1 Introduction

Fires occur in many forest ecosystems around the world, and especially in the Mediterranean region, and they can cause disturbance on vegetation and the soil carbon sequestration and sinks (Dexter et al. 2013; Andrew et al. 2016; Levin et al. 2016). The exact time of fire occurrence, and its intensity can be predicted. Fire danger rating is determining the environmental factors (weather, fuel topography) that influence the occurrence, size, and impact of wildland fires on a given territory (Amatulli et al. 2013).

The ignition and the spread of forest fires can be a result of environmental and anthropogenic factors. Natural weather conditions that influence the fuel distribution play a very substantial role in the fires’ number and their range, especially in regions like the Mediterranean constituency where most fires have an anthropogenic origin (Bajocco et al. 2015).

The simultaneous monitoring of fuel moisture content and weather conditions can be beneficial to indicate forest fire danger. Forest fire danger rating systems are founded on the combination of weather conditions with biophysical features of the predicted combustibility and flammability of the accumulated fuels along with the anticipated anthropogenic risk events and natural occurrences (Pereira et al. 2011).

Fire danger systems have established from several provincial varieties into a National System nowadays. The Canadian Forest Fire Danger Rating System (CFFDRS) embraces all aspects for evaluating and predicting fire weather conditions, fire incidence, and fire comportment. This system includes the Canadian Fire Weather Index (FWI) System and the Canadian Fire Behavior Prediction (FBP) System (Pasqualini et al. 2011).

The FWI was not only developed and used for Canadian forestry (de Jong et al. 2016), but it was similarly comprehensively exercised for other environmental settings (Wang et al. 2017). It was tested against identifying the associations between the FWI scheme and the fuel moisture content observed in confined fire activity periods. Moreover, the FWI index surpasses the routine of other fire danger rating systems concerning high fire activity periods in different environmental settings (de Jong et al. 2016). Since the early 1970s, computer systems have been used in Canada for fire management, and in 1992, the geographic information systems (GIS) was profoundly used by the Canadian Forest Services to simulate the fire management information systems and finally, the spatial configuration of the fire management systems (Lawson and Armitage 2008). The application of the above-mentioned remote automatic weather stations and respective communication machinery in the 1980s and 1990s indorsed the assortment of weather data from distant sites in nearly real-time at a local and regional events level (Taylor and Alexander 2006).

It is fairly assumed that the occurrence and the comportment of forest fires be contingent principally on weather stimulus which is short term and lasts occasionally for a few days (Amatulli et al. 2013). Long-term drying assessment provides information for fire management throughout serious conditions. This supposes that fires can take place also deprived of a period of long-term moisture insufficiency (Dimitrakopoulos and Bemmerzouk 2003). Therefore, the dry climate is not a precondition for the ignition and the blowout of fire in a given area (Amatulli et al. 2013).

The Keetch–Byram Drought Index (KBDI) is an index ranges from 0 to 800, as the higher values are related to higher drought. Data are taken from daily interpretations including the precipitation volume and the air temperature (Keetch and Byram 1968). It is taken into account the soil moisture deficiency, after assessing vegetation cover density, precipitation, and evapotranspiration losses, and is expressed in hundredths of an inch. KBDI index is updated daily, hence, it provides a continuous record and can be easily computed. It is appropriate for localities with fires because the index uses point-source data. For example, since 1990, it has been practiced in certain Mediterranean settings of Greece (Dimitrakopoulos and Bemmerzouk 2003). The KBDI is one of the pointers that are used for tracking fire potential in Hawaii, for categorizing fire-season intensity by the Florida Division of Forestry and by the fire managers in Texas for monitoring the fire possibility (Dolling et al. 2009).

The main objective of the current exploration is to compare the Canadian Forest Fire Danger Rating System and the National Forest Fire Danger Rating System tested on forest fire incidences that occurred in a typical Mediterranean forest of Crete, Greece. The comparison will be used to suggest possible improvements if possible, to adjust the CFFDRS and the NFFDRS more adequate to the Mediterranean forests.

