Changes in Cloud Cover and Cloud Types Over the Ocean from Surface Observations, Ryan Eastman Stephen G. Warren Carole J.

Transcription

1 Changes in Cloud Cover and Cloud Types Over the Ocean from Surface Observations, Ryan Eastman Stephen G. Warren Carole J. Hahn

2 Clouds Over the Ocean The ocean is cloudy, more-so than land Cloud Type Annual Average Amount (%) Land Ocean Fog Stratus (St) Stratocumulus (Sc) Cumulus (Cu) Cumulonimbus (Cb) Nimbostratus (Ns) Altostratus (As) Altocumulus (Ac) High (cirriform) Total cloud cover Clear sky (frequency) Due to large amounts of low clouds: Stratus, Stratocumulus, and Cumulus

3 Low Clouds & Solar Radiation Low clouds scatter sunlight back to space during the day. Albedo ~ 30-40% on average Marine low clouds often have an inverse relationship with Sea Surface Temperature (SST) more low cloud is associated with lower SST (Not everywhere; warm SST can cause convection).

4 Low Clouds & Infrared Radiation Low clouds absorb and emit infrared (IR) radiation during the night and day Surface and cloud temperatures nearly match From a top of atmosphere (TOA) perspective, these clouds are nearly indistinguishable from the surface in IR

5 Low Clouds & Radiation Visible Stratus deck clearly seen off of California IR Stratus deck is barely detectable, but cold, high clouds are visible

6 High Clouds & Radiation High clouds transmit more solar radiation than low clouds High clouds are colder, emit less IR At the TOA, less radiation is emitted to space This produces a weak net warming effect in the atmosphere, since more radiation comes in, and less goes out - compared to a low cloud scenario

7 Low Clouds & SST Low clouds interact with the sea surface locally High clouds not directly affected by SST In marine stratus decks, warm SST brings about lower static stability, causing a transition from stratiform to cumuliform cloud cover (assuming no change in overlying atmosphere)

8 Clouds, Radiation, and SST Low Clouds - Cool the ocean surface High Clouds - Warming in the atmosphere SST - Inversely correlates with low stratus cloud Leads to a positive feedback between low marine clouds and SST: SST warms Clouds break up More solar radiation Surface warms SST cools Clouds spread Less solar radiation Surface cools Other factors also at play subsidence and atmospheric temp, based on surrounding largescale meteorology

9 Cloud Data Clouds can be observed from... Surface looking up Pros: Cloud type information Long record ( ) Cons: Limited geographic coverage Satellite looking down Pros: Good geographic coverage Cons: No cloud type information Short record (1980's to present)

10 Surface Observations Trained observer reports cloud cover and types, among other weather data Observation is transcribed and sent in by radio for use in synoptic charts Report is then archived as 'COADS'

11 Surface Cloud Observations Taken aboard ships, typically every 6 hours Contain meteorological data such as temperature, dewpoint, pressure, wind, and present weather

12 Surface Cloud Observations Cloud data includes: Sky cover, both total and lowest deck in eighths ( oktas ) Cloud type at 3 levels low, middle, and high From a total of 27 possible cloud types Estimated cloud base height

13 Observations Per Year (Millions)

14 Observations Per Year (Hundreds)

15 Cloud Data Averaging Observations are averaged seasonally on the grid shown on the previous slide 27 observed cloud types are pared down to 9: 20 observations per season are required 5 low clouds: fog, Stratus, Stratocumulus, Cumulus, Cumulonimbus 3 middle types: Altocumulus, Altostratus, Nimbostratus 1 high (cirriform) cloud Averages formed for day, night, and day+night

16 Using These Data... Averages are already formed & published, so we will focus on interannual variability Look at time series on a global scale Identify problem with the data Attempt to correct the problem Use corrected data to confirm previous thinking on how low clouds and SST interact Look at local trends in cloud types to see how cloud changes may be related to climate change

17 Producing an Unbiased, Global Time Series Seasonal cloud amounts must be converted to anomalies for each box Anomaly: Long-term seasonal mean subtracted from yearly value in the time series Yearly anomalies are then averaged, weighted by the relative area and % ocean for each grid box Eliminates bias associated with different sized grid boxes, and different climatological means

18 Producing an Unbiased Time Series

19 Producing an Unbiased Time Series

20 Zonal (Red) and Global (Blue) Time Series

21 Global & Zonal Time Series - Cumulonimbus Individual zones tend to vary coherently First seen by Bajuk & Leovy in 1998 No physical explanation exists for this, Likely a spurious variation in the data

22 Global & Zonal Time Series - Cumulonimbus Joel Norris suggests possibilities: Changing nationalities making observations over time Trying to force this failed to produce similar results Observation procedures changing over time None are seen in the data, no 'trade-offs' between types

