Characterization of biomass using instruments – Measurement of forest and crop residues - Biomass as Energy Source: Resources, Systems and Applications - page 60

Figure 3.3. Laboratory trays for use with oven.

samples (Anerud and Jirjis, 2011). Changes in drying time and thermostat settings would also
affect the result. A comparison of results over time would therefore benefit from ring analyses
where sub-samples of a certain sample are processed in different ovens at different laboratories.
The performance of the gravimetric method also has implications for the calibration and vali-
dation of the instrumental methods. Regardless of technique, they are usually calibrated against
a reference method that is likely to be the gravimetric method. Further, instrumental methods
may require additional reference values over time in order to make certain that their calibration
is stable. Deviations in performance may include changes in the hardware. Further, the biomass
mix might change and no longer be covered by the calibration model.
More crucial is that discrete shifts in the reference method would have a severe impact on the
evaluation of an instrumental method. If a discrete shift is introduced after a certain calibration
has been set, one will erroneously conclude that the result of the instrumental method suffers
from a bias. Further, erroneous reference samples are likely to be entered into the subsequent,
revised calibration models and decrease model performance because the error in reference values
is no longer normally distributed.
Another error source is caused by inappropriate storage of samples. Ideally, they ought to be
dried immediately. In practice, samples are dried in batches of, say, 30-60 trays depending on
oven size. A sample of wet fuel might dry significantly if taken in the morning, not weighed
immediately and stored in a warm room during the day before it is put in the oven. A similar
effect would occur if a truck driver, for convenience, takes a sample to be analyzed the next day
and stores it overnight in a non-sealed container in a warm room.
Sampling errors constitute the main error source. The need for representative sampling depends
on the intended use. When the sampling error and the measurement error are uncorrelated, the
total error is computed as the sum of the variances of the sampling error and the measurement
error:

s tot = s sampling + s 2 measurement

3.4.1.2 Sampling
Today, moisture measurement is mainly done for shipment control and subsequent price settle-
ment. The Swedish wood surveying organization VMF (1998) has a recommendation stating that
moisture measurement for price settlement should be unbiased and be based on a sample size that
yields a standard deviation of 1%-unit of moisture. A rising number of samples will yield a more