Sometimes, a VDS time series consists of an annual index in some years and individual measurements in others. The procedure followed then depends on the number of years of individual measurements in the ‘latter part’ of the time series. Suppose that data are reported in years \(t_i, i = 1,...,N\) and let \(t_j\) be the last year in which an annual index is reported; i.e. so individual measurements are reported in years \(t_i, i = j+1,...,N\) (and possibly earlier). The assessment methodology then depends on the value of \(N^{*} = N - j\).
If \(N^* \ge 3\), the data from the years \(t_i, i = i,...,j\) are discarded and the individual measurements from the years \(t_i, i = j+1,...,N\) are assessed using the standard proportional odds model. Information about trend (in the early part of the time series) is sacrificed to get a better assessment of status.
If \(N^* \lt 3\), the individual measurements are converted to annual indices and the whole time series is assessed accordingly. However, there is one embellishment if individual measurements are submitted in the last monitoring year \(t_N\). In many time series, VDS levels have declined so rapidly that the linear models used to assess trends in annual indices cannot track the change completely. The linear models correctly show evidence of a decline, but over-estimate VDS levels in the final monitoring year suggesting that environmental status is worse than it actually is. To overcome this, the assessment of status is based on the individual measurements in the final monitoring year. Specifically, an upper one-sided 95% confidence limit on the mean VDS class in the final monitoring year is obtained from the model used to estimate the cut-points in the standard proportional odds model. This confidence limit is then compared to the assessment criteria.