Correlation between Weather Conditions and Burnt Areas in 25 Years of Forest Fireﬁghting by the Fire Brigade †

: In this paper, research on the relationship between forest ﬁres and meteorology is carried out using simple meteorological variables (temperature, wind, and precipitation) from a few well-selected Greek meteorological stations. The 5 years (1998, 2000, 2007, 2012


Introduction
Nowadays, environmental protection is a very sensitive topic.The degree and manner of the utilization of natural wealth indicates the quality and level of cultural literacy of a community of people and is a reliable indicator of a country's social, political, and economic development.It is now well known that the increase in burnt areas and the occurrence of mega-fires affects, to a greater or lesser extent, every country on every continent.It disrupts national economies and has a major impact on the people, communities, and countries of the present and the future.
In this paper, an attempt is made to correlate meteorological parameters (wind, temperature, and precipitation) with burnt areas within Greece over the last 25 years (1998-2022).Specific years are studied (1998, 2000, 2007, 2012, and 2021) where the burnt area far exceeded the 25-year average (37,000 Ha), as well as other years where the burnt area was below the corresponding average.Finally, temperature, wind, and summer and winter precipitation conditions are investigated in terms of to what extent they impact the total burnt area in each year.

Materials and Methods
The following meteorological data were used for the present analysis: (a) Temperature measurements as well as wind direction and wind speed at 15:00 local time (12:00 UTC) from 1 May to 30 September.(b) The average monthly temperature values.(c) The daily and monthly rainfall amounts of ten (10) meteorological stations of the Hellenic National Meteorological Service (HNMS) from 1998 to 2022.These 10 stations have been selected from a list of HNMS stations, whose data are freely available on the internet through their summary meteorological observations (SYNOPS) in order to provide an overview of the weather conditions of the whole country.More specifically, the following stations have been selected: -Andravida and Corfu for western Greece.-Tripoli, Larissa, and Kastoria for the mainland.-Alexandroupolis and Thessaloniki (Mikra) for north Greece.-Athens, Samos, and Heraklion for eastern and southern Greece.
As for Athens, the Elefsina meteorological station was used for the midday temperature and wind observations, and the station of the National Observatory of Athens (NOA) was used for the monthly mean precipitation and temperature values due to the incomplete meteorological observations of Hellinikon for the years 2013 and 2014.
Regarding the selection of these stations, their representativeness for the abovementioned wider areas and, in general, the justification of the whole methodology, an extensive analysis and documentation has been carried out in the respective report for the year 2016 [1], as well as in the relevant paper of the 18th Panhellenic Forestry Conference [2].It should be noted that: The main sources of the monthly mean air temperature and precipitation were the websites of TuTiempo.net(https://en.tutiempo.net/climate/01--1998/ws-167100.html(accessed on 11 May 2023)) and HNMS (http://www.emy.gr/emy/el/climatology/climatology_month(accessed on 11 May 2023)) [3,4].
The main source of daily values of temperature, precipitation, wind direction, and wind speed was the official website of the US National Oceanic and Atmospheric Administration [5].
Based on the abovementioned data, for each station and for the period 1998-2022, the following were calculated: Precipitation, as a percentage of its normal (climatic) values for the winter period (October-April) and the two most difficult months of the summer period (July-August).For these stations of the HNMS and the NOA, the climatic values for these two periods are given in Table 1. 2.
The deviation of the mean monthly temperature of each five-month period from 1998 to 2022 from its normal values, for the whole of Greece.This is obtained by averaging the deviations of the mean monthly temperatures (of each year) of the 10 meteorological stations from their normal (climatic) values [6], deviations which show a strong symmetry.

3.
The number of days within the five-month period of May-September: With average wind speed (V) > 20 km/h (≥4 Beaufort) at 15:00 on each day.
Finally, sources of the sizes of the burnt forest areas are the Annual Fire Reports of the European Forest Fire Information System (EFFIS) [7].

