This study presents a dynamic model for the prediction of diurnal changes in the moisture content of dead surface fuels in normally stocked Calabrian pine stands under varying weather conditions. The model was developed based on several empirical relationships between moisture contents of dead surface fuels and weather variables, and calibrated using field data collected from three Calabrian stands from three different regions of Turkey (Mugla, southwest; Antalya, south; Trabzon, north-east). The model was tested and validated with independent measurements of fuel moisture from two sets of field observations made during dry and rainy periods. Model predictions showed a mean absolute error (MAE) of 1.19% for litter and 0.90% for duff at Mugla, and 3.62% for litter and 14.38% for duff at Antalya. When two rainy periods were excluded from the analysis at Antalya site, the MAE decreased from 14.38% to 4.29% and R2 increased from 0.25 to 0.83 for duff fuels. Graphical inspection and statistical validation of the model indicated that the diurnal litter and duff moisture dynamics could be predicted reasonably. The model can easily be adapted for other similar fuel types in the Mediterranean region.
The ignition, growth and development of forest fires are highly dependent on the availability and conditions of forest fuels (
In general, two different modeling approaches have been used to predict moisture content of dead fuels: empirical and process-based models (
Regression and process-based models may, on their own right, provide very successful results. However, limitations and complexities of existing models limit their use under different conditions. The power of statistical and physical models may be increased by combining empirical and theoretical modeling approaches (
In the proposed model, environmental conditions determine the surface fuel moisture content, and moisture dynamics are considered to depend mostly on weather variables such as relative humidity, air temperature, wind speed and precipitation (
The objective of this study is to develop a dynamic model to predict surface fuel moisture dynamics on an hourly basis during dry and rainy periods in a “standard fuel type” characterized by fully grown, normally stocked Calabrian pine (
Many efforts have been made to predict moisture content of surface fuels in
A state-dependent model of surface fuel moisture dynamics was developed. The model can be considered a blend of empirical and process-based models. In that, it utilizes empirical relationships to quantitatively describe functional relationships that are common to a wide range of situations rather than fitting limited moisture responses from a set of measured situations. The model is based on the rate of change of fuel moisture contents depending on the difference between the concurrent fuel moisture and the equilibrium moisture content, and on timelag (
where FMC(t) is the fuel moisture content at time
As can be derived from
A simplified model flow chart is shown in
The model requires values for litter and duff moisture to start simulation. Initial values can either be entered manually or calculated by running the model for three days to establish a reference moisture value, using respective weather data with an arbitrary initial moisture content value of thirty-five percent (
The amount of rain is important to consider when estimating dead fuel moisture content as it can raise surface dead fuels moisture content more rapidly than any other weather variable. Surface fine dead fuels, especially litter, react very rapidly to rain and reach the saturation point quickly (
where R is the amount of rain (mm), SMC is the saturation moisture content (%), FMC is the fuel moisture content (%),
while RRFS is the rain amount required for saturation (mm), obtained as follows (
where FL is the fuel load (litter or duff, kg m-2), and
SMC is the maximum moisture content attainable for a given fuel layer (litter of duff). RRFS refers to the amount of rain required for complete saturation of dry fuels. The first part of
RRFS relationship and SMC values were determined in relation to average moisture content values obtained from wetting experiments (see below). SMC values were taken as 150 and 300% for surface litter and duff, respectively. These values are comparable with those reported in the literature (
One of the main problems in fuel moisture modeling is the increase of fuel moisture content due to the water condensation on its surface. Condensation of water on surface fuels can be caused by both water deposition from the atmosphere by distillation and the upward water transport from the soil by turbulent diffusion (
where CE is the condensation effect (mm),
where T is air temperature (°C), RH is the relative humidity (%).
In
The increase of fuel moisture content due to condensation (FMIc, %) was then calculated as (
where FL is the fuel load (litter or duff - kg m-2).
Using the estimated moisture increase due to rain and condensation described above, fuel moisture content of litter and duff (FMC(t), %) were estimated at the beginning of the hour as follows (
EMC is an intermediate variable used to predict fuel moisture contents finally attained by the fuels under constant air temperature and relative humidity (
Timelag is the time necessary for a fuel particle to gain or lose approximately 63.3% of the difference between its initial moisture content and its equilibrium moisture content (
where VPD is the vapor pressure deficit (Pa),
where W is the wind speed (km h-1),
Although it may seem somewhat complex, the TLC relationship has a simple inverse sigmoid shape;
Timelag constant is then used along with EMC and initial moisture content of the fuels to calculate the final moisture content values at the end of the hour (FMC(t+1), %) for litter and duff (
The estimated FMC(t+1) values then become the initial values for the next step of the simulation.
