Canadian Journal of Forest Research, 2015, 45: xxx-xxx doi:10.1139/cjfr-2014-0545
Measuring foliar moisture content with a moisture analyzer Carlos G. Rossa1, 2, Paulo M. Fernandes1, and Anita Pinto1 1
Centre for the Research and Technology of Agro-environmental and Biological Sciences (CITAB), University of Trás-
os-Montes and Alto Douro, Apartado 1013, 5001-801 Vila Real, Portugal 2
Corresponding author (e-mail:
[email protected]).
Abstract Near-instantaneous estimation of the moisture content of live fuels is complicated because of the large control exerted by physiological mechanisms. The commonly accepted reference method for measuring fuel moisture content is oven drying, which is time consuming. This study evaluates the use of a moisture analyzer (ML-50, A&D Company Ltd.) for measuring the foliar moisture content of two common European species. The moisture of live leaves of Arbutus unedo L. (strawberry tree) and Quercus robur L. (pedunculate oak) was measured within a period of 15 min using two drying temperatures and compared with the oven-dry value. Correction factors were determined for estimating oven-dry moisture content based on the analyzer measurement. The power delivered during the drying process plays an important role in the moisture measured by the analyzer in relation to the oven-dry value. Increasing the drying time beyond the minimum period necessary for obtaining a reliable prediction of the oven-dry moisture does not significantly change the moisture measured at lower temperatures. The moisture analyzer is able to estimate live foliage moisture content with high accuracy. Keywords: fuel moisture content, live fine fuels, correction equations.
Introduction Fuel moisture content measures the amount of water, normally expressed as a percentage of its dry weight. It plays a key role in forest fire behaviour (Rothermel 1983) given that, as the mass of water increases so does the energy necessary for pre-heating and igniting the fuel. While in dead fine fuels the moisture content is dominated by the ambient conditions (Simard 1968), in live fuels it is largely a function of physiological mechanisms within the plant (Nelson 2001) and therefore is mostly independent from ambient influences. Although dead fuel fibres saturate at about 35% moisture content (Cheney 1981), after which free water will appear on the surface, the moisture 1
content of live foliage often exceeds 100%. Even in situations where live fuels are not able to sustain combustion by themselves due to high moisture content, the less water they have the less energy will be required for ignition and more energy will be available to ignite other fuel particles, increasing the fire rate of spread (McArthur 1967). Large fire activity in shrub-dominated Mediterranean climate regions can be related with thresholds in live fuel moisture content (Dennison et al. 2008), justifying the existence of moisture content sampling networks for fire management purposes (Weise et al. 1998). In contrast with dead fuels (Viney 1991, Matthews 2013) few methods exist to assess the moisture of live fuels (Chatto and Tolhurst 1997). Indirect estimates of live fuel moisture content can be obtained from ambient weather and drought conditions, fire danger rating systems and remote sensing indices, albeit with considerable limitations and uncertainty brought about by variability in vegetation type and site conditions (Burgan 1979, Yebra et al. 2013). These estimates are crude and reliable assessment of live fuel moisture content implies a measurement. Fuel ovendrying at a given temperature for a given time period, e.g., at 105ºC for 24 h, is the reference method for moisture content determination (Matthews 2010). Because oven drying requires a long period of time for retrieving moisture content, simple and expedite measurement methods that allow accurate estimation of fuel moisture content are desirable. Chatto and Tolhurst (1997) mention two portable devices that measure moisture content within a range wide enough to include live fuels and provide a measurement within a few minutes. One is a portable oven by Neosystems that compresses the fuel sample between two hot plates and a vacuum pump removes the moisture from air. The second method is the Wiltronics Fine Fuel Moisture Meter that is based on the measurement of the electrical resistance of the fuel sample. Jolly and Hadlow (2012) assessed the use of a rapid moisture analyzer named Computrac Max2000XL (Arizona Instruments), which uses a radiant heater for drying the fuel. This paper addresses the use of a commercially available moisture analyzer for measuring the moisture content of live leaves. Water mass loss in time is assessed and correction factors are given for estimating the oven-dry moisture based on the value measured by the analyzer. Material and methods Part of the experiments presented in this paper were carried out in the University of Trás-os-Montes and Alto Douro (UTAD, Vila Real, Portugal), including some preliminary testing that was used to tune the proceedings and to adjust the test parameters. The remaining experiments proceeded in the 2
