Project 1: Forecasting Practice Week
Orlando, FL - KMCO
Week of September 11-16, 2006
For Meteo 410, we've been asked to join a national forecasting contest in competition with other Meteo students from various participating schools around the country. We've been given some additional forecasting techniques, and a chance to hone our skill on a practice week. However, schedules and the class format didn't quite gel, but I think we got enough to be able to analyze how we did on the four days as well as reflect on where our strengths and weaknesses are.
The forecasting practice week occurred the week of September 11-16, with the forecasting parameters for each of the first four days (9/11-9/14) being as follows:
Monday, September 11 06Z 9/12/06 - 06Z 9/13/06
Tuesday, September 12 06Z 9/13/06 - 06Z 9/14/06
Wednesday, September 13 06Z 9/14/06 - 06Z 9/15/06
Thursday, September 14 06Z 9/15/06 - 06Z 9/16/06
This presented quite a challenge with collecting MOS guidance, METARS for the specific timeframe, and forecasting with progs and surface analyses. Following is a brief treatise on the climatology of Orlando, Florida, then my analysis of how close my forecast came to verifying on each of the four days.
Climatology of Orlando, Florida KMCO
For our practice run, we used the station KMCO, which is the Orlando Intl Airport. Orlando sits roughly in the middle of the state, both east-west and north-south. The elevation at the airport is 98 feet above sea level. Due to its proximity to both bodies of water (Gulf of Mexico to the west and the Atlantic Ocean to the east), Orlando is subject to sea breeze fronts in the summer, which often give rise to afternoon showers. The rainy season runs from June through September, sometimes longer, with afternoon shower activity often contributed to by the sea breeze frontal boundaries. Our challenge with this location is the transition from the rainy season to the dry season, which is trying to overlap during our forecast period.
The norms and averages for this week are pretty consistent. Persistence forecasting could be used except for the passage of frontal boundaries and low pressure systems. The challenge then comes down to predicting cloud cover, precipitation amount, and wind speed, as well as the high and low temperatures, as frontal boundaries can affect these areas considerably. Below are the norms and record extremes for the 9/12-9/16 forecasting period. I should have paid more attention to these.
- N O R M A L S -
- R E C O R D S
91 73 0.21 2.52
1962 64 1953 2.01
1967 78 1982 83
Since we needed to access certain things each day in order to form a more cohesive idea of the predicted high, low, precip, and max wind speed for the upcoming forecast period of 06Z-06Z, I formed a daily routine of sites that I accessed regularly. My daily routine included accessing MOS, HPC Surface Analyses (current, 12 hour, 24 hour, and 36 hour Fronts, as well as QPF Day 1 and 2), the eWall progs for 12Z (since I accessed in the late morning) which included NAM, WRF, GFS, UK and CMC (usually 18-42 hour loops depending on when the 06Z-06Z timeframe showed up), the local NWS Area Forecast Discussion for the city in question (KMCO, in this case), as well as Climate Report and METARS for the day before. Though I looked at the eWall prog loops each day, and referenced them in my daily briefee write-ups, I did not include them in this document.
Forecasting and Analysis
[06Z = 2:00 am EDT]
Forecast for 06Z 9/12/06 to 06Z
Actuals (Climatology) for 06Z 9/12/06 to 06Z 09/13/06
High 89 Low 76/75 Precip T (lt RN) Max Wind 18 mph/15.6 kts
Normals for September 12: High 91 Normal Low 73 Normal Precip .21 inches
Extremes for September 12: High 96 (1962) Low 64 (1953) Precip 2.01 inches (1967)
The METARS showed the high as 89, but showed the low as 76, one degree warmer than the Climate Report, as noted in the line above (METAR/Climatology). The low occurred at 06Z 9/12, and the high occurred at 18Z 9/12 and 00Z 9/13. [Note: Watch this METAR - it goes from bottom to top chronologically. The rest of the METARS will go from top to bottom chronologically.]
