NE.Wx Snowfall Forecast Contests

Friday, October 26, 2018

Winter '18 / '19 - VCP32: Just Not The Same

by TQ @ 3:15 PM

Saturday, October 20, 2018

Winter '18 / '19 - How Much For Philly?

by TQ @ 9:27 AM

A weak-to-moderate El Niño continues to be the consensus forecast for Winter '18 / '19.

Weak-to-moderate El Niño conditions -- defined generally as sea surface temperature anomalies at least 0.5°C above normal near the International Dateline in the tropical Pacific ocean -- correlates with more than a few more inches of snow than average in and around Philly.

One problem; average isn't always the best metric to describe what's typical or most likely to occur. That honor frequently falls to the median.

As such ... the median snowfall is four inches less than the 23" 51-year average suggesting this winter's season-total snowfall in Philly could very be up to seven inches more than most other winters.

Saturday, October 13, 2018

Winter '17 / '18 - Correlation of Eurasia's OCT Snow Cover and Season-total Snowfall in NE and M-A Regions

by TQ @ 9:56 PM

REPOST from OCT-17 (lightly edited for clarity)
Updated with Winter '17 / '18 verifications for RIC and NYC and their outlooks for Winter '18 / '19

---
The correlation between the areal coverage of Eurasia/s OCT snow cover and season-total snowfall has become broad-brushed conventional wisdom (CW) following the innovative research by AER climatologist Dr. Judah Cohen.

But ... just how well does the CW hold up for NEWxSFC/s forecast stations across New England (NE) and the mid-Atlantic (M-A) regions?

To find out ... monthly period-of-record areal snow cover data for Eurasia from Rutgers Global Snow Lab were correlated with season-total snowfall data for the 27 NEWxSFC/s stations.

A positive and statistically significant correlation means the greater the areal snow cover over Eurasia in OCT ... the greater the season-total snowfall for the following winter.

An Excel radar chart shown below depicts the results of the analysis.

DISCUSSION: Stations between the inner and outer rings have a positive correlation coefficient statistically different than zero. The coefficients range between 0.289 (CAR) and 0.424 (ORH). Correlation values in this range are classified generally as 'low' (moderate: >= 0.5 - 0.7; strong: >= 0.7 - 0.9). Even though the correlations are weak ... they can still provide useful information for seasonal snowfall forecasts.

Translation: greater season-total snowfall over select stations in the NE and M-A is correlated with greater OCT areal snow cover in Eurasia.

---
The analysis showed other significant correlations of interest.

- RIC/s season-total snowfall has a positive correlation with Eurasia/s AUG areal snow cover.
Eurasia/s AUG-17 snow cover was well below normal ==> lower season-total snowfall @RIC this winter.

VERIFICATION Winter '17 / '18: RIC STP 12.4" (AVG: 13.2")
OUTLOOK Winter '18 / '19: Eurasia's AUG snow cover below average ==> STP below AVG

- NYC/s season-total snowfall has a negative correlation with Eurasia's JUN areal snow cover.
Eurasia/s JUN-17 snow cover was above normal ==> lower season-total snowfall @ NYC this winter.

VERIFICATION Winter '17 / '18: NYC STP 35.4" (AVG: 26.1")
OUTLOOK Winter '18 / '19: Eurasia's JUN snow cover below average ==> STP above AVG

GREEN (RED): positive (negative) correlation between monthly Eurasian areal snow cover and season-total snowfall.

VERIFICATION Winter '17 / '18
Eurasia's OCT-17 areal snow cover was greater than average (12,051,667 km^2 v 10,261,134 km^2).
Positive correlations for the stations listed below suggest Winter '17 / '18 STP would be above average

STN: STP" / AVG"
ABE: 46 / 31.4
BOS: 58.6 / 41.6
BTV: 81.5 / 69.8
PWM: 91.1 / 63.8

---
FINDINGS: data analysis supports the CW for NE forecast stations but not so much across the M-A.

Thursday, October 04, 2018

Winter '17 / '18 - Arctic Oscillation (AO) Analog Verification

by TQ @ 11:35 AM

The analog forecasting technique seeks similarities to the AO state in the run-up to the coming winter with AO run-up states of winters past. Presented here is the verification of AO analogs for the '17 /'18 winter.

The NEWxSFC method ranks analog years by their the sum of square errors (SSE) statistic.
Lower SSE errors ==> stronger analog

Constraining the number of analog winters for analysis to five is arbitrary.

Arctic Oscillation (AO) Index Analog Forecast Verification

In the run-up to Winter '17 / '18 ... '73 / '74 was the leading analog followed by '08 / '09 ... '54 / '55 ... '95 / '96 ... and '75 / '76. Winter '17 / '18 AO started weakly negative and remained weakly negative until meteorological winter's end. MAR-18 AO (not shown) crashed to -0.941.