Agriculture Higher School of Coimbra (ESAC, Portugal). The tests whose data will be analysed in this work were conducted throughout July 2014. Moisture analyzer The ML-50 moisture analyzer (A&D Company Ltd., Tokyo, Japan) (Fig. 1) uses the electric current as a power source, is 32-cm long, 21.5-cm wide, 17.3-cm tall and weighs approximately 6 kg. It is basically a precision balance with a circular sample pan 8.5-cm wide, that weighs samples up to 51 g with an accuracy of 0.005 g and dries them using a 400 W halogen lamp. The drying temperature Td can be varied in the 50–200ºC range with 1ºC increments and the moisture content can be determined with an accuracy of 0.5% for samples in the 1–5 g range and 0.1% for samples above 5 g. During the drying process the instantaneous drying rate is displayed with a 0.01% min-1 precision. It is possible to dry the samples in an automatic mode, in which the analyzer stops the test when the drying rate drops below a given value, or using a timer mode, in which the test is stopped after a selected period of time. We used the timer mode because when the drying rate falls below about 0.1 % min-1 (wet basis) its value becomes inconsistent. Two situations could happen if the automatic mode was used: i) under a relatively high drying rate, e.g., 0.5% min-1, the analyzer could instantly attain that value and stop the test, hindering further sample drying; ii) under a very low drying rate, e.g., 0.1% min-1, the test could run for considerable time without significant changes in the moisture content before that rate was instantaneously attained. Also, one of the objectives was to assess the evolution of water content with time, and a specific drying period served that purpose better. Fuel sampling and test parameters Live leaves of Arbutus unedo L. (strawberry tree) and Quercus robur L. (pedunculate oak) were used. We deliberately choose two species contrasting in their foliar physical characteristics. A. unedo is an evergreen shrub or small tree from the Mediterranean region with dark green and glossy sclerophyllous leaves 5–10 cm long, 2–3 cm broad and about 0.4 mm thick. Q. robur is a Western Europe deciduous tree with short-stalked leaves 7–14 cm long, 4–5 lobes with smooth edges and about 0.17 mm thick. The leaves were collected wearing gloves and using a scissor, from trees shaded all day long, so that fuel moisture could not be directly influenced by solar radiation. The fuel was placed in a thermally insulated bag, which was placed inside a thick black plastic bag to block all solar 3
radiation. Only leaves of one species were collected at a time, carrying an amount enough for two experiments, one at a drying temperature of 150ºC and another at 200ºC. The fuel arrived to the laboratory within less than 10 min. Preliminary tests were made at drying temperatures of 105ºC, the standard value used in ovendrying; 200ºC, the maximum value the analyzer can attain; and 150ºC, an intermediate value. We concluded that the 105ºC temperature was insufficient for properly drying live fuel samples over a period of 15 min. Having in mind the study objectives and the results from preliminary tests we chose the drying temperatures of 150ºC and 200ºC over a 15-min period. For each combination of fuel species and temperature 32 experiments were carried out, totalling 128 tests. Fuel moisture measurement The experiments involved the preparation of two samples, as we intended to compare the moisture content measured with the analyzer (test sample, about 1 g) with that determined by ovendrying at 105ºC during 24 h (control sample, about 3–4 g). The bags were open only for a few seconds, for retrieving some leaves, and rapidly closed to avoid moisture loss as much as possible. The leaves were cut in small pieces with a scissor (because the analyzer was conceived to work with granular material) into the pan of the analyzer and into an aluminium container to be put in the oven. No effort was made to estimate a moisture content representative of the vegetation from where the samples were retrieved. The main concern was to assure that fuel from the test and control samples was identical, although the control sample contained more fuel. To achieve that purpose, small pieces from the same leaf were cut into the pan and big pieces were cut into the container. We also assured that the pan had cooled, in case the analyzer had been previously used. A warm pan can significantly change fuel moisture during sample preparation. Some test samples were purposely larger to evaluate if sample mass mf would significantly influence the difference between the oven-dry and the analyzer fuel moisture contents. The control sample was weighed and put in the oven, fuel in the test sample was uniformly distributed and the pan was placed in the analyzer (Fig. 1). The drying program was initiated after recording the initial mass of the test sample. Fuel moisture content was registered every minute, except for 22 experiments with A. unedo (11 dried at 150ºC and 11 dried at 200ºC) where the records were only taken at 5, 10 and 15 min. Moisture content was displayed in a wet basis (i.e., as 4