Results
The winter (October-April) and summer (July-August) periods precipitation of the 10 meteorological stations of the HNMS considered from the year 1998 to the year 2022 are presented as percentages (%) of their normal (climatic) value (Figure 1).
The deviation of the mean monthly temperature of each five-month period from 1998 to 2022 from its normal values, for the whole of Greece.This is obtained by averaging the deviations of the mean monthly temperatures (of each year) of the 10 meteorological stations from their normal (climatic) values [6], deviations which show a strong symmetry.

3.
The number of days within the five-month period of May-September: (a) With temperatures (T) > 38 °C; (b) With average wind speed (V) > 20 km/h (≥4 Beaufort) at 15:00 on each day.
Finally, sources of the sizes of the burnt forest areas are the Annual Fire Reports of the European Forest Fire Information System (EFFIS) [7].

Results
The winter (October-April) and summer (July-August) periods precipitation of the 10 meteorological stations of the HNMS considered from the year 1998 to the year 2022 are presented as percentages (%) of their normal (climatic) value (Figure 1).In order to estimate the effect of temperature on burnt areas, we calculated the number of days of the fire seasons in the period of 1998-2022, during which temperatures above 38 °C were recorded at 15:00 local time at 10 representative weather stations of the HNMS (Figure 2).In order to estimate the effect of temperature on burnt areas, we calculated the number of days of the fire seasons in the period of 1998-2022, during which temperatures above 38 • C were recorded at 15:00 local time at 10 representative weather stations of the HNMS (Figure 2).
Trials to estimate this parameter by summer mean air temperature during each fire season (June-August) led to poor results (even errors of 3 days).The only sure fact is that more than one day with a temperature greater than 38 • C can be observed during the fire seasons with a summer mean air temperature greater than 1 • C.
The number of days of the fire seasons 1998-2022, during which wind speeds >20 km/h (≥4 Beaufort) were recorded at 15:00 local time at the 10 HNMS weather stations, is given below.As highlighted in previous years' reports [8], this parameter shows relatively little variation from year to year at the country level, while at the local level some stations show an increase and others a decrease in windy days (Figure 3).Trials to estimate this parameter by summer mean air temperature during each fire season (June-August) led to poor results (even errors of 3 days).The only sure fact is that more than one day with a temperature greater than 38 °C can be observed during the fire seasons with a summer mean air temperature greater than 1 °C.
The number of days of the fire seasons 1998-2022, during which wind speeds >20 km/h (≥4 Beaufort) were recorded at 15:00 local time at the 10 HNMS weather stations, is given below.As highlighted in previous years' reports [8], this parameter shows relatively little variation from year to year at the country level, while at the local level some stations show an increase and others a decrease in windy days (Figure 3).Trials to estimate this parameter by summer mean air temperature during each fire season (June-August) led to poor results (even errors of 3 days).The only sure fact is that more than one day with a temperature greater than 38 °C can be observed during the fire seasons with a summer mean air temperature greater than 1 °C.
The number of days of the fire seasons 1998-2022, during which wind speeds >20 km/h (≥4 Beaufort) were recorded at 15:00 local time at the 10 HNMS weather stations, is given below.As highlighted in previous years' reports [8], this parameter shows relatively little variation from year to year at the country level, while at the local level some stations show an increase and others a decrease in windy days (Figure 3).At last, Figure 4 shows the total burned area of forest fires during the summer season in Greece for the last 25 years.It is evident that, during these years, only five times the total burned area in Greece exceeded its mean value of 36,972 ha.
At last, Figure 4 shows the total burned area of forest fires during the summer season in Greece for the last 25 years.It is evident that, during these years, only five times the total burned area in Greece exceeded its mean value of 36,972 ha.Looking back at the previous figures, it is easy to see that a common feature of these five years (1998, 2000, 2007, 2012, and 2021) is much lower-than-normal summer (July-August) precipitation followed by lower-than-normal winter precipitation too (Figure 1).In the years following (1999, 2001, 2008, 2013, and 2022), the burnt areas were below their average regardless of the precipitation and summer temperature patterns.Explanations for that do exist, but they are not interesting from a meteorological point of view.Another common feature of these years is a greater-than-normal mean number of days with a temperature above 38 °C at 12UTC (Figure 2).Of course, this was also observed in 2017 (Figure 2), but a very rainy July (Figure 1) ultimately prevented large burnt areas.However, a common feature of all years with summer (July and August) precipitation higher than normal (including the year 2017 and prior to that of 2002 in Figure 1) is the very few total burned areas (Figure 4) regardless of whether the preceding winter was dry or not.As for the year 2016, which in Figure 1 is presented as very dry, the reason why its burned areas do show an increase, but not exceeded their mean value (Figure 4), may be the small number of very hot days (Figure 2), as well as the great amount of rain at the end of June of the same year.Concerning, finally, Figure 3, its comparison with Figure 4 does not show any obvious correlation between wind and total annually burnt areas, such as that of its previous diagrams which are discussed above.