My high was off by one degree, and I know it's because I didn't factor in climate or afternoon heating. The low in Actuals reflects both the METAR low (06Z 9/12 - 06Z 9/13) and the Climate Report low (9/12/06). My low was a degree or so cooler as I was trying to factor in evaporative cooling from precipitation, which I also overshot. My winds were also under-predicted by 5.6 knots, meaning I didn't factor in the offshore winds enough.
The MOS for September 11th showed the 06Z 9/12 to 06Z 9/13 high and low temps as follows:
GFS Low 73 High 89 ETA Low 73 High 87 NGM Low 70 High 87 GFSX Low 73 High 89
As you can see, the NGM is running low on the low temp, and the models as a whole are running 50/50 between 87 and 89 as a high. Because of this, I predicted 88 for the high. The Actual of 89 was close, off by one degree. For the low, I chose to go one degree higher than the 3-model consensus of 73, with a low prediction of 74. I did this because the HPC map for 24 hour fronts valid 12Z 9/13/06 shows the frontal boundary in the Gulf and the warm front north of the Florida panhandle, and I was considering Orlando as being in the thermal ridge of this approaching front.
As for precip, a trace is less than 0.01 inches and I predicted .10 because I was thinking with the offshore winds, sea breeze fronts, and frontal boundary out in the Gulf, there was more of a chance of convective activity. Also, in this instance, I thought afternoon heating would play into this formula, as well. The 6 hour QPF map valid 18Z 9/12 through 00Z 9/13 shows Orlando in the light green corresponding to the 0.01-0.10 range. However, the 24 hr HPC Fronts map valid 12Z 9/12 and the 36 hr HPC Fronts map valid 00Z 9/13 both showed the thunderstorm symbol in the Orlando vicinity in conjunction with the nearby frontal boundary, and I assumed with the PoPs percentages on MOS at 19 (GSF), 31 (ETA) and 12 (NGM), that some slightly significant rain amounts were possible. Thus my prediction of .10 inches.
I forecast 10 knots for the maximum wind speed, because the 15Z Surface Analysis on 9/12 showed the winds to be fairly light (15 knots at Melbourne). I predicted less winds than Melbourne because Orlando sits further inland, and I was considering the offshore winds as being stronger on the coastline. The actual winds were actually stronger than I anticipated, due to the offshore flow. On the MOS, the winds were light at 06Z 9/12 (6 kts GFS, 9 kts ETA, 6 kts NGM), then showed an increase in the 15Z-21Z timeframe to 12-13 kts. This would correspond with the mixing eddies in afternoon heating.
for 06Z 9/13/06 to 06Z 09/14/06
Actuals (Climatology) for 06Z 9/13/06 to 06Z 09/14/06
High 88 Low 75/72 Precip 0 Max Wind 15 mph/13.0 kts
Normals for September 13: High 91 Normal Low 73 Normal Precip .21 inches
Extremes for September 13: High 95 (1977) Low 67 (1949) Precip 1.48 inches (2001)
The METARS showed the high as 88, but a low of 75, which is 3 degrees warmer than the Climate Report shows for September 13th. The low occurred at 12Z 9/13, and the high occurred at 00Z 9/14. [Note: Remember that this METAR reads top to bottom, as will the rest of them below.]
Here my high was 2 degrees too low, for the same reason as above - I didn't factor in climate or afternoon heating. The low in Actuals again represents the METAR low and the Climate Report low, so if you go by the METAR low, I was off one degree too cool, and 2 degrees too warm by the Climate Report. Cloud cover could have kept nocturnal radiational cooling from occurring as much as I thought it would. My precip forecast was completely inaccurate - I was forecasting for the frontal boundary to be near and bring showers to the vicinity, when it never actually rained in Orlando. My winds were close by 2 knots, though a bit high.
The MOS for September 12th showed the 06Z 9/13 to 06Z 9/14 high and low temps as follows:
GFS Low 73 High 89 ETA Low 73 High 87 NGM Low 70 High 87 GFSX Low 73 High 89
[NOTE: The MOS attached is from the MOS Archive, and is not the MOS I used in my forecast. The numbers above reflect the numbers at the time I accessed the MOS, and are not exactly the same as those on the attached MOS. The original MOS file was somehow lost.]