a percentage of the wet weight of the fuel, Mwb) and used to calculate moisture on a dry basis Mdb using eq. 1: [1]
M db =
100 × M wb 100 − M wb
The control samples were weighed the day after and the moisture content resulting from ovendrying was then calculated. Data analysis For each fuel type we determined the minimum, mean, median and maximum values of the test and control samples mass and oven-dry moisture content of the control samples. The distributions were checked for normality using a Shapiro-Wilk’s test (P > 0.05) and when the significance was below the threshold value the histograms, Q-Q and P-P plots were visually inspected to verify if they were approximately normally distributed. The initial mass of water mw of the test samples depends on fuel mass mf and its moisture content. For each drying temperature Td, assuming that the power supplied by the halogen lamp is constant, the time t necessary for evaporating all water in the fuel depends essentially on mw. As each sample had a different initial mw, the time scale of the tests was adjusted to compare water mass evolution with time between experiments, using the following procedure for each Td: i) the experiment with the highest initial value of mw was selected as a reference; ii) in the remaining tests, the initial time instant was adjusted and instead of zero assumed the time instant where the reference test attained mw approximately equal to the initial sample mw. For the reference experiment, mw values intermediate between the minute observations were estimated by linear interpolation, so that the time scale adjustment had a precision better than 0.01 min (0.6 s). Water loss with time was modelled using an exponential decaying curve (eq. 2): [2]
(
m w = a⋅ exp − b ⋅ t c
)
Goodness of fit was assessed using the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE) and mean bias error (MBE) (Willmott 1982). For each drying temperature and drying period, stepwise linear regressions were fitted with oven-dry moisture content Mod as the dependent variable and the moisture measured by the analyzer Man and the mass of the fuel samples mf (test and control) as the putative independent variables. Goodness of fit was assessed with the coefficient of determination (R2) and standard error (SE), 5
using the regression significance (P-value) to evaluate if the number of samples was sufficient to support the results. The approximately normal distribution of the residuals was verified by visual inspection to their histograms, Q-Q and P-P plots. The independence of the residuals from the predicted values and from the test number (independence over time) was verified through correlation analysis. Correlations were considered significant at P < 0.05. Results and discussion Fuel samples The mf histograms of the test samples were positively skewed for both fuel types (Fig. 2) and in such cases the median (Table 1) is a better indicator of central tendency than the mean, yielding an average sample mass of the order of 1.1 g. Regarding the control samples, Q. robur leaves mf was normally distributed (P = 0.11) and that of A. unedo leaves was approximately normally distributed (P = 0.04). Mean sample mass was 3.2 and 3.8 g for A. unedo and Q. robur leaves, respectively (Table 1). A three-fold variation was found for the oven-dry moisture content of A. unedo control samples (82–244%), contrasting with Q. robur (115–166%). The moisture of Q. robur followed a normal distribution (P = 0.34), which was not the case for A. unedo (Fig. 3). Drying curves When Td = 200ºC there were no observable differences between the drying patterns of A. unedo and Q. robur leaves (Fig. 4), but Q. robur samples dried slightly faster when Td = 150ºC. The leaves of A. unedo are thicker and made of harder tissue, hence likely to offer more resistance to water loss. Those differences were not influent when the drying temperature was high enough, i.e., 200ºC. The drying curves are specific to the tests made in this study and should only be valid for fuel samples with a water content within the observed experimental range and with a thickness of the same order. As expected from Fig. 4, eq. 2 fit is very good for Td = 200ºC and a drying curve for each fuel type is not needed (Table 2). The results are quite good also for Td = 150ºC and, even if separate models have better fit, a joint curve still makes sense since it is more general. An expression for the drying rate can be determined by deriving eq. 2 (eq. 3). If we consider the curve for both fuels (Table 2) for Td = 150ºC and Td = 200ºC we obtain maximum drying rates of 0.33 g min-1 and 0.47 g min-1, respectively. Water loss is practically negligible after a given period 6
of time and the measured fuel moisture does not further change significantly. A drying rate of 0.001 g min-1 represents a 0.1% min-1 change in the moisture content of 1 g of dry fuel, which means that the analyzer value will suffer insignificant or no changes once this rate is achieved. If we define this as a criterion for determining the optimum drying time, that yields 11 min for Td = 150ºC and 7 min for Td = 200ºC.