A qualitative assessment of the forecast's accuracy would rate all analogs except '73 / '74 as 'poor' ... IOW ... useless.

OTOH ... analog #1 mimicked the observed behavior of the AO associated with a weak La Nina; whereas ... . strong La Nina conditions prevailed during the winters of '73 / '74.

CONCLUSION: The analog forecasting technique provided useful guidance for Winter '17 / '18.

---

An alternative forecasting technique looks at the AO/s 'sign' (i.e., positive or negative) for any calendar-year's month preceding the pending winter as a potential leading indicator of the AO's sign for upcoming D-J-F period.

Results from a chi-square 'test for independence' infers a statistically significant relationship ... at the 95% confidence level and a p-value < 0.05 ... between NOV's AO sign and AO's sign of the upcoming D-J-F period.

IOW ... if NOV's AO is negative (positive) ... then the average AO state during the upcoming winter will also be negative (positive); although the classification model is stronger ... i.e., lower false alarm rate ... for the predictor's month with negative signs than positive. This is opposite of the relationship found for the NAO.

BOTTOM LINE: If NOV's AO is negative ... chances are good the AO state will average negative during the D-J-F period.

NOV '17 AO/s sign was negative (-0.078). The 2x2 contingency technique predicted correctly AO state for Winter '17 / '18 would average less than zero. AO for the D-J-F period averaged (0-.076).

Saturday, September 29, 2018

Winter '17 / '18 - North Atlantic Oscillation (NAO) Analog Verification

by TQ @ 8:43 PM

The analog forecasting technique seeks similarities to the NAO state in the run-up to the coming winter with NAO run-up states of winters past. Presented here is the verification of NAO analogs for the '17 /'18 winter.

The NEWxSFC method ranks analog years by their the sum of square errors (SSE) statistic.
Lower SSE errors ==> stronger analog

Constraining the number of analog winters for analysis to five is arbitrary.

North Atlantic Oscillation (NAO) Index Analog Forecast Verification

In the run-up to Winter '17 / '18 ... '90 / '91 was the leading analog followed by '94 / '95 ... '54 / '55 ... '04 / '05 ... and '02 / '03. Winter '17 / '18 NAO started strongly positive ... strengthened through the heart of the season then crashed below zero following meteorological winter's end

A qualitative assessment of the forecast's accuracy would rate the analogs #2 ... #3 ... and #4 as 'poor' ... IOW ... useless.

OTOH ... analogs #1 and #5 mimicked the observed behavior of the NAO associated with a weak La Nina. Moderate El Niño conditions prevailed during the winters of '94 / '95 and '02 / '03.

CONCLUSION: The analog forecasting technique provided useful guidance for Winter '17 / '18.

---

An alternative forecasting technique looks at the NAO/s 'sign' (i.e., positive or negative) for any calendar-year's month preceding the pending winter as a potential leading indicator of the NAO's sign for upcoming D-J-F period.

Results from a chi-square 'test for independence' infers a statistically significant relationship ... at the 95% confidence level and a p-value < 0.05 ... between NOV's NAO sign and NAO's sign of the upcoming D-J-F period.

IOW ... if NOV's NAO is negative (positive) ... then the average NAO state during the upcoming winter will also be negative (positive); although the classification model is stronger ... i.e., lower false alarm rate ... for the predictor's month with positive signs than negative. This is opposite of the relationship found for the Arctic Oscillation (AO).

BOTTOM LINE: If NOV's NAO is positive ... chances are good the NAO state will average positive during the D-J-F period.

NOV '17 NAO/s sign was zero (0) for the 1st time in the index's 68 year record. The 2x2 contingency technique offered no useful guidance.

Saturday, April 14, 2018

Winter '17 / '18 - 19th Annual 'Regular Season' Snowfall Forecast Contest: Season Summary and Historical Perspective

by TQ @ 9:03 PM

FINAL results here.

EIGHT contest-worthy storms this season ... one more than average.
The storm count tied for 6th place with four other Winters ('07 / '08 ... '08 / '09 ... '10 / '11 ... and '13 / '14).

Contest-worthy Storm Count by Month
DEC - 1 [25-DEC-17]
JAN - 2 [04-JAN-18; 17-JAN-18]
FEB - 2 [07-FEB-18; 17-FEB-18]
MAR - 3 [07-MAR-18; 12-MAR-18; 21-MAR-18]
TOT - 8

Verified forecasts

http://www.newx-forecasts.com/NEWxSFC_19/storms/storm1_verifications_25Dec17.htm

http://www.newx-forecasts.com/NEWxSFC_19/storms/storm2_verifications_04Jan18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm3_verifications_17Jan18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm4_verifications_07Feb18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm5_verifications_17Feb18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm6_verifications_07Mar18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm7_verifications_12Mar18.htm
http://www.newx-forecasts.com/NEWxSFC_19/storms/storm8_verifications_21Mar18.htm

FINAL Results and Storm Summaries

http://www.newx-forecasts.com/NEWxSFC_19/storms/storm1_summary_25Dec17.htm

Wednesday, April 11, 2018

Winter '17 / '18 - 19th Annual 'Regular Season' Snowfall Forecast Contest: FINAL Results

by TQ @ 4:17 PM

Under the ‘two-thirds’ rule … forecasters who entered at least SIX forecasts are included in this season's FINAL standings.

