La Nina and Oklahoma Summer Precipitation

NCEP analysis

Figures are from: ENSO Temperature & Precipitation Composites


Note the relatively large signal in Oklahoma and northern Texas in the upper left of the figure immediately below. There is some "folk lore" here in Norman Oklahoma that La Nina summers tend to be dry. The figure below supports that conjecture.

lanina.jja.precip.gif


Here is the corresponding El Nino image. The El Nino effect is not investigated on this page.

elnino.jja.precip.gif


Reinvestigate NCEP data

Obtain daily data from ftp://ftp.cpc.ncep.noaa.gov/precip/CPC_UNI_PRCP/GAUGE_CONUS/ for years 1948 through 2010. (The missing data for day 20070226 was faked by substituting an average of 20070225 and 20070227). Composite into 756 months. Analyze with Python:

jjarain.png

See a slide show of the annual cycle of precipitation.


From the text on the NCEP image for JJA precipitation in La Nina years, we take the La Nina years to be:

1950 1954 1955 1956 1964 1970 1971 1973 1974 1975 1985 1988 1998 1999 2010

My composite images retain the resolution in the original dataset. The pixels, or bins, in the NCEP composite (see inset) appear larger. There seems to be good agreement between the image and the inset, indicating the data has been composited correctly:

laninajjachangeINSET.png


Same image as above, but a percent change:

jjarainpercent.png


Presumably, climate customers prefer an accurate forecast for a specific location (their house, their farm). As in my windfarm study, we forego that customer preference and examine average rainfall over a region. Except for reservoir managers and similar people, such information is less useful than a point forecast. But hopefully the climate signal will be stronger and have more significance. We make a box (magenta) surrounding Oklahoma and northern Texas.

Here is plot of all the t-values:

jjatvalue.png

The average t-value in the box is t=-1.25. The t-value of average rain in the box is t=-1.55, and p=.07. The fraction of the valid data area that lies within the box is f=0.048. Note that in my windfarm study I looked for p<.05f to declare significance. Some details of the statistical analysis will appear below.


Here is the anomaly in the box for all years 1948-2010. Note the if we could kick 1950 out of the set of La Nina years, we could have a convincing case that La Nina years are significantly drier:

rainbar.png


Here is an image from Kim et al. (2011) that I was familiar with (but note it was prepared for the months July through October). The image suggests there is justification for kicking out 1950: the East Pacific Cooling is very modest; the normalized Nino-3 index of Kim et al. (2011) was only about -0.2. Also, in 1974 it was only -0.3, which had a slightly wet summer. All the other years designated as La Nina by NCEP have a normalized Nino-3 index of <-0.75. In the image I have written an A to indicate NCEP's La NinA designation.

nino3nino4.png

I downloaded the ersstb data and as able to reproduce the above plot. Coincidently, 1950 is the one year that has a discernible difference:

JASOindices.png

But the above is from the detrend of the seasonal composite time series. An alternative is to detrend the entire time series of the anomalies in the data file, up to the end time (December 2006, for Kim et al.), before making the seasonal composites. This looks closer to the Kim et al. figure:

JASOindicesD.png


So having convinced myself I was using the data correctly, I plotted the indices for June through August. Here 1950 looks more consistent as a La Nina year:

JJAindices.png


1950, a wet La Nina year in Oklahoma:

jjad1950.png

Statistical Analysis

retain 1950 as La Nina year

n=15
mean= -286.23 (.1 mm) 
standard error= 185.20
t-value= -1.55
2*p= 0.14
90% confidence interval: -612.44 < true mean < 39.96 
resample method 90% interval: -575.3 < true mean < 12.01  

=== kick 1950 out as La Nina year ===
 -575.301945808 12.0111291557
{{{
n= 14  
mean= -415.32 (.1 mm)
standard error= 142.66
t-value= -2.91
2*p= 0.012
90% confidence interval: -667.95   < true mean < -162.69
resample method 90% interval: -644.5 < true mean <  -192.54 

  1. NCEP ENSO page has animations of recent SST.

  2. Australia Bureau of Meteorology ENSO monitoring page

  3. Australia Bureau of Meteorology Model Outlooks of ENSO Conditions

  4. Australia ENSO wrap-up

  5. Yearly Climate Graphs for Selected Locations in Central/Western Oklahoma and Western North Texas

  6. OCS drought monitoring

On 5 June 2012: Australia ENSO wrap-up states: No climate models favour a return to La NiƱa. As of 16 June 2012, OCS drought monitoring shows no drought this year in Oklahoma.

Appendix

# in the following: st=1 -> La Nina , st=-1-> El Nino
# delta is the JJA anomaly in units of 0.1 mm 
#year st delta 
1948 0  596.25
1949 0 -103.40
1950 1 1520.96
1951 0  373.87
1952 0 -1149.30
1953 0 -313.70
1954 1 -1358.54
1955 1 -609.56
1956 1 -1181.49
1957 -1 -258.58
1958 0  538.12
1959 0  424.53
1960 0  463.94
1961 0  530.71
1962 0  585.26
1963 -1 -366.67
1964 1 -496.76
1965 -1   47.62
1966 0   55.40
1967 0   44.13
1968 0  143.21
1969 0 -263.35
1970 1 -1019.90
1971 1   85.14
1972 -1  -91.75
1973 1 -239.61
1974 1   30.85
1975 1  137.25
1976 0 -724.09
1977 0  -44.96
1978 0 -768.82
1979 0  511.22
1980 0 -1283.33
1981 0  130.76
1982 -1   10.99
1983 0 -905.43
1984 0 -772.69
1985 1  136.71
1986 0  373.37
1987 -1   85.75
1988 1 -519.55
1989 0 1125.00
1990 0 -540.43
1991 -1  127.18
1992 -1 1236.93
1993 0 -185.67
1994 -1 -184.31
1995 0  641.34
1996 0 1306.06
1997 -1  496.62
1998 1 -838.16
1999 1 -236.21
2000 0 -132.17
2001 0 -915.12
2002 -1    6.17
2003 0   31.09
2004 -1 1120.59
2005 0  597.69
2006 0 -471.55
2007 0 1316.23
2008 0  577.01
2009 -1  271.83
2010 1  295.32
rainbox.txt