2 Materials and methods

2.1 Study site and data collection

The field tests were conducted in the Akrotiri region, Chania (Crete, Greece). The area is situated at an altitude of 185 m above the sea level, approximately 35° 31′ N and 24° 03′ E (Fig. 1). December, January, and February are the months with the highest precipitation and the mean annual rainfall is about 600 mm. The investigated ecosystem is represented by Pinus halepensis with 50% crown closure which is a typical Mediterranean forest (Elhag and Bahrawi 2016). It has a southwest exposure with a slope of 20%. The investigation was carried out in two fire seasons (2008 and 2009), with daily observations at 14.00 h.

Fig. 1
figure 1

Location of the study area

Full size image

Under canopy litter, duff, and soil samples were congregated in a triplicate manner and were placed into hermetically sealed aluminum canisters to diminish evaporation. The daily moisture was calculated as the average of the three values. The daily necessary weather data were collected to perform the two fire risk indices from the closet weather station located in the vicinity of the study area.

Three annual plants, Avena sterilis, Parietaria diffusa, and Piplatherum miliaceum, representing the typical understory species of the Pine forest in Crete, were chosen. These plants have a shallow root system and form the understory of Pinus halepensis. Their daily moisture content was measured on an oven-dry weight (ODW) basis from March 17th, 2009, until the end of July 2009, when the plants were completely dried out in the field. The needle moisture content of Pinus halepensis (a tree species with deep root system) was determined as well, on an oven-dry weight, from August 10, 2009, daily during the fire season and every week during the rest of the year. From each species, three samples were collected.

2.2 Canadian Forest Fire Danger Rating System

The DC and DMC starting values should be considered at 0 when at least 200 mm of rain has been recorded over a period of about 3–4 days or less. One should go back in time to a point where it is relatively certain that the zero value was achieved, because of abundant rainfall. Then he should calculate forward to the period of interest. This amount of rain (200 mm) is, in the writer’s opinion, rarely recorded in a Mediterranean-type climate over a 3–4 days period. Furthermore, the difficulty of getting meteorological data did not allow the researcher to return in periods with heavy rainfalls. Thus, they made the assumption that following the period of the higher annual rainfall (December, January and February), the forest floor would be fully saturated. In these conditions, Alcañiz et al. (2016) suggest the following values: FFMC = 85, DMC = 6, DC = 15.

Fine Fuel Moisture Code (FFMC)

$$F = \frac{{59.5\left( {250 - m} \right)}}{147.2 + m}$$
(1)

Duff Moisture Code (DMC)

$$P = c\left[ {\log \left( {Mmax} \right) - \log \left( {M - E} \right)} \right]$$
(2)

Drought Code (DC)

$$D = 400\ln \left( {800/Q} \right)$$
(3)

Fire Weather Index (FWI)

$$f\left( D \right) = 0.1Rf\left( D \right)$$
(4)
$$\ln S = 2.72\left( {0.434\ln B} \right)^{0.647}$$
(5)

2.3 The National Forest Fire Danger Rating System

According to Keetch and Byram (1968), the drought index signifies the net influence of evapotranspiration and precipitation in generating collective moisture insufficiency in deep duff or topsoil layers. The material is organic and can be soil humus or suppressed wood such as roots in unpredictable degrees of decay. The KBDI index can reach maximum values up to 800 (in hundredths of an inch), and the higher values correspond to higher drought levels. The index is obtained by calculating rainfall amount and maximum air temperature from daily measurements (Keetch and Byram 1968). Under constant variation of the weather data, and the soil-duff is based on only on the water availability in the soil, then, the initial equation is formed as follows.