dmw = −a ⋅ b ⋅ c ⋅ t c−1 exp − b ⋅ t c dt
(
[3]
)
Correction factors For each drying period (t = 1–15 min), in the stepwise regressions that allowed obtaining an estimate of Mod with R2 > 0.9 (e.g., Fig. 5) the first and second variables included were Man and mf of the control samples. It makes sense that oven-dry moisture content has some correlation with the control samples mass that increases with higher water content. Addition of test samples mass as a variable produced very little improvement in R2, showing its small influence in the quality of the estimates. With no exception, Man absorbed almost all of Mod variation. In fact, the maximum R2 increase obtained by including more variables beyond Man was of the order of 0.01. For that reason and for the sake of parsimony we decided to use Man as the single independent variable in the regressions. Also, because in all regressions the intercept was very close to zero and did not influence the goodness of fit significantly it was not included in the regression equations. Therefore, for each Td and for a given drying period, the oven-dry moisture content was estimated by multiplying the analyzer measurement by a correction factor F (eq. 4): [4]
M od = F ⋅ M an Fig. 5 shows a typical fit of oven-dry moisture content to the analyzer moisture content using eq.
4. Table 3 gives the parameters and performance evaluation for regressions with R2 > 0.9, which yielded a minimum drying period of 8 min for Td = 150ºC and 5 min for Td = 200ºC. The P-value was always < 0.0001, showing that the number of tests was sufficient to support the results. The residuals were approximately normally distributed and their independence from the predicted values and from the test number was confirmed by the absence of significant correlations. As expected, the correction factor is lower for the higher drying temperature (Fig. 6). For Td = 150ºC an increase in the drying period does not significantly improves the analyzer measurement, whereas F decreases progressively with time when Td = 200ºC and for t > 10 min a correction factor is practically unnecessary. This means that, in this fast drying process, the same amount of energy 7
used for removing water from the fuel produces different results in case it is delivered in a short period of time (higher power) or in a longer period of time (lower power). So, F tends to an asymptote of about 1.08 for Td = 150ºC and for Td = 200ºC it tends to unity. When Td = 150ºC the moisture measurement does not change significantly beyond the previously determined optimum drying period of 11 min, given that F remains practically constant. In the case of Td = 200ºC, although F still decreases beyond the optimum t of 7 min, its variation becomes minimal. Comparison with other methods The alternatives to the ML-50 analyzer mentioned by Chatto and Tolhurst (1997) have practical disadvantages. The Neosystems oven needs a car battery as a power source and has a warm-up waiting time of approximately 15 min. The Wiltronics Fine Fuel Moisture Meter requires a ground sample and calibration for each fuel type. Although the ML-50 analyzer was not designed as a portable system, given its dimensions and weight, it can be used in the field provided that a source of electric power is available. The Computrac Max2000XL moisture analyzer (Jolly and Hadlow 2012) operates similarly to the ML-50 and also measures the moisture content within about 15 min with similar accuracy and therefore is a valid alternative. Part of the more volatile components will evaporate at drying temperatures of 150 and 200ºC but such fact does not compromise the results, as also found by Jolly and Hadlow (2012) who measured the moisture content of Pinus contorta Douglas (lodgepole pine) needles, representative of volatilerich conifer live foliage. Conclusions We have assessed the use of a moisture analyzer (ML-50, A&D Company Ltd.) for measuring the moisture content of live foliage. The moisture of live leaves of A. unedo and Q. robur was measured at drying temperatures of 150 and 200ºC for 15 min and compared with the oven-dry value. A model was proposed for describing water mass loss with time and correction factors were determined for estimating the oven-dry moisture content based on the value measured by the analyzer. Both drying temperatures allow accurate estimates of the oven-dry moisture. Q. robur leaves dry slightly faster than A. unedo leaves at 150ºC but at 200ºC no significant differences are observed between the fuels. Defining as a criterion the attainment of a rate of water mass loss of
