TRENDS IN MARINE WINDS ADJUSTED FOR CHANGES IN

OBSERVATION METHOD, 1980-2002

Bridget R. Thomas1, Elizabeth C. Kent2, Val R. Swail3 and David I. Berry2

1 and 3Meteorological Service of Canada, Climate Research Branch, 1Dartmouth, NS & 3Downsview, ON

2National Oceanography Centre, Southampton, UK

Trends in Marine Winds Aim: Identify and remove spurious trends in ICOADS Winds

o Cardone et al. (1990, On trends in historical marine data. J. Climate) found that in the period 1946 to 1984, trends could be removed by adjusting winds using observing method metadata. Is this still true?

Datao ICOADS winds and observation method flago Metadata on observing heights (WMO Pub. 47)

Methodo Use only data of known method (restricts to 1980 - 2002)o Adjust for anemometer height and observing methodo Compare datasets of 5˚ x 5˚ area monthly mean winds for

subsets defined by sufficient metadata Results

o Using metadata and adjusting for height removes most but not all of the increasing trend

o Spurious trend in visual winds increases following adjustment

ICOADS Quasi-global (all available data) mean wind speed, no adjustments applied for measurement methods

The Evolving Marine Wind Observing System

Changing observing method:

o transition from visual estimates based on sea

state to more measurements by anemometers

Changing ships:

o Ships are getting bigger over time

• higher anemometers

• higher observing platforms for visual winds

• as a result, flow distortion biases may change

over time (see poster by Moat et al.)

The Evolving Marine Wind Observing System (2)

Changing logging methods (measured winds):

o Anemometer "Eye-ball" average of dial or readout

(biased to gusts?)

o Average of continually logged data stream (true mean)

o TurboWin height correction (only for a few years;

stopped at 2002)

o Changing method to calculate true wind from the

relative wind and the ship’s motion

Partial Solutions to Account for Changes

ICOADS contains a wind indicator flag (WI) which

indicates measured or estimated, and original

units; WMO Pub. 47 gives ship metadata

For wind reports identified as measured, we

adjust the winds for anemometer height, to a 10

m reference level

For wind reports identified as estimated, we

adjust the winds using the Lindau (1995)

improved Beaufort equivalent scale

Correction Methods for Wind Observations

Visual Winds (or estimated: WI = 0, 2, 3, 5)o Beaufort Scale adjustments applied following Lindau (1995): A

new Beaufort equivalent scale, Proceedings of the International COADS Winds Workshop, Kiel, Germany, 232-252.

Anemometer Winds (or measured: WI = 1, 4, 7, 8)o Adjusted to 10m height assuming neutral stability

• Over much of the ocean a good approximation• In unstable conditions winds will be slightly over corrected• In stable conditions winds will be under corrected, possibly

significantly, but these conditions are not commono Individual anemometer heights from WMO Pub. 47o Anemometer wind data were not used in this analysis if no

height was available

1) Before adjustment to 10m

Comparison of Winds of Different Methods

2) After adjustment to 10m

Comparison of Winds of Different Methods

3) Comparison to Reanalyses 10m winds

Comparison of Winds of Different Methods

Results so far

Adjusting the winds to 10m reduces the overall trend from 0.7 ms-1 to 0.5 ms-1 over 23 years: 1980 - 2002

Agreement between the trend in visual and anemometer winds worsens over the period

The trend in the visual winds actually increases (the correction applied increases as the winds increase)

NCEP1 and ERA40 Reanalysis 10m winds show little or no increase over the period for this area

10m anemometer winds increase by 0.2 ms-1, the visuals by 0.7 ms-1

What's going on?

What could affect temporal trends in estimated winds?

Ships getting largero Expect visual estimates to get smaller (observers are further

from the sea surface)o However visual estimates are getting largero And estimated winds are higher on bigger ships than on smaller

ships, the opposite of what we expect Are observers influenced by the presence of

anemometer onboard?o This would explain the trends seen: visual winds are getting

closer to the uncorrected anemometer winds over time Is there evidence for this?

o The VSOP-NA showed that visual winds at night on ships carrying anemometers were significantly higher than on those ships without anemometers

Annual mean day-night differences are larger for estimated than measured winds; differences decrease over time

What could affect temporal trends in anemometer winds?

Flow distortiono as ships increase in size, the geometry of the anemometer location

changes (see Moat et al. poster)o a change in fleet composition could lead to trends (see Thomas et

al. poster)o Thomas et al. (2005, CLIMAR Special Issue of IJC) found that 10m

ship anemometer winds were 6% higher than co-located 10 m buoy winds over the period 1980 - 1995.

Reduction in true wind calculation errorso Due to the introduction of automatic coding and logging softwareo Automated averaging of winds removes human tendency to report

gusts

Mixture of winds at observation height and 10mo Some versions of the TurboWin logging software apply wind

reduction to 10m, unfortunately we don't know which reports were corrected. More work is needed to identify them.

Conclusions

Adjustment of ICOADS wind speed to 10m reduces trends in the period 1980 to 2002

Metadata on both observation method and instrument height is vital to make the necessary corrections

But the visual winds show greater trends than either the anemometer winds or the Reanalyses

There is some evidence that observers making visual wind estimates are "influenced" by the (uncorrected for height) anemometer wind

Future work .......

Future work Extend analysis back in time to 1950

o Requires improved assignment of measurement method flag And forward in time

o Requires identification of winds already corrected to 10m by TurboWin coding software after 2001

Develop corrections to remove spurious trends in visual windso Time varying Beaufort Scale?o Identify individual ships which report anemometer winds as

visuals?o 'Calibrate' visual winds using daily pressure fields in well

sampled regions? (using dataset described by Kent and Berry poster)

Quantify the effects of flow distortion on anemometer windso Variations with ship typeo and ship size, geometry, wind direction ..... etc.

Produce corrected wind speed dataset from ICOADS

Other slides, if needed during questions

Wind Speed Indicator from ICOADS Wind speed is stored in tenths of ms-1. WI shows the method used and units in which wind speed was originally recorded

(0, 1, 3, 4 follow WMO code for Wind Indicator).

WI Description (method, units) Grouping for Analysis

0 Estimated, ms-1 VISUAL

1 Measured, ms-1 ANEMOMETER

2 Estimated, original units unknown VISUAL

3 Estimated, knot VISUAL

4 Measured, knot ANEMOMETER

5 Estimated, Beaufort force (conversion of original data, or based on documentation)

VISUAL

6 Estimated, original units unknown, or unknown method

not used in analysis

7 Measured, original units unknown ANEMOMETER

8 High resolution measurement (e.g., hundredths of ms-1)

ANEMOMETER

Frequency distribution of global wind speed (in m/s) for each ICOADS Wind Indicator,

for Jan1955

•in January 1955, most wind reports have an ambiguous wind indicator; they could be estimated or the method is unknown

•the second most frequent method indicator is for estimated winds, originally reported as Beaufort force; nearly all values are at midpoints of the Beaufort intervals

•measured winds show a more continuous distribution than estimated

Can we assume WI = 6 are all visual reports? The histograms show the distributions of wind speeds from 3

individual ships in the WI = 6 (visual OR unknown) category in January 1950.

One is clearly in Beaufort intervals and therefore visual, one is probably using an anemometer, the one on the right is unclear.

Conclusion: Exclude WI = 6 from analysis and start in 1980. In the future we must try and understand the winds with unknown

method.