23 Testing Observed Variations Prove these variations to be spurious by comparing ship data to island data Island data only spans (currently) Compare high and low frequency variation for numerous cloud types Low frequency variation 5th degree polynomial High frequency variation residual after polynomial is removed Over central Pacific Ocean Region with lots of islands and consistent ship traffic

24 Ocean & Island Comparison

25 Ocean & Island Comparison Island data is averaged identically to ship observations Observations Grid Boxes Anomalies Time Series 5th degree polynomial is then fit to each time series Residuals are calculated, then correlated

26 Ocean & Island Comparison Low Frequency Visual comparison of polynomial fits shows little agreement Gives no reason to believe low frequency variations are real Provided residuals show agreement, we can determine that low frequency variation is spurious

27 Ocean & Island Comparison High Frequency Correlation of residuals shows agreement at high frequency All correlation coefficients are positive and significant (99%), except Cumulus (Cu) Expect Cu to correlate less, due to smaller scale motions & processes

28 Ocean & Island Comparison Low frequency variation shows little agreement High frequency variation does show agreement It is not likely that long-term variability would be completely unrelated, given the correlations on the year-year time scale Land data has not shown a tendency for different latitude bands to vary coherently Ocean low-frequency variations are spurious, and should be removed from the data before attempting to study cloud cover changes

29 Removing Spurious Variations Use 10th degree polynomial to approximate coherent global variation in 55 year data set Captures variations nearly equal to that shown in island comparison Subtract scaled polynomial from each grid box

30 Using Adjusted Data Correlate cloud cover variations with sea surface temperature (SST) for all seasons Cloud types & SST Location of strong correlation Correlate only high frequency (year-year) variation Remove 10th degree polynomial curve fit to data in each box Investigate long-term changes in cloud cover Regional changes instead global Assume a null hypothesis of little global change Remove the global 10th degree polynomial curve Emphasis on cloud types that correlate well with SST

31 Total Cloud Cover & SST

32 Low Cloud Cover & SST

33 Stratiform Clouds & SST

34 Cumulus & SST

35 Clouds & SST Negative Correlations Negative correlations in eastern subtropical oceans Low stratiform cloud cover is responsible Cumulus clouds show inverse relationship Reinforces previous thinking about SST and stratiform cumuliform clouds Warmer SST related to broken clouds

36 High Cloud Cover & SST

37 Clouds & SST Positive Correlations Positive Correlations in Central Pacific High and low clouds are responsible All types but Cumulus correlate positively Possibly SST driven convection causing more clouds when SST is warmer? Opposite to previous case

38 Correlation with ENSO Total Cloud Cover

39 Correlation with ENSO Total Cloud Cover

40 Correlation with ENSO Stratiform Clouds

41 Correlation with ENSO Cumulus Clouds

42 Correlations with ENSO ENSO correlations show strong seasonality, much more than SST-cloud relationships alone Much stronger signal in DJF than JJA SON trade-off between Cumulus and Stratiform clouds in the Eastern Pacific Ocean ENSO may be responsible for the strong positive correlation in the Central Pacific Appears to persist through seasons Agrees with SST driven convection idea ENSO index shows a slight increase over this time period ~ / Decade

43 Trends of Total Cloud Cover 25 observations are required per season Data must span at least 30 years, with no fewer than 20 individual years Global polynomial subtracted from time series

44 Trends of Total Cloud Cover Computed using the median of pairwise slopes method Shown if a trend is greater than its uncertainty, or if the uncertainty is less than 2% per Decade

45 Trends of Total Cloud Cover (0.1 % / Decade)

46 Trends of Low Clouds (0.1 % / Decade

47 Trends of Stratiform Clouds (0.1 % / Decade)

48 Trends of Cumulus Clouds (0.1 % / Decade)

49 Trends of High Clouds (0.1 % / Decade)

50 Conclusions Spurious variation is present in time series of cloud types 10-degree latitude zones vary coherently This variation must be removed for the study of local trends Cloud types and SST correlate in certain parts of the ocean In eastern oceans Stratus/stratocumulus decks are present and persistent SST and low cloud cover are inversely correlated In the central pacific Positive correlation suggests SST driven convection

51 Conclusions Pronounced regional changes in cloud cover are occurring Decreasing low stratiform clouds in eastern oceans, in regions where low stratiform clouds persist Increasing clouds in central Pacific, where SST drives convection and cloud formation Some transition to cumulus, but overall a reduction in low cloud cover May represent a positive cloud feedback SST trend? Correlation likely reflects ENSO, but ENSO trend is not large High cloud trends could suggest jet stream migration

52 Thank You Steve Warren, Carole Hahn Coauthors Robert Wood, Joel Norris, Louise Leahy Project Feedback