To determine the surface fuel moisture content changes under varying weather conditions, measurements were carried out in three geographic locations (
The first study area was in the Yaras State Forest in Mugla, south-western Turkey (37° 08′ N, 28° 30′ E; average elevation 750 m a.s.l.). The area has Mediterranean climate with prolonged dry summers and mild, moist winters. Average air temperature from May to October varies from 15 to 25 °C, and the average monthly rainfall ranges from 7.8 to 68.3 mm for the same period. The fire season in Mugla region generally lasts from late May until mid-September.
The second study area is in Antalya, southern Turkey (37° 08′ N, 28° 30′ E; average elevation 240 m a.s.l.). This area has climatic conditions similar to Mugla, characterized by prolonged dry summers and mild, moist winters. Average air temperature from May to October varies from 20 to 28 °C, and the average monthly rainfall ranges from 3.1 to 80.1 mm for the same period. The fire season in the Antalya region generally lasts from late May until mid-September.
The third study area was in the Trabzon region, north-eastern Turkey (40° 59′ N, 39° 46′ E; av. elev. 50 m a.s.l.). The study area has a Black Sea climate characterized by warm and humid summers, long cool/cold and damp winters with high and evenly distributed rainfall the year round. Average air temperature from November to May varies from 7 to 16 °C and average monthly rainfall is between 58.5 to 96.1 mm for the same period. The fire season in Trabzon differs from other regions and generally lasts from November until early-May.
The study was conducted in pure, normally stocked (~80-90% crown closure), and even-aged Calabrian pine stands. Sampling plots, 20 × 20 m in size, were established in normally-stocked and nearly fully-closed Calabrian pine stands in all three study areas. Plots were set up inside the stands at least 50 m away from open areas to avoid edge effect on fuel moisture. Measured stand characteristics included stand age, diameter at breast height (dbh), crown width, tree height, crown base height, canopy closure, stand density and basal area. Stand age was 41, 63 and 50 years; mean dbh 26.8, 37.1 and 34.2 cm; mean crown width 4.6, 8.3 and 5.6 m; mean tree height 14.0, 17.8 and 15.5 m; mean crown base height 8.7, 10.2 and 8.9 m; mean canopy closure 95, 80 and 85%; mean stand density 725, 250 and 450 stem ha-1; mean stand basal area 40.9, 27.0 and 41.3 m2 ha-1 for Mugla, Antalya and Trabzon, respectively.
No living plants were present in the understory within the stands, and living trees made up 100% of the overstory. Surface fuels consisted primarily of needle litter along with some branch and cone litter. Average litter and duff fuel loads in the measurement plots were 0.335, 0.396 and 0.405 kg m-2 and 1.647, 1.005 and 1.853 kg m-2 at Mugla, Antalya and Trabzon sites, respectively. Stand and surface fuel characteristics are given in
A fully automated weather station (Davis Vantage Pro™, Davis Instruments Corp., Hayward, CA, USA) was set up at the study sites to record weather variables. Measurements involved rainfall (mm), air temperature (°C), relative humidity (%), wind speed (km h-1) and direction. Weather measurements were recorded continuously during the period of fuel measurements. Weather measurements (
Surface fuel moisture measurements were carried out in the sampling plots every two hours between 09:00-19:00. Five fuel samples were taken from each plot for each sampling time. Each sample was obtained from a sub-plot measuring 15 × 20 cm. All fuels within the sub-sampling plot was removed and separated as surface litter (L layer, newly fallen surface litter) and duff (F layer, decomposed and/or decomposing organic matter below L layer). Fuel samples were sealed in plastic containers, weighed with 0.01 g precision and taken to the laboratory. Measurements continued until the end of the study period. All litter and duff samples were oven-dried at 105 °C for 24 h and weighed to obtain fuel moisture contents (
Some descriptive statistics of weather and fuel moisture are provided in
To determine moisture content increase after a rain event, a wetting experiment was further conducted immediately before, during and after a rain event. To determine fuel moisture increase after wetting, fuel containers measuring 15 × 20 cm in size were prepared for litter and duff using nets of 1 × 1 mm mesh size. Containers were placed in plots measuring 1 × 1 m with three replicates for litter and duff. To wet the samples, tap water was used in quantities of 0.2, 0.5, 1.0, 2.0, 5.0, 7.0 and 10.0 liters (rain equivalent of 0.2 mm to 10 mm). Uniform artificial raindrops were produced over the 1 × 1 m study plot by a water sprinkler (
Furthermore, litter and duff samples were taken from the study plots in Trabzon before, during and immediately after a rain event. As a result, a total of 13 rain events over 8 different days were documented throughout the study period.
Parameter values for fuel moisture increase due to rain, condensation, timelag constant, wind effect on timelag were obtained by fitting the model by non-linear least squares regression (
After calibration, the model was further tested against the independent data sets from Mugla (M2) and Antalya study sites for validation.