Discussion
The importance of the extracted results above is enormous!First of all, as they were retrieved easily without any difficult statistics, they can also easily be understood by people who are not familiar with mathematics, such as firemen, ministers, politicians, and others.Secondly, it facilitates the choice and application of the appropriate, more complex statistics for deeper learning and practical use.So, based on these findings, an expectational relationship between a rainfall index (x) and total annually burnt area (y) has been found for Greece and is shown in the diagram of Figure 5, where: Looking back at the previous figures, it is easy to see that a common feature of these five years (1998, 2000, 2007, 2012, and 2021) is much lower-than-normal summer (July-August) precipitation followed by lower-than-normal winter precipitation too (Figure 1).In the years following (1999, 2001, 2008, 2013, and 2022), the burnt areas were below their average regardless of the precipitation and summer temperature patterns.Explanations for that do exist, but they are not interesting from a meteorological point of view.Another common feature of these years is a greater-than-normal mean number of days with a temperature above 38 • C at 12UTC (Figure 2).Of course, this was also observed in 2017 (Figure 2), but a very rainy July (Figure 1) ultimately prevented large burnt areas.However, a common feature of all years with summer (July and August) precipitation higher than normal (including the year 2017 and prior to that of 2002 in Figure 1) is the very few total burned areas (Figure 4) regardless of whether the preceding winter was dry or not.As for the year 2016, which in Figure 1 is presented as very dry, the reason why its burned areas do show an increase, but not exceeded their mean value (Figure 4), may be the small number of very hot days (Figure 2), as well as the great amount of rain at the end of June of the same year.Concerning, finally, Figure 3, its comparison with Figure 4 does not show any obvious correlation between wind and total annually burnt areas, such as that of its previous diagrams which are discussed above.

Discussion
The importance of the extracted results above is enormous!First of all, as they were retrieved easily without any difficult statistics, they can also easily be understood by people who are not familiar with mathematics, such as firemen, ministers, politicians, and others.Secondly, it facilitates the choice and application of the appropriate, more complex statistics for deeper learning and practical use.So, based on these findings, an expectational relationship between a rainfall index (x) and total annually burnt area (y) has been found for Greece and is shown in the diagram of Figure 5, where: x(year) = 100 − {[3 * x(October-April) + x(July-August)]/4 (1) x = % percentage of the total rainfall amount of each time period (whole year, October of the previous year to April of the current year or July and August of the current year) in terms of its climatic values.
Environ.Sci.Proc.2023, 26, x 6 of 7 x(year) = 100 − {[3 * x(October-April) + x(July-August)]/4 (1) x = % percentage of the total rainfall amount of each time period (whole year, October of the previous year to April of the current year or July and August of the current year) in terms of its climatic values.This relationship, combined with the other mentioned results, is very useful as already in the early days of each fire season it will be possible to have an estimation of its outcome and to prepare more appropriately when the accuracy of the long-term seasonal forecast of temperature and rainfall anomalies is improved.It must be mentioned that x is essentially a drought index for the whole country.The significant impacts of the summer as well as of the droughts of the previous winter period on the forest fires of Greece before the year 1998 can be found in [8][9][10].

Conclusions
During the 25 years of forest firefighting by the Hellenic Fire Corps, timeseries of specific meteorological parameters without any complex statistical processing easily reveal certain common characteristics of some years.Specifically:

•
The 5 years (1998, 2000, 2007, 2012, and 2021) which recorded burnt areas more than the 25-year average (37,000 ha) had a much drier-than-usual two-month period of July and August following a winter period with a shortage of rainfall as well as a number of days with very high temperatures greater than the average of the last 25 years.The latter implies that the average temperatures in all three summer months were at least 1 °C above their normal values.

