EXPLANATION OF FORECAST AND FORECASTED PARAMETERS

Send comments to Marc Sarazin (msarazin@eso.org) or Andre Erasmus (erasmus@saao.ac.za).

1. Forecast Times

The first forecast (based on the 12UT (previous day) run of the European Centre for Medium-Range Weather Forecasting (ECMWF) meteorological numerical model and the 23h45 UT satellite image) is made at approximately 03h00 UT each day and includes the 2–8 day long-range outlook. The forecast and outlook are updated at about 06h00 UT after the 02h45 UT satellite image is received. After receipt of the output from the 00UT run of the ECMWF model, additional forecast updates are made as new satellite data for 06h00, 09h00, 12h00, 15h00, 18h00, and 21h00 UT are received. The arrival of new data automatically triggers an updated forecast. If no new data are received, the most recent forecast is displayed. If input data have not been received for 24 hours or longer a message to that effect is displayed in the time and date field.

2. Forecast Periods

The forecasts of the humidity and cloud cover parameters described below are for 3-hour periods and are indicative of conditions for the entire period. The first line of data is an analysis of conditions at the observation time of the satellite image used in the forecast.

3. Forecast Products

There are two main categories of products — image products and text products. For each 3-hour forecast period, the satellite image used as input for a particular forecast run is displayed with the values of the forecasted parameters at the bottom of the image. It is important to note that the forecasted values are indicative of conditions for the entire period and not for a specific time. In addition, the observatory site locations are shown and lines are drawn to indicate the image pixels most likely to move over the sites. Text products include a tabulation of forecast parameter values for between 8 and 11 (depending on when the forecast is made) 3-hour periods extending between 18 and 30 hours into the future and the long-range outlook for two to eight days ahead.

4. Humidity Parameters (AvgUTH(%), SigUTH(%) and PWV(mm))

The Upper Tropospheric Humidity (UTH) is the relative humidity for the middle to upper troposphere in the layer between 700 mb and 300 mb (approx. 3000 m – 9500 m). There are no significant amounts of water vapour in the atmosphere above this layer. However, there may occasionally be intrusions of low-level moisture near the ground that would not be included in the UTH.

AvgUTH is the average (mean) UTH for the forecast period using all the pixels in the corresponding satellite image area. Opaque pixels are assigned a UTH of 100%. For clear and transparent pixels the computed UTH for each pixel is used.

PWV is the Precipitable Water Vapour in the atmospheric column above the observatory and is equivalent to the amount of liquid precipitation that would result if all the water vapour in the column condensed (in mm). PWV values are based on the UTH for the middle and upper troposphere and a surface relative humidity value derived from the observations of surface relative humidity at the observatory sites in the 24 hours preceding the forecast. The UTH value used in the computation of PWV is for clear and transparent pixels only. The PWV is therefore indicative of moisture conditions where or when observations are possible. If no surface relative humidity measurement is available, PWV is not computed and “n/a” appears in the column.

SigUTH is the standard deviation of the UTH for the forecast and indicates moisture variability. Usually when conditions are dry, SigUTH is small. A rise in SigUTH values would indicate a transition to more moist conditions. A minimum number of 25 clear and transparent pixels are needed for SigUTH to be output, otherwise “n/a” appears in this column.

When conditions are too cloudy for humidity parameters to be computed “No clear pixels” appears in the humidity listing.

5. Cloud Cover Parameters (%Total, %Opaque, %Transparent, Transparency Index, CldAdj)

For each forecast period the percentage of Opaque, Transparent and Total (sum of opaque and transparent) cloud cover forecasted for that period is indicated. The cloud cover parameter values are based on the detection of cloud in both the Water Vapour (6.7µm) image and the Infra-red window (10.7µm) image. The usually observed transition from clear to cloudy conditions occurs with the passage of a cloud band. As the cloud band approaches, there is a gradual increase in the amount of transparent cirrus, then the relative amount of opaque cloud increases while the transparent cirrus decreases. The amount of opaque cloud will reach a peak and then start to decrease, gradually giving way to transparent cirrus and finally clear skies. This cycle can take from a few hours to a day or more, depending on the extent, thickness and movement of the cloud band.

Cloud cover percentages should be interpreted as the fraction of the sky expected to be covered by that cloud type, on average, during the forecast period. For example, a 50% opaque cloud forecast may mean that 50% of the sky will be clear and 50% overcast or it could mean that for 50% of the forecast period it will be clear and for the remainder of the period it will be overcast or any combination of these extremes. The user can gain insight on the nature of the forecasted cloud cover by zooming in on the forecast period images and noticing how the cloudy pixels are distributed.

The Transparency Index (see below) is indicative of atmospheric transparency when the sky is generally free of opaque clouds. For this reason, when more than 10% opaque cloud is forecast, the Transparency Index (see below) is not computed and “n/a” appears in this column. In the absence of opaque cloud but when transparent cirrus is forecasted, conditions may be suitable for spectroscopy but marginal for photometry. Under these conditions, the Transparency Index may be used as a guide.

The Transparency Index is defined as follows:
TI = 1.0: All pixels are clear (photometric)
TI = n/a: 10% ≤ % of the pixels that are opaque < 100% (marginally spectroscopic)
TI computed (0 < TI ≤ 1): 0% < % of pixels that are opaque < 10% (marginally photometric). Under these conditions, transparent cirrus may be present. The value of TI ranges from near zero (less transparent) to 1 (more transparent).

The column CldAdj contains a symbol indicating upward or downward adjustment in the cloud cover forecasted for a given forecast period. Consideration is given to in-place (orographic) cloud formation at La Silla and the dissipation of cloud moving from the east across the Andes at Paranal. A zero “0” indicates that no adjustment in the cloud cover forecasts is expected. An expected increase (decrease) is indicated by “+” (“−”) or by “+ +” (“− −”) depending on the magnitude or likelihood of the expected adjustment.

6. The 2–8 Day Long-Range Outlook

The long-range outlook is an objective assessment of moisture conditions over the 2–8 days following the day of the forecast. It is provided to enable observers and schedulers to estimate how long a dry or moist period may last. The day for which the 3-hourly forecasts are issued, is classified as “dry”, “drying”, “moistening” or “moist” and then, using information on the development of pressure systems, the persistence of winds from the tropics and westerly wave propagation over the area of interest, days 2–8 are classified into one of the same categories.

Two important factors to note regarding the long-range outlook are:

  1. the outlook is for moisture conditions and not cloudiness, so cloud may be present under dry conditions and moist conditions may be cloud free (It is reasonable to conclude that clouds are more likely to be present when it is also moist).
  2. the outlook is not a forecast and is therefore not to be used as a predictor of day-to-day conditions but rather as an indicator of when moisture conditions are expected to change. Frequent changes from “drying” to “moistening” over the week suggests low predictability for that week. A gradual change from “dry” (“moist”) to “moist” (“dry”) over several days indicates that predictability is higher. Based on the results from the feasibility study, the algorithm is expected to estimate when a “moist” (“dry”) period will end with an accuracy of ±1 day 89% (74%) of the time at Paranal (La Silla).