To assess the accuracy of model predictions (
All statistical analyses were conducted using the software package SPSS® ver. 22.0 (
Data used for calibration and validation of the model covered a significant range of weather conditions (
The relationships between measured and predicted litter and duff moisture contents for Mugla and Antalya sites are given in
The visual and statistical evaluation of the results indicated that the accuracy of model predictions for both litter and duff moisture was satisfactory for Mugla and Antalya sites. However, we observed some discrepancies between the observed and predicted values of duff at the Antalya site, where the model systematically underestimated duff moisture contents during the period April 20-23 and April 28-29 (
The validation statistics of the comparison between predicted and observed fuel moisture content values for the two sites are given in
As for the rainfall effect, despite there were only three rainfall occurrences with limited amount of precipitation, the response and performance of the model in case of a rain event was also satisfactory (
The proposed model produced reasonable fits to the validation data. As expected, the accuracy of model predictions was higher during dry periods when moisture content is low, though litter moisture dynamics were predicted reasonably well under all weather conditions (
The accuracy of model predictions in Mugla was similar to other studies (
(i) the range of the moisture content predictions: the range of LMC was 5.48-51.81% and that of DMC 9.07-81.06% for the Antalya site, covering rainy and dry periods (
(ii) Slight differences in stand structure and surface fuel characteristics (
(iii) Potential soil moisture effect not accounted for: the effect of soil moisture on the moisture content of surface fuels has been clearly demonstrated (
In this study, an attempt was made to predict diurnal surface fuel moisture dynamics in Calabrian pine stands in Turkey by developing a deterministic dynamic model assembled from empirical relationships over a wide range of weather situations. Model inputs were hourly air temperature, relative humidity, wind speed and the amount of rain readily measured in the field. The performance of the model under varying weather conditions and different stand characteristics was reasonably accurate for dry periods or periods with occasional rain events. Lower performances of the model at the Antalya site was probably caused by the presence of high soil moisture affecting surface fuel during the first three days of observation.
The proposed model has been developed for fire danger rating predictions in fully grown (tree height: 15-20 m), normally stocked (basal area: 30-50 m2 ha-1; crown closure: 80-100%) pine stands. Its application to other Mediterranean stand types should be performed with caution, and only when specific model parameters (
More experimental research on the effects of soil moisture on surface fuel moisture content dynamics are needed for future model development. Moreover, future efforts should further improve model accuracy by testing the model over a wider range of stand characteristics and conditions.
We would like to extend our appreciation and thanks to Mugla and Antalya Regional Forest Directorate and its staff. This study was supported by The Scientific and Technological Research Council of Turkey, project no. TOVAG-112O809. We are grateful to two anonymous reviewers for their useful suggestions and comments that greatly improved the manuscript.
Flowchart of the model.
Geopraphic locations of study sites and sampling plots; Mugla (a), Antalya (b) and Trabzon (c). Dashed areas in the map shows the geographical distrubition of Calabrian pine in Turkey.
Temperature, relative humidity, rainfall, wind speed and vapor pressure deficit values during the study for Mugla (M1 and M2) and Antalya sites.
Time series of hourly predicted and observed litter (a, c) and duff (b, d) moisture contents for the period of August 14-27, 2014 in Mugla (a, b) and April 20-29, 2014 in Antalya (c, d), respectively.
The relationships between measured and predicted litter (a, c) and duff (b, d) moisture contens for Mugla and Antalya.
Definition of variables used in model specification.
Variable | Description | Unit |
---|---|---|
CE | Condensation effect on surface fuels | mm |
EMC | Equilibrium moisture content | % |
FMIr | Fuel moisture increase due to rain | % |
FMIc | Fuel moisture increase due to condensation | % |
SMC | Saturation moisture content | % |
FMC | Fuel moisture content | % |
R | Rainfall | mm |
RRFS | Rain required for saturation | mm |
RH | Relative humidity | % |
T | Air temperature | °C |
TLC | Timelag constant | hour |
W | Wind speed | km h-1 |
VPD | Atmospheric vapor pressure deficit | Pa |
FL | Fuel load | kg m-2 |
Mean structural and surface fuel characteristics at the study plots.
