Salt marsh vegetation radiometry Data analysis and scaling Sonia Silvestri a, *, Marco Marani a , Jeff Settle b , Fabio Benvenuto c , Alessandro Marani c,d a Dipartimento di Ingegneria Idraulica, Marittima, Ambientale e Geotecnica, Universita ` di Padova, via Loredan 20, I-35131 Padua, Italy b ESSC, University of Reading, Reading, UK c Dipartimento di Scienze Ambientali, Universita ` Ca’ Foscari di Venezia, Venice, Italy d Istituto Veneto di Scienze Lettere ed Arti, Venice, Italy Received 18 July 2001; received in revised form 29 September 2001; accepted 9 October 2001 Abstract This paper aims to determine the optimal procedure for classifying salt marsh vegetation from hyperspectral data and to establish relationships between airborne and ground measurement properties. The study is carried out on data collected in the Lagoon of Venice (Italy). Spectral angle mapping proves to be a reliable classification procedure, and spectral differentiation is seen to improve separability of vegetation types. Further, scaling relationships are derived to link the value of the variance of data aggregated at different scales, allowing the determination of data variability at coarse resolution on the basis of ground measurements. The comparison between the theoretical, up- scaled, values of standard deviation and those computed from remote sensing data shows a good agreement supporting the derived relations. Finally, a scaling relationship is established for spectral angles, which may be useful in determining an optimal threshold angle from ground data. D 2002 Elsevier Science Inc. All rights reserved. 1. Introduction A large number of techniques exist for characterising surface cover using remotely sensed data from sensors on satellite or aircraft platforms. Thanks to these methods, remotely sensed data in the form of aerial photographs, satellite multispectral images, and airborne digital multi- spectral camera images have been successfully used to characterise coastal areas and salt marshes (Dale, Hulsman, & Chandica, 1986; Donoghue & Shennan, 1987; Phinn, Stow, & Van Mouwerik, 1999); good improvements have been shown in this field by recent applications of classifica- tion techniques to hyperspectral airborne data (Bajjouk, Guillaumont, Populus, 1996; Eastwood, Yates, Thomson, & Fuller, 1997; Smith, Spencer, Murray, & French, 1998). In order to be effective, classification techniques require ground reference data, ideally acquired simultaneously. Such ancillary ground data are used to identify training areas for classification procedures, to measure ground leaving radiance to validate an atmospheric correction, or to gather ‘‘end-member’’ spectra for use in spectral unmix- ing or matched filtering. Despite the extensive literature on classification and related techniques, and the costs of ground data collection, there is still a lack of clearly defined protocols for the collection of such data. It is important to stress that reliable ground measurements are essential for the calibration of remotely sensed data and for the validation of the quantitative information extracted from it. The selection by direct inspection of reference areas for the calibration and validation of a classification procedure is, in fact, not a reliable and repeatable procedure, and it is operator depend- ent. The human operator tends, for example, to add informa- tion when this is incomplete (e.g., a patch of a given vegetation type may be perceived as uniform even when the vegetation does not cover the soil entirely) or to overlook information, which is not perceived as necessary (e.g., the variability of soil moisture, or water depth over the soil in the case of tidal area) but which may substantially influence the signal measured by a sensor and, conse- quently, the classification procedure. Hence, the use of 0034-4257/01/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved. PII:S0034-4257(01)00325-X * Corresponding author. E-mail address: sonia@unive.it (S. Silvestri). www.elsevier.com/locate/rse Remote Sensing of Environment 80 (2002) 473 – 482