8
0.001 g min-1, we obtain an optimum drying period of 11 min for a drying temperature of 150ºC and 7 min for 200ºC. The power delivered during the drying process plays an important role in the moisture measured by the analyzer. At 150ºC a drying time increase beyond the minimum period necessary for obtaining a reliable prediction of the oven-dry moisture (8 min) does not produce a significant change in the measurement and the correction factor is always about 1.08. At 200ºC the measurements are always closer to the oven-dry value and for a drying period greater than 10 min the correction factor is always less than 1.01. The ML-50 moisture analyzer provides accurate and robust user-friendly estimates of the moisture content of live foliage and thus can be recommended for both fire management and fire research applications. Acknowledgements The authors acknowledge Joaquim Sande Silva, Rosinda Pato and Jorge Bandeira from the Agriculture Higher School of Coimbra (ESAC), for having made possible the use of the laboratory where part of the experimental program was carried out. This research was funded by the Fundação para a Ciência e a Tecnologia under post-doctoral grant SFRH/BPD/84770/2012. References Burgan, R. 1979. Estimating live fuel moisture for the 1978 National Fire Danger Rating System. USDA For. Serv. Res. Pap. INT-226. Chatto, T., and Tolhurst, K. 1997. The development and testing of the Wiltronics T – H fine fuel moisture meter. Vict. Dep. Nat. Resourc. Environ., Fire Manag. Branch Res. Rep. 46. Cheney, N.P. 1981. Fire behaviour. In Fire and the Australian Biota. Edited by A.M. Gill, R.H. Groves, and I.R. Noble. Australian Academy of Science, Canberra, Australia. pp. 151–175. Dennison, P.E., Moritz, M.A., and Taylor R.S. 2008. Evaluating predictive models of critical live fuel moisture in the Santa Monica Mountains, California. Int. J. Wildland Fire 17: 18–27. doi:10.1071/WF07017 Jolly, W.M., and Hadlow, A.M. 2012. A comparison of two methods for estimating conifer live foliar moisture content. Int. J. Wildland Fire 21: 180–185. doi:10.1071/WF11015 Matthews, S. 2010. Effect of drying temperature on fuel moisture content measurements. Int. J. Wildland Fire 19: 800–802. doi:10.1071/WF08188 9
Matthews, S. 2013. Dead fuel moisture research: 1991–2012. Int. J. Wildland Fire 23(1): 78–92. doi:10.1071/WF13005 McArthur, A.G. 1967. Fire behaviour in eucalypt forests. Forestry and Timber Bureau Leaflet 107, Canberra, Australia. Nelson, R.M. 2001. Water relations of forest fuels. In Forest Fires: Behaviour and ecological effects. Edited by E.A. Johnson, and K. Miyanishi. Academic Press, San Diego, Calif. pp. 79– 149. Rothermel, R.C. 1983. How to predict the spread and intensity of forest and range fires. USDA For. Serv. Res. Pap. INT-143. Simard, A. 1968. The moisture content of forest fuels – II comparison of moisture content variations above the fibre saturation point between a number of fuels. For. Can. Inf. Rep. FF-X156. Viney, N.R. 1991. A review of fine fuel moisture modelling. Int. J. Wildland Fire 1(4): 215–234. doi:10.1071/WF9910215 Weise, D.R., Hartford, R.A., and Mahaffey, L. 1998. Assessing live fuel moisture for fire management applications. In Proceedings of the 20th Tall Timbers Fire Ecology Conf., Fire in Ecosystem Management: Shifting the Paradigm from Suppression to Prescription, 7–10 May 1996, Tallahassee, Florida. Edited by T.R. Pruden, and L.A. Brennan. Boise, Idaho. pp. 49–55. Willmott, C.J. 1982. Some comments on the evaluation of model performance. B. Am. Meteorol. Soc. 63: 1309–1313. Yebra, M., Dennison, P.E., Chuvieco, E., Riaño, D., Zylstra, P., Hunt, E.R., Damson, F.M., Qi, Y., and Jurdao, S. 2013. A global review of remote sensing of live fuel moisture content for fire danger assessment: moving towards operational products. Remote Sens. Environ. 136: 455–468. doi:10.1016/j.rse.2013.05.029 List of symbols a, b, c (–), parameters used in eq. 2 and 3. F (–), correction factor for the moisture measured by the analyzer. mf (g), initial mass of the fuel samples (wet fuel). mw (g), mass of water. Mdb (%), fuel moisture content (dry basis). Mwb (%), fuel moisture content (wet basis). 10
Man (%), fuel moisture content determined by using the moisture analyzer (dry basis). Mod (%), fuel moisture content determined by oven-drying (dry basis). n (–), number of experiments. t (min), drying time. Td (ºC), drying temperature.