Feature | Variable | Units | Mugla | Antalya | Trabzon | |
---|---|---|---|---|---|---|
Stand | Stand Origin | - | Plantation | Natural | Plantation | |
Stand Age | Year | 41 | 63 | 50 | ||
Number of Plots | # | 3 | 3 | 2 | ||
DBH (d1.30) | cm | 26.8 | 37.1 | 34.2 | ||
Crown Width | m | 4.6 | 8.3 | 5.6 | ||
Tree Height | m | 14.0 | 17.8 | 15.5 | ||
Crown Base Height | m | 8.7 | 10.2 | 8.9 | ||
Canopy Closure | % | 95 | 80 | 85 | ||
Stand Density | Stem ha-1 | 725 | 250 | 450 | ||
Stand Basal Area | m2 ha-1 | 40.9 | 27.0 | 41.3 | ||
Surface Fuel | Litter | Fuel Depth | cm | 1.8 | 2.1 | 2.5 |
Fuel Load | kg m-2 | 0.335 | 0.396 | 0.405 | ||
Fuel Density | g cm-3 | 0.019 | 0.020 | 0.016 | ||
Duff | Fuel Depth | cm | 4.1 | 3.2 | 3.9 | |
Fuel Load | kg m-2 | 1.647 | 1.005 | 1.853 | ||
Fuel Density | g cm-3 | 0.040 | 0.031 | 0.048 |
Weather and fuel moisture variables for the study areas. (M1, M2, A, T): different study periods at the three sampling sites (see text); (a): wind speed was measured at 10 m standard height in the open ground (
Feature | Variable | Stats | Mugla | Antalya | Trabzon | Trabzon WE (b) | |
---|---|---|---|---|---|---|---|
M1 | M2 | A | T | T | |||
Weather | Temperature(°C) | Min | 17.1 | 19.9 | 8.6 | 2.4 | - |
Mean | 26.7 | 27.2 | 17.2 | 13.1 | - | ||
Max | 38.2 | 37 | 25.5 | 20.4 | - | ||
Relative Humidity (%) | Min | 16 | 10 | 26 | 8 | - | |
Mean | 35.4 | 44.5 | 73.4 | 72.4 | - | ||
Max | 71 | 69 | 95 | 96 | - | ||
Wind Speed (a)(km h-1) | Min | 0.2 | 0 | 0 | 0 | - | |
Mean | 7.4 | 1.9 | 2.1 | 4.2 | - | ||
Max | 22.7 | 6.3 | 8 | 20.9 | - | ||
Rainfall (mm) | Max | 0 | 0 | 1.2 | 2.8 | - | |
Fuel Moisture | Litter Moisture(%) | Min | 3.6 | 4.8 | 5.5 | 15.4 | 22.8 |
Mean | 8.8 | 8.4 | 17.9 | 24.7 | 72.7 | ||
Max | 17.2 | 11.6 | 51.8 | 50.7 | 160.2 | ||
SD | 2.9 | 1.4 | 17.8 | 9.3 | 21.9 | ||
Duff Moisture(%) | Min | 6.3 | 5.7 | 9.1 | 44.2 | 71.2 | |
Mean | 8.1 | 9.2 | 32.3 | 67.8 | 160.9 | ||
Max | 10.3 | 12.8 | 81.1 | 98.9 | 360.3 | ||
SD | 1 | 1.9 | 24.6 | 13 | 59.2 |
Parameter values used in the model.
Variable | ModelParameter | Value | |
---|---|---|---|
Litter | Duff | ||
Fuel moisture increase due to rain (FMIr), % |
|
25 | 30 |
|
2.8 | 3.5 | |
Rain required for saturation (RRFS), mm |
|
50 | 40 |
|
1 | 1 | |
Condensation effect (CE), mm |
|
0.07 | 0.25 |
|
15 | 20 | |
|
1.4 | 2 | |
Timelag (TLC), hour |
|
50 | 75 |
|
50 | 75 | |
|
3 | 4 | |
|
3 | 3 | |
Wind effect on TLC (kw), km h-1 |
|
0.3 | 0.1 |
|
10 | 10 |
Statistics of the comparison between observed litter (LMC) and duff (DMC) moisture content values with validation data. (MAE): mean absolute error; (MAPE): mean absolute percentage error; (R2adj): adjusted coefficient of determination; (SE): standard error. (*): DMC values were excluded from the validation data set at Antalya during the periods April, 20-23 and April, 28-29 when soil moisture was high.
Area | Number of measures | MAE | MAPE | R2adj | SE | |||||
---|---|---|---|---|---|---|---|---|---|---|
LMC | DMC | LMC | DMC | LMC | DMC | LMC | DMC | LMC | DMC | |
Mugla (M2) | 81 | 81 | 1.19 | 0.90 | 14.68 | 10.83 | 0.66 | 0.84 | 0.80 | 0.76 |
Antalya | 55 | 55 | 3.62 | 14.38 | 22.19 | 38.73 | 0.67 | 0.25 | 5.41 | 15.06 |
Antalya* | - | 27 | - | 4.29 | - | 19.55 | - | 0.83 | - | 4.93 |
Fig. S1 - The relationship between fuel moisture contents and temperature and relative humidity.