121 
3.1. Errors in field spectroscopy 
Error is defined as the difference between the measured value and the "true" value of the entity. Error can result 
from random or systematic sources. The effect of random error can often be reduced by repeated measurement, 
or by averaging, which increases the measurement precision. Systematic errors cannot be reduced by averaging 
or making repeated measurements. 
The accuracy of field spectral measurements can be defined as the difference between the 
mean indicated value and the true value. It therefore refers to the measuremnet in absolute terms and implies 
the use of absolute or traceable standards. The precision of the measurement defines the reproducability or 
repeatability of the measurement made with the same sensor. 
Precision and accuracy cannot be determined exactly, but an estimate of the the measurement 
uncertainty can be derived from consideration of the observed data, and an understanding of the sources of 
error for a particular measurement situation. In field spectroscopy the main sources of measurement uncertainty 
are identified as: sensor calibration and characterisation factors, environmental factors, and measurement 
methodology. These are summarised in Table 1. 
1. Sensor and panel characteristics and calibration 
Radiometric linearity 
Radiometric precision 
Wavelength precision and accuracy 
Dark current characteristics 
Temperature sensitivity 
Field-of-view characteristics 
Radiometric calibration 
Reference panel characteristics 
2. Environmental factors 
3. Measurement methodology 
Solar angle 
Dual beam vs. single beam 
Atmospheric conditions 
Sequential vs. simultaneous measurement 
Adjacent object scattering 
Cosine corrected sensor vs. panel 
Surface contaminants 
Hand held vs. fixed support 
Table 1. Some of the factors which affect the accuracy of field spectroscopy 
3.2. Example: temperature sensitivity of Spectron SE590 spectroradiometer 
The Spectron Engineering SE590 spectroradiometer has a specified working range of 0° to 50°C and the two 
SE590s in the NERC-EPFS have been used across this range of temperatures, in environments from Antarctica 
to the Negev desert. The radiometric sensitivity of the CE390 sensor head (s/n 1569) was measured by making 
a series of 10 replicate scans of the EPFS radiance calibration source integrating sphere. The procedure was 
performed with the sensor head cooled to around 2°C, and then repeated with it heated to around 38°C and 
again at room temperature (23°C). In each case the CE390 was left to stabilise for at at least one hour at the 
required temperature before measurements were taken. The 10 replicate spectra for each temperature were 
averaged and compared. The ratio of both the high and low temperature measurements to that for the room 
temperature measurement was also calculated. 
Figure 3 shows that the sensor response has a strong temperature dependence, especially at wavelengths 
above 700nm. The SE590 has an internal correction for dark current errors so the change in responsivity 
observed here is not simply a change in dark level signal. The ratio of the high and low temperature mean 
spectra to that at 23°C confirm the spectral variation in the temperature dependence. The ratio indicated a 10 % 
increase in sensitivity at 800nm rising to 35% increase at lOOOnm for the high temperature case. The reduction 
in sensitivity for the low temperature measurement is of a similar magnitude.