Calibrated models are used to correct the Hemispherical-Conical Reflectance Factors (HCRF) measured in a continuous mode in the field, and the influence of each model is discussed. Results suggest that effects of thermal sensitivity and non-linearity partially cancel out when reflectance is computed using channel A (irradiance) and B (radiance) ratio, however, in the case of non-linearities, this may not occur when signals are very different in each channel. Dark measurements showed a bias inversely dependent on temperature that was added to dark current. Wavelength calibration showed a dependency on temperature; however, this was small considering the spectral features of the instrument (Full Width at Half Maximum~10nm, interval sampling ~3.3nm). Finally, the cosine directional response correction model produced the largest differences between he corrected and the non-corrected HCRF. This correction requires accounting for the diffuse-to-global radiation ratios. Differences between corrected and non-corrected reflectances were larger in the near infrared region than in the visible. We conclude that characterization of spectroradiometers installed outdoors in automated continuous systems is necessary to ensure comparability and quality of data. Thermal insulation of the instruments could reduce errors related with dark current, thermal sensitivity and wavelength calibration; however these still should have to be known in order to compare with data from other instruments. Moreover, non-linearities and directional response of the cosine receptors would have to still be corrected in order to achieve reliable measurements under different ranges of irradiance and sun elevation.

Hyperspectral sensors are increasingly being used to continuously collect optical data that can be related to carbon and water ecosystem exchanges. Automated proximal sensing can solve the temporal mismatch existing between the periodic remote observations and the continuous acquisition of the Eddy Covariance systems, and can provide also tools for the up-scaling. However, characterization of spectroradiometers used continuously outdoors is necessary to assure data quality; because environmental conditions can influence the instrumentation performance, but also are drivers of the vegetation physiology estimated through the optical measurements. We describe the laboratory characterization previous to field deployment of a Unispec-DC, a dual channel spectroradiometer integrated in an automated multi-angular system (AMSPEC-MED). The instrument operates in a savanna ecosystem in Majadas del Tiétar, Cáceres, Spain, under a wide range of temperatures, radiation levels, illumination angles and internal settings.

Laboratory experiments were conducted in order to characterize several features of the spectroradiometer and to estimate correction models. Dark current, the signal produced by thermally generated electrons, was modelled as a function of the temperature and the integration time set. Thermal sensitivity, the sensor’s responsivity dependence on temperature, and spectral calibration were also modeled as a function of temperature. Moreover, non-linearities, deviations from a linear relationship between the input radiance and the output signal, were characterized as a function of the grey level and (innovatively) of the integration time. Since a cosine receptor is used to sample irradiance; we also modeled the diffuser directional response deviations from the ideal response, the cosine of the incident angle of illumination.

Peer reviewed