2 Observational data and their processing

The OJ-94 Project has collected a large data base on OJ 287 (Pursimo & Lehto 1996; Sillanpää et al. 1996a,b; Kidger et al. 1998) with good time coverage, especially in the optical region. This is by far the most complete light curve ever obtained for an extragalactic object, with over 3000 data points in the V-band and slightly less in the other optical bands (Table 1). All the collected data is archived in Tuorla Observatory (Pursimo & Lehto 1996) and can be obtained from the Project WEB site: http://astro.utu.fi/oj94/. In UV and IR the data are somewhat sparse but the IR light curves are more detailed than what has previously been observed.

UV-observations were carried out with IUE in Nov.-Dec. 1994, during the optical flare, in two spectral bands (for details see Pian et al. 1996). Seven flux measurements are available both at $\lambda$2550 and at $\lambda$1550.

The majority of the IR data were obtained during 1993/1994 observational season with two 1.5-m telescopes; TIRGO at Gornergrad, Switzerland and TCS on Teide Observatory, Tenerife, Spain ( Kidger & Gonzalez-Perez 1994). The light curves are quite well covered, especially from early December 1993 till April 1994. A few additional observations were obtained in autumn 1994. The data base contains 75 points in the K band, 73 in H and 47 in J.

The optical observations discussed here cover the time range from September 1993 till March 1996. Table 1 contains the basic information about optical observations: photometric band, the amount of individual brightness measurements and the number of telescopes (or observatories) participating in observations (see also Sillanpää et al. 1996a,b; Kidger et al. 1998). The majority of V,R,I data have been obtained by Boltwood.

  

Table 1: Basic information about optical observations

An examination of the whole data base shows that there are some observations with large errors. In the archive we have saved ALL the data sent by the different observers without special quality check. Bearing in mind that it is always better to reject shady data than to use them in the analysis we have decided to remove such measurements from the data base. To find an acceptable level of observational errors we used the following method. In each spectral band the magnitudes were transformed to fluxes using absolute calibration by Mead et al. (1990). Then the variability ranges were divided in segments of 1 mJy length. The mean flux value $\bar {F_j}$, corresponding mean observational error $\bar {\sigma_j}$ and its dispersion $\Delta_j^2=\frac{\sum_{k=1}^{n_j}(\sigma_k-\bar {\sigma_j})^2}{n_j -1}$ were calculated for each segment j which contains n_j points. An observation was considered as unreliable when its relative error $\sigma_k/F_k$ exceeds $(\bar {\sigma_j}+3\Delta_j)/\bar{F_j}$.

Figure 1 demonstrates the method of data reduction for the V band. Crosses give relative errors for all individual observations. Solid line is the exponential approximation of the dependence on F of the values $(\bar {\sigma_j}+3\Delta_j)/\bar{F_j}$ (triangles in Fig. 1). All points lying above this line were rejected from further consideration. The number of data points removed at this stage is indicated in the second column of Table 1 in brackets. Note that the number of rejected data points is less than 3% of the data set.

  

Figure 1: V band; the dependence of individual observational relative errors on flux (see text)

The next step is to check the data homogeneity. Figure 2 shows a fragment of the light curve in the V band for 1993/94 and 1994/95 observational seasons. The data of various authors are plotted by different symbols. The examination of the figure (and the whole data base also) reveals that the V band light curves of various observers agree quite well with each other (see also Sillanpää et al. 1996a,[b]Si96b).

To examine the agreement in other bands the dependences of colour indices B-V, V-R and V-I on V magnitude were constructed. As an example the dependence of V-R versus V is shown in Fig. 3. One can see that there are some points very much scattered relative to a general dependence. They are labeled by numbers being the last four figures of JD. These points as well as analogous points on the other graphs were also rejected from data base as unreliable.

  

Figure 2: The V band light curve of OJ 287 for seasons 1993/94 and 1994/95. The different symbols correspond to different observers: $\Box$ -Boltwood, $\bigtriangleup$ - RoboScope, $\Diamond$ - Italian, $\circ$ - Crimean Observatory, + - IAC-80 telescope, $\times$ - Nordic Optical Telescope, * - Jakobus Kapteyn Telescope

  

Figure 3: The dependence of the colour index V-R on the V magnitude (the symbols are the same as in Fig. 2)

At the same time all three graphs show that systematical shifts between the colour indices found by various observers are smaller than $\rm 0.1^m$ (if any). Therefore no correction for this effect was made.

After rejecting all shady data the mean values of fluxes for each JD were calculated taking into account the errors of individual observations. Note however, that here we have ignored the microvariability observed during some nights (e.g. Dultzin-Hacyan et al. 1997). Values for F and their mean square errors were found for 36 dates in U band, 175 in B, 458 in V, 406 in R, 321 in I, 28 in J, 36 in H and 37 in K band.

All the data (including those for UV bands) were corrected for galactic absorption using the standard curve of interstellar reddening and $A_V =\rm 0.06^m$ according to Smith et al. (1987). This data base was used for further analysis.