Forecast Methods Their Strengths and Weaknesses 2431 Climatology

Until the recent advent of operational oceanography, navies have had to rely on climatologies or point observations to make operational decisions (Jacobs et al. 2009). Climatologies can be useful for planning purposes, but they are of limited use where oceanic variability is high. In the extreme case of a bimodal system, climatology shows the mean of the two modes, which may be a physical situation that never arises in reality (e.g. south or north of a front, inside or outside an eddy). Figure 24.6 illustrates the limitations of climatology, by showing the September monthly mean SST in the Tasman Sea as depicted by the World Ocean Atlas 2001, and the daily mean SST on 16 September 2009 from the BLUElink forecasting system.

Conversely, where variability is low, or where it occurs on timescales longer than the averaging period (normally monthly), climatology can give a very good indication of expected conditions. Furthermore, the expected error of a forecast based on climatology is independent of the lead time of the forecast (see Martin (2010, Fig. 8b)). Forecasts based on deterministic models or persistence perform better, on average, than climatology in the early part of the forecast period. This is illustrated in Fig. 8b of Martin (2010), which shows the median RMS errors of global Sea Surface Height (SSH) forecasts based on climatology, persistence and a deterministic model. Climatology gives better guidance than other forecast methods, such as persistence or deterministic models, at long lead times (Murphy 1992), because forecasts based on persistence or unbiased deterministic models asymptote to twice the climatological variance at long time periods. This is because, once these forecasts are completely decorrelated from reality, they have errors resulting from having anomalies in the wrong places, as well as errors from not having anomalies in the right places (Kalnay 2003). For this reason, defence forces normally use cli-matological oceanographic data when conducting long-range planning.

Fig. 24.5 36 h Forecast of risk of pirate attacks off the Somali coast, based on significant wave height, valid 28 August 2009

Khartoum sudXn


| Less than x Piracy Probable rr. x to y Piracy Possible

More than y Piracy Unlikely

0 0

Post a comment