Winter '09 / '10 - Do Different Flavors of +ENSO Affect BWI Snowfall?

"Looking back at past winters since 1950, approximately 17 were influenced by an El Niño episode.

"Seasonal snowfall averaged above normal for weak and moderate El Niño winters while below normal for strong El Niño episodes."

 - El Niño and DC/Baltimore Winters (WFO LWX - NOV-09)

Judging from the LWX chart...those conclusions would appear reasonable and true.  The lime green column is tallest of them all.  The red column is below climo.

Frank Royance...quoting LWX 'Warning Coordination' meteorologist Chris Strong in his Maryland Wx blog...

"With moderate strength El Nino's (sic) [like this one] we have statistically the greatest chance of above-normal snowfall."

"What they're saying is that not all El Niño winters are alike for the mid-Atlantic states. Some will be snowy; some not. Here's how they tend to break down, according to Klein:

"On average, weak El Niño winters bring below-average temperatures and below-average precipitation. Not generally conducive to lots of snow.

"Strong El Ninos, on average, bring us above-normal temperatures and precipitation. The cold air tends to remain well to our north, so most of the precipitation falls as rain rather than snow."

Sadly..no.

The LWX analysis lacks any significance testing to determine whether the average season-total snowfall for the various +ENSO states are statistically different from each other or climatology (1971 - 2000) or the period-of-record since 1950.

Bottom line up-front:  BS.  Despite appearances...they are not different.  There is no statistically significant difference...between average season-total snowfall observed during weak...moderate...or strong +ENSO winters when compared to each other nor to the period-of-record since 1950 or climatology (1971 - 2000).

The study's conclusions about the probability of observing more or less season-total snowfall relative to 'normal' is unsupported at the 95% confidence level.  There is weak evidence...at the 90% confidence level...to conclude 'moderate' +ENSO winters observed more snowfall...on average...compared to climatology (1971 - 2000) and the period-of-record beginning in 1950...which is pretty weak tea.

Data and analysis after the jump.

+ENSO years / season-total snowfall

Weak (ge 0.5°C < 1.0°C)
1963-64 / 51.8"
1977-78 / 34.3"
1969-70 / 21.0"
1987-88 / 20.4"
2004-05 / 18.0"
1976-77 / 11.1"
2006-07 / 11.0"
AVG:  23.94"

Moderate (ge 1.0°C < 1.5°C)
2002-03 / 58.1"
1986-87 / 35.2"
1965-66 / 32.8"
1968-69 / 18.6"
1994-95 / 8.2"
AVG:  30.58"

Strong (ge 1.5°C)
1957-58 / 43.0"
1982-83 / 35.6"
1991-92 / 4.1"
1997-98 / 3.2"
1972-73 / 1.2"
AVG: 17.42"

-----
z-Test: Two Sample for Means - Moderate and Weak +ENSO
Mean 30.58 23.94
Known Variance 285.60 181.75
Observations 5 7
Hypothesized Mean Difference 0
z = 0.7282
P(Z<=z) one-tail  = 0.2333
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.4665
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis...i.e...there is no difference in average snowfall.

There is no statistical difference in the average season-total snowfall between 'moderate' and 'weak' +ENSO events.

-----
z-Test: Two Sample for Means - Weak and Strong +ENSO
Mean 23.9 17.4
Known Variance 181.75 325.51
Observations 7 5
Hypothesized Mean Difference 0
z = 0.6835
P(Z<=z) one-tail = 0.2471
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.4943
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis.

There is no statistical difference in the average season-total snowfall between 'weak' and 'strong' +ENSO events.

-----
z-Test: Two Sample for Means - Moderate and Strong +ENSO
Mean 30.58 17.42
Known Variance 285.6 325.51
Observations 5 5
Hypothesized Mean Difference 0
z = 1.1904
P(Z<=z) one-tail = 0.1170
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.2339
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis

There is no statistical difference in the average season-total snowfall for 'moderate' and 'strong' +ENSO events.

 -----
z-Test: Two Sample for Means - All Years and Weak +ENSO
Mean 19.83 23.94
Known Variance 196.8 181.75
Observations 60 7
Hypothesized Mean Difference 0
z -0.761
P(Z<=z) one-tail 0.223
z Critical one-tail 1.645
P(Z<=z) two-tail 0.447
z Critical two-tail 1.960

Z < z-critical fails to reject null hypothesis.

There is no statistical difference between average season-total snowfall for 'all years' and 'weak' +ENSO events.

 -----
z-Test: Two Sample for Means - All Years and Moderate +ENSO
Mean 19.83 30.58
Known Variance 196.8 285.6
Observations 60 5
Hypothesized Mean Difference 0
z - = 1.384
P(Z<=z) one-tail = 0.083
z Critical one-tail = 1.645
P(Z<=z) two-tail = 0.166
z Critical two-tail = 1.960

Z < z-critical fails to reject null hypothesis.

There is no statistical difference in the average season-total snowfall between 'all years' and 'moderate' +ENSO events.

-----
z-Test: Two Sample for Means - Climo (1971 - 2000) and Weak +ENSO
Mean 18.15 23.94
Known Variance 167.6 181.75
Observations 30 7
Hypothesized Mean Difference 0
z = -1.0313
P(Z<=z) one-tail = 0.1512
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.3024
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis.

There is no statistical difference in the average season-total snowfall between Climo (1971 - 2000) and 'weak' +ENSO events.

-----
z-Test: Two Sample for Means - Climo (1971 - 2000) and Moderate +ENSO
Mean 18.15 30.58
Known Variance 167.6 285.6
Observations 30 5
Hypothesized Mean Difference 0
z = -1.5697
P(Z<=z) one-tail = 0.0582
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.1165
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis.

There is no statistical difference in the average season-total snowfall between Climo (1971 - 2000) and 'moderate ' +ENSO events.

-----
z-Test: Two Sample for Means - Climo (1971 - 2000) and Strong +ENSO
Mean 18.15 17.42
Known Variance 167.6 325.51
Observations 30 5
Hypothesized Mean Difference 0
z = 0.0868
P(Z<=z) one-tail = 0.4654
z Critical one-tail = 1.6449
P(Z<=z) two-tail = 0.9308
z Critical two-tail = 1.9600

Z < z-critical fails to reject null hypothesis.

There is no statistical difference in the average season-total snowfall between Climo (1971 - 2000) and 'strong' +ENSO events.

-----
BWI snowfall data
CPC ENSO data
Statistical testing w/Excel/s 'z-Test: Two Sample for Means' (Data Analysis Tools Add-in)