11
Tables Table 1. Descriptive statistics for fuel mass mf and oven-dry moisture content Mod of the samples. n Fuel type A. unedo
64
Q. robur
64
mf (g) test samples Mean Min.-max. Median (s.d.) 1.19 0.69-2.45 1.07 (0.42) 1.18 0.69-2.34 1.10 (0.35)
mf (g) control samples Mean Min.-max. Median (s.d.) 3.23 1.76-5.69 2.94 (0.99) 3.80 2.39-5.63 3.66 (0.78)
Mod (%) control samples Mean Min.-max. Median (s.d.) 149.0 82.2-243.8 158.9 (5.9) 138.8 114.6-166.1 139.5 (1.4)
Table 2. Parameters and statistical measures of performance of the model for estimating water loss with time (eq. 2). n
Model parameters a b c
Data set Td = 150ºC (A. unedo + Q. robur) Td = 150ºC (A. unedo) Td = 150ºC (Q. robur) Td = 200ºC (A. unedo + Q. robur)
848 336 512 848
1.32 1.25 1.33 1.41
0.18 0.10 0.20 0.17
1.55 1.78 1.57 1.92
R
2
0.960 0.982 0.976 0.993
Model evaluation RMSE MAE (g)
MBE (g)
0.043 0.032 0.030 0.018
0.0002 0.0015 -0.0010 -0.0040
0.020 0.016 0.013 0.010
Table 3. Parameters and statistical measures of performance of the linear regressions for oven-dry fuel moisture as a function of analyzer moisture content (eq. 4). F, regression coefficient corresponding to the correction factor for the moisture measured by the analyzer (for all regressions P < 0.0001); SE, standard error of the estimates. Drying temperature
Td = 150ºC
Td = 200ºC
t (min)
n
Regression parameters F SE
8 9 10 11 12 13 14 15 5 6 7 8 9 10 11 12 13 14 15
53 53 64 53 53 53 53 64 64 53 53 53 53 64 53 53 53 53 64
1.094 1.087 1.088 1.083 1.082 1.082 1.081 1.085 1.052 1.036 1.028 1.021 1.017 1.017 1.009 1.006 1.004 1.001 1.003
0.006 0.005 0.004 0.005 0.005 0.005 0.005 0.004 0.005 0.005 0.005 0.005 0.005 0.004 0.005 0.005 0.005 0.005 0.004
Regression evaluation R2 SE (%) 0.942 0.957 0.980 0.963 0.963 0.964 0.964 0.981 0.977 0.960 0.964 0.964 0.965 0.983 0.965 0.964 0.964 0.963 0.983
5.93 5.11 4.64 4.74 4.75 4.70 4.69 4.52 5.56 5.00 4.79 4.80 4.72 4.79 4.72 4.79 4.75 4.83 4.84
12
Figure captions Fig. 1. A&D ML-50 moisture analyzer with a sample of live leaves of A. unedo in the pan. Fig. 2. Histograms of the initial mass (wet fuel) of the test and control fuel samples. Fig. 3. Histograms of the oven-dry moisture content of the control fuel samples. Fig. 4. Mass of water evolution in time for the fuel samples dried with the moisture analyzer. The fitted model is eq. 2 with parameters given in Table 2. Fig. 5. Oven-dry moisture content as function of the value measured by the moisture analyzer. The linear fit is given by eq. 4 with correction factor F = 1.017 (Table 3). Fig. 6. Correction factor for the analyzer fuel moisture as a function of drying time. Statistics for the regressions from which F was derived are given in Table 3.
Figures Figure 1
13
Figure 2
Figure 3
14
Figure 4
Figure 5
Figure 6
15