...
t'=(t-t_last)*(FWHM/0.8)*sqrt(median_sky/sky).
This formulation allows us to give "partial credit" for observations through clouds, or in poor seeing.
Field and Passband Prioritization
Not all locations on the sky are of equal interest to the survey, and this priority can differ across passbands. So we need to allocate a Field Priority Factor FPF(field, band) to tune the relative priority of observations.
Slewing Overhead Matrix
For N fields there is an N x N symmetrical matrix that lists the time overhead required to move the telescope from field i to field j. These times depend only on the angular separation between two fields, and the matrix only needs to be computed once.
Filter Exchange Overhead Matrix
For F filters there is a time penalty associated with changing filters. This can be represented by a symmetrical F x F matrix, but it's probably fine to just use a scalar t_filter.
...
Oct 18 2013, C. Stubbs
Observations obtained at angles z from zenith suffer from two effects:
...
At 35 seconds per visit and 9.6 square degrees per field, we cover the sky at a rate of 7900 square degrees in an 8 hour night. That means that on average we revisit interval (no weather) is 3 days for 18,000 square degrees. In the longest night in the year, 10 hours, we'd get about 1000 images.
Sky rotation.
The position of objects on the sky changes in right ascension direction at an angular rate of 15 degrees per hour times cos(declination).
One potential approach:
- determine the rank-ordered priority of all fields on the meridian, in each passband, for different potential values of seeing.
- reject the fields that never appear in the top 1000. These have such low priority we'd never get to them.
- For each parametric value of seeing, compute the sequence of observations that maximizes the merit function, including the slew overhead contribution.
Some references
LSST science book
...