area) which have a FRP value. Lately I found out that there are at least 2 fire risk indici in use when you look at the 2020 datasets:
- risco_fogo is a number (integer) between 0 and 5,
- currently used for the predictions the next 1 to 3 days
- riscofogo is used in the observation data and the 2019 report:
- is a value between 0 and 1 (1 = maximum risk).
| Risco | Risco de Fogo (RF 2019) | RF 2020 (predictions) | ||
|---|---|---|---|---|
| Mínimo | RF ≤ 0,15 | 1 | ||
| Baixo | 0,15 ≤ RF ≤ 0,40 | 2 | ||
| Médio | 0,40 ≤ RF ≤ 0,70 | 3 | ||
| Alto | 0,70 ≤ RF ≤ 0,95 | 4 | ||
| Crítico | RF ≤ 0,95 | 5 |
NASA FIRMS Fire Information for Resource Management System Data¶
For that particular outlier found in the Brazilian dataset, we found a match. It turns out that the NASA annotates its data with a confidence column. The confidence for this outlier has been marked as being "low".
Perhaps there is a reason for that strange value in the Brazilian dataset: to be able to filter it out more easily.
finding the points contained within the borders of Brazil¶
- GeoSeries crs mismatch: epsg:4326 and epsg:4674
- so we set the correct CRS: Latitude - Longitude
- now we can filter out the data lying within the Brazilian border
Background info for plot 16/03/2020¶
- value 0412 on the x-axis is the time of the observation expressed in GMT Greenwich meantime 04:12:00 AM.
- This would indicate a local time at Manaus in winter time of 00:12:00 AM
- No data after 17:00 due to no coverage by any satellite, or bad atmosferical conditions.
FRP observations cumulated for 16/03/2020¶
We add up the FRP's during 1 day, so we can compare it to the average.
FRP observations cumulated for the Legal Amazon 16/03/2020¶
In human terms: on March the 16th the observating satellites detected the emission of 3300 MegaWatts of energy in wavelengths that are known to quantify correctly the burning of diff. kind of vegetations.
At this moment of the year it is still the wet season in the Amazon.
FRP observations statewise 16/03/2020¶
So now we know the sums of the FRP for 3 states in Amozonia Legal at March 16th.
FRP plots of 2018¶
Filtering out the rows without frp or fire risk values.
Averaging 2018 data¶
April 2018 new data became accessible with Suomi NPP.¶
Notice the sudden drop in average radiative power emitted starting from April 2018.
Starting April 2018 new data became accessible: the instrument Suomi NPP. The introduction of the Suomi NPP had 2 main reasons: the higher spatial resolution provided better location precision, and an extra orbiting satellite meant more coverage throughout 24 hours (mainly during the dark hours in this case).
But there is a drawback to witness here: adding much more numbers with lower values leads to lowering the average radiative power overall.
These gaps in observation windows are noticable in the fig. "FRP observations 16/03/2020".
Conclusions:
- we should avoid to mix the data of Suomi NPP with other data without some weighting measures.
- in order to use 2018 data for comparison, you'll need to split data in 2 series: before and after 01/04/2018.
Let's add up the numbers for FRP in 2018 cumulatively...