Again, the NGM came in low on the low, by 3 degrees, while the other three models were in consensus on 73. And we had another 50/50 split between a high of 87 and a high of 89. However, I predicted 3 degrees cooler than MOS for the high because the current surface analysis for 12Z 9/13 showed the warm frontal boundary just offshore in the Gulf, and the 48 hr Fronts V 00Z 9/14 showed Orlando sitting in the rain sector with thunderstorm possibilities. Therefore, I figured the high might be a bit cooler due to evaporation cooling. Also, I upped the low a degree to account for probable cloud cover with these stormy conditions, and the proximity to the low pressure system.
My precip prediction of .25 inches was based on the same thing as the high/low temps. Especially based upon the 24 hr Day 1 QPF valid 00Z 9/13 to 00Z 9/14, which clearly shows precip chances within this timeframe of roughly .5 for Orlando. Of course, these QPFs have been known to over-state these precip chances, which is why I went with .25 instead of .5. Regardless, the Climate Report for September 13 stated 0 precip, or at least less than 0.01 inches. I felt sure that with the proximity to the frontal boundary, there would be some rain. But as someone in class pointed out, frontal boundaries that make it all the way to tropical Florida often lose their punch in that climate. I did not consider this, and frankly was doing a bit of upstream forecasting in that this same front had come through the Midwest and packed a good punch, dropping our temps by 30 degrees and giving us a brief spate of thunderstorms during the overnight hours. I was cautioned not to think the same way about tropical climates as I do about land-mass climates such as the Midwest. Fronts and storms usually do not behave the same between the two. Other factors are solar heating and warm waters nearby, which diffuse the fronts. But in addition, surface and upper level dynamics are not as strong an influence - not a lot of temperature gradient going on in the warm season in the tropics. Good lesson.
I was only 2 knots too high on the max wind speed this time. I was trying to factor in the expected rain/thunderstorms plus offshore winds, as well as the pressure gradient with the approaching front. And in looking again at the surface analysis for 00Z 9/14, the trough is sitting just west of Orlando, with the frontal boundary not far behind in the Gulf. On the MOS, the winds were again light at 06Z 9/13 (6 kts GFS, 6 kts ETA, 9 kts NGM), then showed an increase at 15Z/21Z to-12-13 kts. This would indicate the same mixing eddies with afternoon heating as yesterday's prediction. MOS was way off on the winds.
Forecast for 06Z 9/14/06 to 06Z 9/15/06
Actuals (Climatology) for 06Z 9/14/06 to 06Z 9/15/06
High 89 Low 76/73 Precip .21 Max Wind 21 mph/18.25 kts
Normals for September 14: High 91 Normal Low 72 Normal Precip .20 inches
Extremes for September 14: High 96 (1990) Low 67 (1956) Precip 2.24 inches (2001)
The METARS showed the high as 89, but the low as 76, again 3 degrees warmer than the Climate Report. The low occurred at 12Z 9/14, and the high occurred at 00Z 9/15. This METAR was more fun because from 2341Z 9/14 to 0053Z 9/15, the remarks indicate thunderstorms and rain throughout this period. Actually, CB (cumulonimbus) is remarked upon at 1753Z 9/14, which is the time period the convective activity was forming. These various thunderstorms were located WNW, N, and E of the airport, and all were moving E. Also of note is that the temperature dropped during these thunderstorms to 27° C (80.6°F) at 2341Z 9/14, to 25°C (77°F) at 0015Z 9/15, and to 23°C (73.4°F) at 0046Z 9/15. This low of 73 was recorded in the Climate Report, whereas the METARS low of 76 was recorded from the "1 group," which is one of the standard times of 06Z, 12Z, 18Z, 00Z, and 06Z for the period. Though I was 2 days late on obtaining a radar image from the RAP-NCAR site by the time I thought of including it, I was able to access the archived loop from the SPC Archive for 2042Z 9/14 to 0040Z 9/15. The shower activity is so small and so quick near Orlando that you really have to watch closely to catch it on the loop! (I couldn't save the loop or link directly to it, but if you choose the Base Reflectivity Mosaic for September 15 00Z Past 6 hours, it will take you to the loop).