Full table with all other error statistics at the Contest/s web site here (direct link).
Individual forecaster's storm statistics here (direct link).

Best Forecasts by Storm

Top 10 Forecasts (Error statistic: SUMSQ Z)

SUMSQ Error Z is the primary measure of forecaster skill (lower the better).
Accounts for the magnitude and distribution of the storm-total snowfall for all stations.

Top 10 Forecasts (Error statistic: Total Absolute Error Z - TAE)

TAE Error Z is the secondary measure of forecaster skill (lower the better).
Accounts for the magnitude of snowfall at each station.

Top 10 Forecasts (Error statistic: RSQ Z)

R-squared (RSQ Z) is a supplementary measure of forecaster skill (higher the better).

Accounts for how well the variability of the observed snowfall was accounted for by the forecast.

---
Eighteen unique forecasters submitted a total of 2,582 stations forecasts.
Eight forecasters entered all 8 contests.
Three forecasters entered 7 contests.
The remainder entered fewer than six

---
Hope to see y'all again next winter.

Saturday, April 07, 2018

Winter '17 / '18 - 'Season-total' Snowfall Forecast Contest: Month of MAX Snowfall

by TQ @ 9:45 PM

BOS Red Sox Opening Day
14-APR-53

Forecast stations observing their MAX monthly snowfall during Winter '17 / '18 in ...
DEC - 0
JAN - 7 (CAR ... PVD ... ACY ... SBY ... RIC ... ORF ... RDU)
FEB - 0
MAR - 18 (all the rest)

Leaving aside for the moment the fact FEB has 28 days and all other snow-season months have 31 ... 72% of our forecast stations observed their highest monthly snowfall total in MAR -- yes MAR -- this season.

Climo finds most of the contest/s forecast stations (14) observe their MAX monthly snowfall in JAN. MAX monthly snowfall happens in FEB at 10 of the remaining 11 stations with only BDL having its greatest monthly climo snowfall in DEC.

No station/s monthly snowfall climo reached its maxim in MAR which makes the winter '17 / '18 all the more remarkable.

Friday, April 06, 2018

Winter '17 / '18 - 17th Annual 'Season-total' Snowfall Forecast Contest: FINAL Results

by TQ @ 8:21 PM

UPDATE 2: 10:45 PM EDT ... FRI ... 06-APR-18

All better (I hope)

UPDATE: 10:15 PM EDT ... FRI ... 06-APR-18

Senior NEWxSFC forecaster Roger Smith was inadvertently omitted from the FINAL Results summary (operator error - Excel copy down mistake). Fix in the works

h/t Roger 'Eagle Eye' Smith

---

Complete station-by-station forecaster verifications at the Contest/s web site here (direct link)

Perfect forecasts
iralibov @ PVD (47")

Near misses (< 1" error)
Mitchel Volk
0.1" @ ALB (74.9")
0.2" @ PHL (29.8")
0.2" @ ORF (15.2")
0.3" @ CON (85.3")
0.4" @ NYC (35.4")

Any.wx
0.1" @ BDR (39.9")
0.5" @ BTV (81.5")

33andrain
0.1" @ IAD (11.9")
0.6" @ BOS (58.6")

iralibov
0.1" @ RDU (8.9)
0.6" @ BOS (58.6")

kevinmyatt
0.4" @ NYC (35.4")
0.4" @ RIC (12.4")

Brad Yehl
0.2" @ BDL (48.4")

donsutherland1
0.6" @ EWR (34.4")

MarkHofmann
0.8" @ DCA (7.8")

Smallest station forecast error yet percent error >= 10%
Mitchel Volk
11.1" @ PWM (91.1")
2.2" @ SBY (17.2")

TQ
11.1" @ PWM (91.1")

Any.Wx
11.1" @ PWM (91.1")

donsutherland1
16.5" @ BGR (109")

33andrain
8.9" @ ORH (88.9")

iralibov
10.2" @ ACY (34.2")

Roger Smith
2.2" @ SBY (17.2")

snocat918
1.6" @ BWI (15.4")

MarkHofmann
0.8" @ DCA (7.8")

---

Season-total snowfall from all stations (1,234") came in 35% above the period-of-record normal (914"). Winter '17 / '18 ranks 4th among other Contest years since winter '04 / '05 ... the season when BGR and BGM were added to the station list.