•
In the years following destructive ones, the burnt areas were below their average regardless of the precipitation and summer temperature patterns.

•
When precipitation during the two-month period (July-August) was two or more times more than normal, fire seasons had, at most, one-third of the 25-year average of burned areas regardless of the precipitation of the preceding winter periods and summer temperatures in those years.This relationship, combined with the other mentioned results, is very useful as already in the early days of each fire season it will be possible to have an estimation of its outcome and to prepare more appropriately when the accuracy of the long-term seasonal forecast of temperature and rainfall anomalies is improved.It must be mentioned that x is essentially a drought index for the whole country.The significant impacts of the summer as well as of the droughts of the previous winter period on the forest fires of Greece before the year 1998 can be found in [8][9][10].

Conclusions
During the 25 years of forest firefighting by the Hellenic Fire Corps, timeseries of specific meteorological parameters without any complex statistical processing easily reveal certain common characteristics of some years.Specifically:

•
The 5 years (1998, 2000, 2007, 2012, and 2021) which recorded burnt areas more than the 25-year average (37,000 ha) had a much drier-than-usual two-month period of July and August following a winter period with a shortage of rainfall as well as a number of days with very high temperatures greater than the average of the last 25 years.The latter implies that the average temperatures in all three summer months were at least 1 • C above their normal values.

•
In the years following destructive ones, the burnt areas were below their average regardless of the precipitation and summer temperature patterns.

•
When precipitation during the two-month period (July-August) was two or more times more than normal, fire seasons had, at most, one-third of the 25-year average of burned areas regardless of the precipitation of the preceding winter periods and summer temperatures in those years.

Figure 1 .
Figure 1.Average winter (October-April) and summer period (July-August) precipitation of the 10 selected meteorological stations of HHNMS, as percentages (%) of their climatic value, from 1998 to 2022.Orange circles are only for visualizing 5 specific years.

Figure 1 .
Figure 1.Average winter (October-April) and summer period (July-August) precipitation of the 10 selected meteorological stations of HHNMS, as percentages (%) of their climatic value, from 1998 to 2022.Orange circles are only for visualizing 5 specific years.

Figure 2 .
Figure 2. Average number of days with temperature values >38 °C (12UTC) of the 10 selected meteorological stations of HNMS (indicative for the whole country) during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for T > 38 °C.

Figure 3 .
Figure 3. Average number of days with wind speeds (V) > 20 km/h (12UTC) of the 10 selected meteorological stations of HNMS, as indicative of the whole country, during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for V > 20 km/h.

Figure 2 .
Figure 2. Average number of days with temperature values >38 • C (12UTC) of the 10 selected meteorological stations of HNMS (indicative for the whole country) during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for T > 38 • C.

Figure 2 .
Figure 2. Average number of days with temperature values >38 °C (12UTC) of the 10 selected meteorological stations of HNMS (indicative for the whole country) during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for T > 38 °C.

Figure 3 .
Figure 3. Average number of days with wind speeds (V) > 20 km/h (12UTC) of the 10 selected meteorological stations of HNMS, as indicative of the whole country, during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for V > 20 km/h.

Figure 3 .
Figure 3. Average number of days with wind speeds (V) > 20 km/h (12UTC) of the 10 selected meteorological stations of HNMS, as indicative of the whole country, during the fire seasons from 1998 to 2022.The red horizontal line is the mean time series value for V > 20 km/h.

Figure 4 .
Figure 4. Total burned area of forest fires during the summer season in Greece, from 1998 to 2022.The blue horizontal line is the mean time series value for burned area 36,972 Ha.

Figure 4 .
Figure 4. Total burned area of forest fires during the summer season in Greece, from 1998 to 2022.The blue horizontal line is the mean time series value for burned area 36,972 Ha.

Figure 5 .
Figure 5.Total annually burned area in Greece (y) as a function of the rainfall index (x).

Figure 5 .
Figure 5.Total annually burned area in Greece (y) as a function of the rainfall index (x).