Every morning I create a 6-day forecast for Connecticut and I keep track of what the various computer models and forecast techniques are predicting.
In April, I had 29 days out of a possible 30 to go back and check verification.
The above graph shows that forecast error generally increases with time. It is interesting to note a spike at Day 5 and a decrease at Day 6. That goes back to two particular days that had poor Day 5 forecasts vs. actual temperatures. The spread is relatively uniform as well. There’s really no case of one model or forecast having significantly better skill at one period, but not another. The error with the NAM model does seem to increase faster with time than the others.
My own forecasts happen to rank more accurate than any of the other models. There are several factors for this. One could be that my forecast is for an average of all inland Connecticut ASOS reporting stations, while for the models I used the location of Meriden (KMMK). Starting tomorrow, I will also create 6-day forecasts for Meriden specifically, to see how the results end up for May. Another factor could be that my Day 1 forecast comes out a few hours after the 00z Euro and 06z NAM/GFS/DGEX.
The Euro and MAV MOS rank fairly close, but it is very interesting to note that the negative (cold) bias the Euro has is almost a mirror reflection of the MAV MOS positive (warm) bias.
The MAV MOS appears to correct some of its bias towards Days 5 and 6. That can perhaps be partially explained by the fact that MOS is skewed towards climatological temperatures.
For my own forecasts, the bias ends up becoming close to zero.
Explaining the models/forecasts…
Q: My forecast high temperatures for inland Connecticut. (mean of inland stations)
MAV MOS: Forecast high temperatures for KMMK. (06z model run)
ECMWF: Forecast grid-point high temperatures for KMMK. (00z model run)
NAMDGEX: Approximate high temperatures for KMMK. These values are interpolated off of a graphical forecast, so the numbers are estimated. I use the NAM for Days 1-4 and the DGEX for Days 5 and 6. (06z model runs)
850mb: An 850mb forecast technique that I have been working on for quite some time. Because this technique is based off of Danbury (KDXR), that station is used for verification.
LAMP MOS: Forecast high temperatures for KMMK. (most recent run in morning)
NAM Re-projection: This takes into account the actual 9 a.m. temperature vs. the 06z forecast for 9 a.m. for KMMK. That error is then re-projected into the high temperature forecast. Example: If the 9 a.m. temperature was 2°F warmer than forecast, then 2°F is added to the high temperature forecast.
Consensus: A mean of each forecast above, including my previous forecast (continuity)
James Heuschkel suggested to use a blend of the ECMWF and MAV MOS, based on their opposing biases. Well, not only does the bias of the blend move close to zero, but the overall forecast error was less than any other forecast technique for Days 2-6:
It’s pretty interesting to see the results. To become a better forecaster, it really comes back to experience and being able to see how forecasts verified in the past. If a forecaster is stuck using one technique and never deviates from that, there’s no room to grow or increase forecast skill. Part of why I make 6-day forecasts every single day is so that I can become a better forecaster.