If weather variables which affect transpiration, such as temperature, relative humidity, and solar radiation are constant, and if the rate at which the soil-duff layer loses water is directly proportional to the amount of water in the layer, then the KBDI index can be initially estimated as follows:

$$\frac{1}{{w_{c} - Q}} \frac{{{\text{d}}Q}}{{{\text{d}}t}} = \frac{{f_{1} \left( T \right)}}{{t_{{T_{0} ,\infty }} f_{1} \left( {T_{0} } \right)\left[ {1 + y_{0} \exp \left( { - aR} \right)} \right]}}$$
(6)

When dQ is constant, it gives the final equation:

$${\text{d}}Q = \frac{{\left[ {800 - Q} \right]\left[ {0.968 \exp \left( {0.0486T} \right) - 8.3} \right]{\text{d}}t}}{{1 + 10.88\exp \left( { - 0.0441R} \right)}}*10^{3}$$
(7)

When dt is equal to one day. Therefore, dQ can be finally obtained following equation mentioned by Leverkus et al. (2014):

$${\text{d}}Q = \frac{{\left[ {203.2 - Q} \right]\left[ {0.968 \exp \left( {0.0875T + 1.5552} \right) - 8.3} \right]{\text{d}}t}}{{1 + 10.88\exp \left( { - 0.001736R} \right)}}*10^{ - 3}$$
(8)

where a is a constant and K is a function of T.

3 Results and discussion

3.1 Canadian Forest Fire Danger Rating System

The FFMC can reach maximum value up to 101. According to Flannigan and Harrington (1988), 65 is the threshold where the fire occurrence probability begins to increase exponentially. In the present observation, FFMC values range from 10.3 to 96.5 (mean value of 84.95) for the two investigated seasons and 90% of the values are lower than 75. Based on the above-mentioned results, it can be suggested that most of the days during the fire season are characterized with conditions appropriate for fire occurrence. It should be pointed out that even a small increase in the FFMC high values can influence the fuels flammability and thus making them intensively burning.

Two FWI Fuel Moisture Codes were included: the FFMC which was compared to observed litter moisture content and the DMC which was compared to observed duff moisture content (Table 1). Since both codes are inversely related to fuel moisture, fine fuel moisture (FFM) demonstrated in Fig. 2, is obtained by subtracting FFMC from 101, while DMC is converted to duff moisture (DM) according to Fernandes et al. (2014).

Table 1 Incidence percentages of FWI within fire danger temporal analysis
Full size table
Fig. 2
figure 2

Fine fuel moisture content assessment (estimated in red, projected in blue)

Full size image

Since the correlation of DC index with the observed soil moisture content was not significant at the 95% confidence level, it was later excluded from the present study. As stated previously, this code can be used as a marker of periodical drought magnitude on forest fuels, and the sum of smoldering matter in deeper duff layers and large logs. It was originally designed by Ganteaume et al. (2013) to indicate the drought effect in deep soil horizons. In the present case, observed data concern only the topsoil layer immediately exposed to weather conditions.

Daily dissimilarities of moisture content of projected and estimated fine fuel are shown in Fig. 3. The trends describing daily fluctuations of predicted and observed fine fuel moisture content are similar during both seasons, especially in the summer period, when there is a lack of rain (r ≈ 0.90). This represents the model sensitivity to daily fluctuations in relative humidity, temperature, and wind speed. It can be said that the model can marginally mispredict the moisture content immediately after rain (Dalaris et al. 2013). This misprediction in a higher range of moisture content is not a foremost inconvenience since the applied significance of the lower range with regard to fire risk assessment is profound (Brown et al. 2015).

Fig. 3
figure 3

Duff moisture content assessment (estimated in red, projected in blue)

Full size image

3.2 National Forest Fire Danger Rating System

Three typical Mediterranean plant species (Piplatherum miliaceum, Parietaria diffusa, and Avena sterilis) were tested for the water content best fit valid for KBDI assessment (Table 2).