Orlando was sitting in the warm sector of the frontal boundary and I didn't factor that into the high - though it only made a difference of 2 degrees. The same flip/flop low factor occurs here, where the METAR is higher and the Climate Report is lower than my forecast low. Given the METAR, I was 2 degrees too cool, or a degree too warm by the Climate Report. Again, cloud cover was probably a factor. My winds were way too low on this one, by 8.25 knots. I don't think I'm catching the sea breeze possibilities, as these winds are most likely offshore, and they do have the frontal boundary behind them.
The MOS for September 13th showed the 06Z 9/12 to 06Z 9/13 high and low temps as follows:
GFS Low 74 High 89 ETA Low 74 High 86 NGM Low 73 High 88 GFSX Low 74 High 89
The only consensus on the high here was between the GFS and the GFSX, and since I saw the frontal boundary coming through the Orlando area between 09Z and 18Z, I went with the cooler 86 degree high in expectation of evaporational cooling from precip activity. My low of 74 was in conjunction with the consensus of three of the models. However, the actual high of 89 is in line with afternoon heating, which I thought would not have the chance to occur to that extent. The rains came in overnight, which blew my theory on that one.
But the rains did come this time! I predicted .5 inches, and actual was .21, but again, I was factoring in the 24 hr Day 1 QPF valid 12Z 9/14 to 12Z 9/15, which shows Orlando in the .5-.75 range. Though storms did occur, they did not produce the rainfall amounts that the QPF expected. I was also thinking that the frontal passage would enhance the sea breeze fronts, but this would not be the case in the overnight hours. The FOUS data for NAM 06Z 9/14 and NGM 12Z 9/14 indicated small amounts of precip in each 6 hour period with as much as 1.15 inches in the 36 hour period after 06Z on the NAM (note that this is after 06Z 9/14 for NAM and after 12Z for NGM). This would correspond to 12Z-18Z 6 hour period on 9/15, which is after our forecast period, so I did not find this useful.
On max winds this time, I was 8.25 knots too low, which can probably be accounted for by the presence of the convective activity and the pressure gradient with the frontal passage. The MOS was even lighter this time at 06Z 9/14 (GFS 5 kts, ETA 6 kts, NGM 5 kts), with an increase to only 8-9 kts in the 15Z-21Z timeframe. This is a major bust, and I now realize why you can't just go by MOS for precip and winds. MOS is all about trends, whereas frontal boundaries and upper level troughs can't be figured into that since they're not consistent or uniform.
for 06Z 09/15/06 to 06Z 09/16/06
Actuals (Climatology) for 06Z 09/15/06 to 06Z 09/16/06
High 92 Low 81/73 Precip 0.01 Max Wind 17 mph/14.77 kts
Normals for September 15: High 91 Normal Low 72 Normal Precip .20 inches
Extremes for September 15: High 94 (1992) Low 63 (1956) Precip 1.67 inches (1957)
The METARS show the high as 92, but the low as 72. The low occurred at 12Z 9/15, and the high occurred at 00Z 9/16. After the frontal passage and the thunderstorms in the last forecast period, instead of a cool-down, there was a warm-up in temps.
Again, big bust in the high by not factoring in the warm sector and afternoon heating. The low on the METAR crept up even higher, as well, busting my low forecast by 7 degrees - an egregious error - unless you look at the Climate Report, where my bust was only by one degree too warm. My prediction of .5 of rain was based on the QPF and the eWall progs, which overestimated the amounts in the Orlando area. Winds again were way too low, though only by 4.77 knots this time.
