1 Introduction

Integral-field spectroscopy (IFS) provides a spectrum simultaneously for each spatial sample of a two-dimensional field. This technique has merited much attention in recent years due to its advantages with respect to classical sequential 2-D spectroscopic techniques (e.g., long-slit scans, Fabry-Perots) when studying relatively small extended objects. Several approaches to IFS have been developed, based in the use of fibres (e.g. Vanderriest & Lemonier 1988; Barden & Wade 1988; Arribas et al. 1991), microlenses (e.g. Bacon et al. 1995), micro-mirrors (e.g. Weitzel et al. 1996; Content 1998), or mixed solutions (e.g. Allington-Smith et al. 1997).

Most of the advantages of the IFS technique are direct consequences of the simultaneity when recording spatial and spectral information. The simultaneity not only implies a more efficient way of observing but, more importantly, it guarantees a great homogeneity in the data. In addition, IFS has other advantages. For instance, with IFS systems there is no need for an accurate centring of the object in the slit or to adapt the slit width (spectral resolution) to the seeing conditions; neither are the data affected by "slit effects'' when determining radial velocities (cf. Vanderriest 1995).

Here we want to demonstrate another relevant advantage of IFS. In many circunstances it is possible to determine and to correct in the spectra the effects of differential atmospheric refraction, using an a posteriori procedure. This is obviously important to preserve the spectrophotometric properties of the spectra without the use of atmospheric dispersion correctors (ADCs). Note that for long-slit observations the presence of differential atmospheric refraction imposes strong restrictions which cannot be corrected by any means. Thus, the slit must be orientated along a direction defined by the parallactic angle (i.e. the spatial direction is predefined) if it is required to preserve the relative fluxes in the spectrum. If the slit is orientated at a desired position angle (no coincident with the one along which the differential atmosperic refraction takes place) special filters for the acquisition/guiding systems should be used to optimize the detection in a particular wavelength range, other spectral ranges being affected by light losses. To have a complete feel for these difficulties, we should note that the effects of differential atmospheric refraction are active, in the sense that they change with time. All these problems are strong drawbacks to obtain 2-D spectroscopic maps by scanning with a long slit. However, an approach using the a priori (theoretical) estimates of the shifts induced by differential atmospheric refraction to correct long slit data was proposed by Walsh & Roy (1990).

In this paper we study the differential atmospheric refraction effects in IFS, and the basic concepts on which its correction should be performed. These steps will be illustrated with an example drawn from observational data.