No, we do it for Ecan. We like to think we know a bit about local weather conditions

We have been doing this forecast for about ten years now. The relationship between smog level and weather parameters works well at a general level, but not at all well in detail. Thats why its a general forecast. We and Ecan did a lot of analysis a few years ago and found that the best indicator of whether there would be an exceedance of the guideline (which is 50micrograms of PM10 per cubic metre of air over a 24 hour period (9am to 9am) ) is at least four consecutive hours of inversion and light winds (less than 2m/s) during the evening between 4pm and midnight. We issue the forecast around 8am each morning, and have the option to update it during the day if conditions (or our opinions of them!) change.
So the forecast (over three categories, low risk, medium risk, high risk) is actually a forecast of the likelihood of getting the four hours in the time period specified. Even so, the relationship is not perfect - a 50mg average over 24 hours can be obtained with fewer than 4 hours during the evening, if, for example, levels are very high the following morning. Conversely, a few nights with more than four hours can end up under 50mg if, for example, a NW starts up mid evening. So the relationship between 4 hours and the 50mg is not perfect, but it does seem to produce the best results overall. For the last few years we have averaged around 75-85% accuracy. Thats driven largely by the number of low nights, which are usually fairly easy to predict. High nights are harder, and we seem to average around 65% for those - ie 2 out of every three high nights we predict as high, but about 1 in three sneak up on us, usually on nights when we are unsure and have issued a medium forecast.
As for tonight, experience shows that a dying SW wind drops out rather slowly if the flow is more S than SW as has been the case today, plus the inversion doesnt form as quickly or strongly when there is moisture around. The showers have cleared late enough and the flow stayed strong enough through the afternoon to make me think that the inversion wont become well established until later in the night. Thats the way I saw it at 8am, and I had no reason to want to change it at 4pm. Yes, I too expect a frost in the morning but I dont expect four consecutive hours of inversion and no wind before midnight, hence the low risk forecast. I did go for medium risk for Timaru, on the basis they would be nearer the ridge and have the wind drop out earlier, and have had a few extra hours for the ground to dry out prior to sunet.
So you see, there is quite a bit of thought that goes into it, and its a frustrating forecast to make because of the imperfection in the relationship so that sometimes we call the weather right, but the smog outcome doesnt quite work.