The MOS for September 14th showed the 06Z 9/15 to 06Z 9/16 high and low temps as follows:
GFS Low 74 High 92 ETA Low 74 High 88 NGM Low 72 High 91 GFSX Low 74 High 92
Predicted high temps rose for this day, as did the Actuals. I failed to factor this in as I was still thinking in upstream forecasting mode wherein the frontal passage would be expected to cool temps. Instead, immediately upon the passage of the front, the temps warmed back up into the 90s, for which climate would be accountable. This may be the end of summer, but summer it is, still. And Florida is not letting it go too easily. I was reminded by a classmate that Hurricane Wilma was threatening Florida last year in October, and the temps were in the mid-80s at that time. Good lesson, again. And I didn't factor in climate norms, which would have helped with precip, too.
Precip was minimal this go around, even though as I stated above, my prediction was based on the QPF and the eWall precip progs (06Z 9/15, 12Z 9/15, 18Z 9/15, and 00Z 9/16). The QPF actually shows Orlando in the .10 range, but the eWall precip progs show potential for larger amounts.
My max winds were low by 4.77 knots, not as much as the last forecast, but still not accurate. The average wind speed for 9/12-9/16 (from a link on the Climate Summary provided in the text) is 7-7.5 knots, which would make my low prediction of 10 knots a bit high. The MOS at 06Z 9/14 was lighter still (GFS 3 kts, ETA 5 kts, NGM 3 kts), with the usual increase in the 15Z-21Z timeframe to 6-8 kts. I feel that the sea breeze fronts and offshore winds are skewing these MOS totals, as they are not factored into the MOS calculations.
Overall, as a forecaster, I need to delve more deeply into details, look more closely at surface and upper air analyses, and pay attention to local climate of the forecast site rather than trying to "upstream" or "downstream" based on my location. My strength is looking at the big picture (at least, I think it is). Winds and precip are my weaknesses, and warrant review of relevant chapters in Meteo 101. I also think I should seek out the graphics to backup the NWS AFDs in order to better understand them. In fact, that might be the biggest key to becoming more accurate. Backup graphics for the AFDs is one goal, and another would be to develop a reference sheet of the MOS weaknesses to refer to when reviewing them.
In the discussion threads, no one really discussed winds much, and now I wish they had. There was one good point brought up about lapse rates being dry adiabatic which would indicate winds mixing down to the surface. I believe a good information source for the winds might be the forecast soundings to get a feel for this mixing/lapse rate situation. However, it was pointed out that cloud cover would cancel this factor out. Another tool that I should have used that might have helped my wind prediction problem is the 2m grid interpolations and 850 mb method (Delta Method).
One thing I didn't use was the FOUS for predicting precip. I was never comfortable with it, and found reviewing the rest of the data charts (surface analyses, QPFs, progs, MOS, etc.) so time-consuming I never got around to dealing with the FOUS indepth. I posted once on the 14th to test my interpretation of the data, and found that I was off on Z timing. Also, I found it disconcerting that I couldn't get a FOUS readout for our specific city, and didn't catch on for awhile to which city listed in FOUS was closest to Orlando. Obviously, this needs to be revisited so this tool can be used in the contest, but I admit I was very discouraged at first at my inability to grasp its usefulness..
In summary, improvement would most likely come from the following goals:
Of note is the fact that our practice week was rather chaotic due to our online-only-student inability to follow the daily deadline format. Even though the format was changed, it was still difficult for some to be able to post in a timely or consistent manner during daytime hours. Also, we weren't able to practice like we'd wanted to on the new format since we needed to concentrate this week on our ePortfolio assignments. I note these things not as an excuse for anything, but as a diary entry for future 410 classes to see the growing pains in this Certificate Program's first incarnation.
I do want to say, however, that in dealing with the format problems, trying to make a post by noon and trying to get a forecast down by the end of the workday, plus trying to follow threads, going through all the model loops and accessing various sites for information, plus the work I was actually supposed to be doing at my desk, I did not make use, or make full use, of a number of tools that would have helped me to be more accurate. This is my main shortcoming at this time.
September 24, 2006