Spectroscopy Field Strategies and their Effect on Measurements of Heterogeneous and Homogeneous Earth Surfaces
Mac Arthur, Alasdair1; Alonso, Luis2; Malthus, Tim3; Moreno, Jose2
1NERC FSF, Geosciences, University of Edinburgh, UNITED KINGDOM; 22Laboratory for Earth Observation, Faculty of Physics, University of Valencia, SPAIN; 3CSRIO Land and Water, Canberra, AUSTRALIA

Field spectrometers are portable non-imaging electro-optical devices used to measure spectral radiance or spectral irradiance either, normally, through the visible to near infra-red region (400 nm to 1,000 nm) or the visible to short-wave infra-red region (400 nm to 2,500 nm) of the solar electromagnetic spectrum in approximately 200 to 1,000 or more sampling intervals. These instruments are primarily used by scientists to measure upwellling Earth surface radiance or exitance (normally radioed to a reference and, hence, termed reflectance measurements in the rest of this work), or to measure downwelling solar irradiant flux (Milton 2009). Irradiance measurements will not be considered further here. When being used to measure reflectance, unlike imaging spectroscopy, these instruments measure from a single Earth surface sampling unit - the measurement support. These reflectance measurements are normally acquired either to gain an understanding of the interaction of light with Earth surfaces and relate features of these measurements to (bio)physical or (bio)chemical processes or state variables or for the validation, or calibration, of Earth observation remote sensing (RS) measurements made by imaging spectrometers mounted on airborne and satellite platforms. For each of these three measurement purposes different Earth surface types may be selected. Surfaces considered to be either spectrally homogeneous or heterogeneous may be selected and heterogeneous surfaces may be considered as having either regular (agricultural row crops, for example) or irregular (semi-natural dwarf shrub land covers, for example) distributions of spectrally distinct reflecting elements, at some scale of observation. These surfaces may be natural, semi-natural or man-made. Field spectroscopists may then adopt either single; random; regular grid; transact; or 'swiping' (continuously moving the fore optic above a surface while the spectrometer is making a measurement integrated over the area covered) sampling strategies when making measurements of these surfaces. However, little comparative work has been done to determine the most appropriate sampling strategy for each surface types and measurement purpose neither have the inherent uncertainties (accuracy and/or precision) of each of these measurement approaches been quantified (the work by Goetz (2012) being the somewhat limited exception). Furthermore, Mac Arthur et al (2012) have demonstrated that, for two frequently used full wavelength (400 nm to 2,500 nm) field spectrometers with a selection of commonly used fore optics, the measurement support is defined by each spectrometer/fore optic system's directional response function (DRF) rather than the field-of-view (FOV) specified by spectrometer manufacturers. The manufacturers' specifications were found to be at best incomplete and possibly misleading. Mac Arthur et al (2012) also demonstrated that each reflecting element within the support was not weighted equally in the integrated measurement recorded by the spectrometer (Figure 1). There were nonuniformities of spectral response with the spectral 'weighting' per wavelength interval being positionally dependent and unique to each spectrometer/fore optic system investigated. However, Mac Arthur et al (2012) did not provide any advice on how to compensate for these systematic instrument induced measurement errors or quantify the associated uncertainties introduced by these inherent spectrometer characteristics when making measurements of Earth surfaces. The work being carried out here will provide the first systematic study of the effect of field spectroscopy sampling strategies and performance characteristics (the DRFs) for a range of different earth surface types and measurement purposes (bio-processes or state variables, RS image validation or calibration).

To conduct this study, synthetic and stylised Earth surface data cubes for each of the surface types, introduced in the previous section, are being generated by first producing raster surfaces indexed with individual surface element reflectance classes and with spatial distribution of individual reflecting components typical of each of the surface types being considered (natural; semi-natural; and man-made and spectrally homogeneous and heterogeneous). Then library spectra for each individual reflecting component will be assigned to each indexed raster position to generate 3-dimensional spectral data cubes, with 'x' and 'y' positions and reflectance per wavelength interval (Figure 2). Spectrometer DRF data cubes ('x' and 'y' positions and response per wavelength interval) for two full wavelength spectrometers, with a selection of fore optics in common use, generated by Mac Arthur et al (2012) will then be convolved with these Earth surface data cubes, following the method adopted by Mac Arthur et al (2013), to simulate spectroscopic measurements of each of the synthetic Earth surfaces data cubes generated and the results compiled for later comparison and analysis. This convolution and compiling of resulting simulated reflectance measurement spectra will be repeated by moving the DRF data cubes in spatial increments to replicate each of the different field spectroscopy sampling strategies for each Earth surface type and each measurement purpose being considered here. This will enable the different field sampling strategies to be directly compared and their suitability for each measurement purpose to be assessed.

Consequently, this work will enable recommendations on the most appropriate field sampling strategies and spectrometer/fore optic combinations, with indications of expected accuracies, to be made. This will assist robust field spectroscopy methodological development for different Earth surface types and measurement purpose and will further the use of field spectroscopy for quantitative Earth observation.

Milton, E.J., Schaepman, M., Anderson, K., Kneubuhler, M. and Fox, N. (2009). Progress in field spectroscopy. Remote Sensing of Environment, 113, S92S109. Mac Arthur, A.A., MacLellan, C. and Malthus, T.J. (2012). The fields of view and directional response functions of two field spectroradiometers. IEEE Transactions on Geoscience and Remote Sensing, 50,10. 3892-3907.
Mac Arthur, A.A., (2013) Field spectroscopy and spectral reflectance modelling of Calluna vulgaris. PhD thesis, Geosciences, University of Edinburgh.
Goetz, A.F. H. (2012) Making Accurate Field Spectral Reflectance Measurements. Available online from ASD Inc. Boulder, Colorado, 80301, USA