Table 2 Estimation of the designated plant moisture content using the KBDI system
Full size table

Foliage moisture content of Pinus appeared to be indifferent to the instabilities of the KBDI, reflecting Pinus profound rooting system. Its moisture content reflects the deep moisture status even though the moisture content of the shorter lived and relatively superficial rooted other three herbaceous species reflects the topsoil moisture status which is more susceptible to weather discrepancies (Dolling et al. 2005; Elhag et al. 2018).

When applying KBDI for fluctuations in plant moisture content the variance explained 88% for Piplatherum miliaceum (Fig. 4), 68% for Parietaria diffusa (Fig. 5), and 72% Avena sterillis (Fig. 6). The Student test (Table 3) showed a strong correlation with the KBDI, hence, the estimated and the projected (from the KBDI) moisture content values are significantly indifferent. On the other hand, and according to Parr et al. (2007) and Turco et al. (2016), forecasts of herbaceous plant species moisture content within 20% are methodically profound and acceptable for prescribed forest fire (Dolling et al. 2005). Consequently, the result of this study indicates that moisture content of the three annual plant species monitored can be adequately predicted using the KBDI. The equations developed are applicable only to the region investigated.

Fig. 4
figure 4

KBDI moisture content of Piplatherum milaceum (estimated in red, projected in blue)

Full size image
Fig. 5
figure 5

KBDI moisture content of Parietaria diffusa (estimated in red, projected in blue)

Full size image
Fig. 6
figure 6

KBDI moisture content of Avena sterillis (estimated in red, projected in blue)

Full size image
Table 3 T test for observed and predicted moisture content values
Full size table

Fine fuels can burn even at low values of KBDI that is why “days-since-rain” is a better indicator of actual burning potential than the index, even during the early spring when the KBDI has values below 200. As the index increases up to 300 and above, “days-since-rain” has less importance because of the drying in the deep fuel layers which results in fire intensity increasing. Fuel humidity increases normally each night. Regardless of the location and thus of the difference between day and night humidities, this humidity increase determines the fuel available for fires. At KBDI below 300, the humidity increase during the night can control fires near or above the extinction level. This is not valid when the drought reaches higher levels. On the other hand, the KBDI tends to over predict in the aspect of time the termination of the fire season. This emphasizes the fact that long-term moisture insufficiency in itself cannot be used for predicting precarious fire conditions since fires are a result of the influence of many factors in a combination with drought conditions. Leverkus et al. (2014) point out that there cannot be determined a specific “threshold value” at which forest fires may or may not progress. Jimenez-Gonzalez et al. (2016) found that drought indexes are a good measure of the state of deep organic layers; nevertheless, their study conveys a mixed message: that long-term weather has some effect on forest fire activity, but does not provide much information by itself since many other factors are involved in this phenomenon. The present results support such conclusions since the KBDI values were poorly correlated to forest fire incidence (r = 0.24) and area burned by wildfire (r = 0.03) for the investigated period in Chania, Crete.

4 Conclusions and recommendation

The tested CFFDRS in the Mediterranean environment (Crete, Greece) showed several tendencies in the meteorological fire risk assessment method usage, following the mandatory modifications and concerning its abilities as a guide to fire conditions. Although different meteorological conditions have been covered during the study period, it seems to be highly unlikely that the data include the predictable range of fuel moisture and weather discrepancies found in the Mediterranean climate.

The developed separate models for Piplathemm miliaceum, Parietaria diffusa and Avena sterillis showed that the National Forest Fire Danger Rating System predicted with high accuracy in their moisture content. It can be concluded that the index was appropriate, to a confident degree, to signify the moisture content in the topsoil layers for the trees with the shallow root system, while it was inapplicable for forecasting the needles moisture content of trees with deep root system such as Pinus halepensis. We suggest that more species that are representative, in terms of abundance and dominance, for the Mediterranean region should be used for monitoring of their foliage moisture content. The relationship between moisture content and the KBDI index would contribute predictive models and would ease to settle phases with conditions for forest fire management.