Processing of Visibility Observations 25 

It is preferable to use βe as the verifying parameter since it is the measured quantity. When helpful for interpretation or comparison with other techniques, results will be converted to visibility using the appropriate ASOS algorithm from Table 2. While the uncertainty existing in the conversion of βe to visibility is perhaps a significant source of error, it will not be the focus of this research. In addition to the several imperfect assumptions detailed above, producing visibility observations in practice is also subject to error from differences in the shape or color of the objects or lights being viewed, the viewing angle with respect to the horizon, and the position of the sun. Some of the assumptions made to mitigate these are necessitated by the use of automated instrumentation, and some are required even with a human observer simply due to the nature of the measurement.

Raw, one-minute βe observational data for the seven verification sites was obtained from the National Climatic Data Center website (2011). In order condense this data into a single hourly βe observation suitable for verification, the 10 βe values during and prior to the top of each hour (spanning 10 min) were averaged. Other measured parameters, such as temperature, dewpoint temperature, wind direction, and current weather condition were taken directly from the official METAR observation.

The basic process used by ASOS to determine the current weather condition plays a critical role in preparing the data and is summarized in Figure 6. As with all ASOS measurements, the process is completely automated except during equipment malfunction or other extenuating circumstances (e.g., smoke in vicinity, presence of a funnel cloud,

etc.). In the overwhelming majority of cases, any reduction in reported visibility to below 7 mi as measured by the forward scattering sensor can be ascribed to precipitation (of some form), mist, fog, or haze. Precipitation is detected by the ASOS precipitation gauge and reported accordingly, regardless of the visibility. Independently, if the reported visibility is <7 mi and the dewpoint depression is <2.2 K, mist or fog is reported. The distinction between mist and fog is one of severity; fog is used if reported the visibility is <0.625 mi, while mist is used otherwise (hereafter, both will be called fog for simplicity). Note that fog and precipitation can be reported together if both conditions are met. Lastly, if the reported visibility is <7 mi but the dewpoint depression is >2.2 K, haze is reported, unless precipitation is also reported, in which case the precipitation takes precedence (National Oceanic and Atmospheric Administration/DOD/FAA/U. S. Navy 1998).

Figure 6. Summary of basic logic used by ASOS to determine present weather. Only the aspects of the logic relevant to this research are shown.

This logic makes the following the assumptions that must be deemed acceptable before using the observations as ground truth:

 Fog and haze cannot coexist

 If the reported visibility is <7 mi, the dewpoint depression is <2.2 K, and it is not precipitating, then fog must be present

Determining the presence of fog based only on the visibility and dewpoint depression may seem a crude approximation but it is consistent with a lack of distinction between fog, haze, and mist. Automated instrumentation aside, the distinction between haze and fog is quite inexact. Haze is defined as aerosol particles that “increase in size with relative humidity”, but not so large that they reach their activation radii, at which point they would become mist droplets (American Meteorological Society 2012). The exact RH at which this occurs depends on the aerosol characteristics (Rogers and Yau 1989), and cannot possibly be known in every case. If the RH remains high enough, the droplets will continue to grow and eventually be classified as fog droplets. The ASOS dewpoint depression threshold of 2.2 K (which corresponds to an RH of 80–90% in most cases) is likely to be below the activation threshold of most haze particles (Rogers and Yau 1989). Referring to haze, mist, and fog, the American Meteorological Society Glossary (2012) states “there is no distinct line…between any of these categories”. Given the indistinct transition between haze and fog from an observational standpoint, the ASOS logic seems reasonable. At worst, some instances of moist haze whose particles have not yet reached activation radii but are causing a visibility restriction will be misclassified as fog.

Once the hourly reports of temperature, dewpoint temperature, wind direction, and present weather have been combined with an hourly βe value, additional processing is needed to isolate just the contribution of fog to the measured βe. First, any observation with βe <0.29 km-1 (approximately corresponding to daytime visibility of 6.5 mi and

nighttime visibility of 8 mi), is simply classified as a no-fog case. In these cases, the actual value of βe is not retained because 1) except for precipitation, ASOS does not report the phenomenon responsible for any reduction in visibility, and 2) it is outside the range of visibilities relevant for most DoD operations.

Next, since haze and fog cannot coexist, any observation reporting haze is also classified as a no-fog case, even if βe>0.29 km-1. In these cases, βe is reassigned a value of 0.10 km-1, an arbitrary figure that simply ensures these observations are not confused with cases of fog.

Finally, observations with βe>0.29 km-1 and precipitation occurring were removed from the dataset al.together, even if fog was also reported. For a given βe, the relative contributions of fog and precipitation are inseparable in this case.

After the filtering described above, the remaining observations are those with

βe>0.29 km-1 due to fog alone, thus comprising the fog cases of the verification dataset. In these cases, the βe value was preserved.

A small percentage of the verification data did not fit into one of the above categories and required special treatment. If a nighttime observation reported a βe value in the range 0.29-0.37 km-1 with no precipitation, no present weather was normally reported since this βe range corresponds to reported visibilities >7 mi using the nighttime algorithm and subsequent rounding. In these cases, the present weather was deduced to be either haze or fog using the same dewpoint depression criteria used by ASOS.

Figure 7. Summary of the processing of the hourly observations to isolate the effects of fog on the observed βe values.