 Let's have a look at some of the tools that forecasters used to forecast precipitation type using a case from early December 2019. The WPC surface forecast map, valid at 12Z on December 1st, showed a strong, occluded low pressure system located near the Wisconsin-Illinois border. As we would expect, the bulk of the snow, shaded in blue, was predicted to fall north of the low, wrapping around to its northwest and west in the cold conveyor belt. Most of the low, WPC was predicting a variety of precipitation types north of the warm front, which fits our conceptual models, as warm air overruns cooler air near the surface in that region. WPC had painted much of central Pennsylvania into New York, as having a chance of ice at this time. The radar image from 12Z on December 1st shows that WPC generally had the right idea, with rain throughout much of western Pennsylvania, but the radar's algorithm suggested a mix was occurring across much of central Pennsylvania into New York. So how could a forecaster have diagnosed this situation ahead of time? Let's start with a look at the big picture, using this GFS 4-panel forecasting prog, valid at 12Z on December 1st. This run was from 12Z on November 30th, so it was a 24-hour forecast. The top left panel shows a deep 500 millibar low over the Midwest, basically right on top of the predicted position of the surface low, meaning that the low was predicted to be in late occlusion. But take a look at the 1,500 millibar thickness on the top right panel. The red dashed line here marks 540 decimeters, which is roughly the critical thickness at most of the elevations in the eastern United States. The line dives southward on the western flank of the low, where colder air is moving southward, but surges way north into the northern Great Lakes, and eastward to New England on the eastern flank of the low. Locations to the south of the red line here have a thickness greater than 540 decimeters, which statistically translates to odds that favor rain over frozen precipitation types. But the only thing we can conclude from critical thickness is that statistics favor rain more than frozen precipitation to the south of the critical thickness contour, which included central Pennsylvania. In reality, though, we already saw that radar indicated mixed precipitation in central Pennsylvania at 12Z. But the fact that precipitation wasn't all rain throughout Pennsylvania wasn't a surprise to the forecasters who took the time to assess the thermal structure of the lower troposphere. We can start by looking at temperature forecasts at mandatory levels in the lower troposphere, starting with 850 millibars. Here's the GFS forecast for 850 millibar heights, temperatures, and winds, valid at 12Z on December 1st from the 12Z run on the day before. The zero-degrees Celsius isotherm runs like this. And blue shadings mark temperatures below zero-degrees Celsius. Over the town of Johnstown, Pennsylvania, marked by the asterisk, clearly the GFS predicted warm air to surge northward on southerly winds, and 850 millibar temperatures were several degrees above zero, which would cause falling snowflakes to melt on the way to the ground. Now let's look a little closer to the surface at 925 millibars. Here the forecast tells a different story. Winds are out of the southeast over Johnstown, but there's some lingering low-level cold air, the GFS predicted temperatures to be slightly below zero at 12Z. That opens the door for frozen precipitation, but to really get the full picture, we have to look at forecast soundings. Here's the GFS forecast sounding for Johnstown, valid at 12Z on December 1st from the 12Z run the previous day, and I've highlighted the zero-degrees Celsius isotherm in blue. The forecast sounding shows a big, deep warm layer from about 700 millibars down to near 925 millibars, with the thin layer from 925 millibars to the ground being colder than zero degrees. Following snowflakes would melt in this warm layer, but would not have time to re-freeze in this thin cold layer near the ground. Rain drops would be super-cooled when they reached the surface, and would freeze on contact, so a forecaster could conclude that this GFS forecast favored freezing rain in Johnstown around 12Z. And indeed, that's what happened. The Mediogram shows a few reports of light freezing rain at Johnstown on the morning of December 1st, with surface temperatures hovering around 30 to 31 degrees Fahrenheit. By 18Z, the GFS predicted that the 540-deckometer thickness contour would push even farther north, clearly north of Lake Ontario and Rochester, New York along its southern shore. But could a forecaster making a prediction the day before be confident that rain would fall at 18Z and Rochester purely because of the 1,500 millibar thickness value? No, the thickness just suggested odds favoring rain over frozen precipitation, but a complete look at the thermal structure in the lower troposphere tells the story. Here's the GFS forecast sounding for Rochester valid at 18Z from the 12Z run the previous day. And again, I've highlighted the zero-degree Celsius isotherm in blue. The GFS predicted a rather deep melting layer. Temperatures were slightly above zero from near 700 millibars down to just below 800 millibars, while the layer from around 825 millibars down to the surface was below zero-degree Celsius. That thick cold layer would allow for snowflakes that had melted in this warm layer to completely re-freeze before striking the ground. Forecasters could have concluded that the GFS favored sleet in Rochester at 18Z and observations showed that's what occurred. Precipitation actually started as freezing rain earlier in the day around 15Z, but by 17Z it had changed to sleet which continued for a few hours. So critical thickness only tells us whether rain or frozen precipitation is statistically more likely. To get a better handle on precipitation type forecasts, forecasters must close the examined temperatures throughout the lower troposphere and how they change over time.