short-term aerosol trends: reality or myth?€¦ · trend computation artifact? • need a better...
TRANSCRIPT
SHORT-TERM AEROSOL TRENDS: REALITY OR MYTH?
Gregory Leptoukh1 and Viktor Zubko2,1
1NASA Goddard Space Flight Center2Wyle IS
https://ntrs.nasa.gov/search.jsp?R=20100003357 2020-07-12T20:26:58+00:00Z
Gregory Leptoukh, IGARSS '09 2
Spatial Distribution of AOT (Terra) Annual Anomaly Trend (Time range Sep 2002 – Aug 2008)
Annual Anomaly: Deviation of monthly AOT time-series in each 1 x 1 grid (72 points for 6 years) from the AOT averaged over 6 years
Least squares linear trend: slopes (per year) for each grid box 7/17/2009
Gregory Leptoukh, IGARSS '09 3
Spatial Distribution of AOT (Terra) Anomaly Percentage Change per Year
(Time range Sep 2002 – Aug 2008)
7/17/2009
Main questions addressedShort-term Trends of MODIS AOT over six years
– Why are the trends different in different regions?
– How these trends be so high?
– Why are they “coherent” in many areas?
– Are these changes in aerosol concentrations real, i.e., are they monotonic changes in emissions?
7/17/2009 4Gregory Leptoukh, IGARSS '09
Possible alternative explanations
• Trend computation artifact?
• Need a better deseasonalization?
• Related to changes in meteorology patterns?
7/17/2009 Gregory Leptoukh, IGARSS '09 5
MODIS Terra AOT Anomaly Trend (Least Squares)
7/17/2009 Gregory Leptoukh, IGARSS '09 6
2003 Siberian Fires and “negative” AOT trend
Outlier
Negative LS trend
The regular Least Squares (LS) linear trend is simple but sensitive to outliersAn alternative (Sen) linear trend is less sensitive to outliers
7/17/2009 7Gregory Leptoukh, IGARSS '09
Eastern China: real positive trend
7/17/2009 8Gregory Leptoukh, IGARSS '09
MODIS Terra AOT Anomaly Trend Alternative (Sen) computation
7/17/2009 Gregory Leptoukh, IGARSS '09 9
An alternative linear trend based on Kendall-Sen non-parametric statistics is more robust and less sensitive to outliers
60
40
20 "
o
-20
-40
Orig. Time Series: B(Sen), Av~era~ge~A~O~T~~X~~~
-60 ~~~~.,~~~~~~~~~~~,-~~~~~~~~
-LSO -LOO -50 o 50 LOO LSO
MODIS Terra AOT Anomaly Trends:Difference between LS and Alternative (Sen)
7/17/2009 Gregory Leptoukh, IGARSS '09 10
SenLS
Difference
Characteristics of outlier hot and cold spots
7/17/2009 Gregory Leptoukh, IGARSS '09 11
PDFs for four case study regions (Bezier approximations are shown in black)
MODIS Terra AOT Percentage of change per year
Sen shrinks outliers’ PDFs and pushes them closer to zero
SenLS
SenLS
Possible alternative explanations
• Trend computation artifact?
• Need a better deseasonalization?
• Related to changes in meteorology patterns?
7/17/2009 Gregory Leptoukh, IGARSS '09 12
MODIS AOT Annual Anomaly Linear Trends
7/17/2009 Gregory Leptoukh, IGARSS '09 13Terra
Sen Sen
LS LS
Difference Difference
Sen Sen
LS LS
Difference Difference
Aqua
Deseasonalized AOT Monthly Anomaly Linear Trends
7/17/2009 Gregory Leptoukh, IGARSS '09 14AquaTerra
Sen Sen
LS LS
Difference Difference
Possible alternative explanations of AOT short-term changes in some areas • Trend computation artifact?
• May disappear after deseasonalization?
• Related to changes in meteorology patterns?
7/17/2009 Gregory Leptoukh, IGARSS '09 15
Similarities in MODIS Cloud Properties Anomaly trends with the Coincident 52-Months of CERES and AIRS-V5 OLR
From Joel Susskind, IGARSS’08
?
7/17/2009 16Gregory Leptoukh, IGARSS '09
AIRS OLR CERES OLR
MODIS Cloud Fraction MODIS Cloud Optical Thickness
SST
AOT
Cloud Fraction
Western Pacific Warm Pool displacement (5S-5N, 160E-160W)
7/17/2009 17Gregory Leptoukh, IGARSS '09
MODIS AOT Standardized Anomaly Trend (LS)
Standardized or normalized anomaly: divided by anomaly std
60 J?Wj~
40 '
""' 20 o '-'
~ ° a .~
'oi ...:i -20
-60 -150 -100 -50 0 50 100 150 -150
Longitude e) -100
0.2
0.15
0, I
0,05
° -0,05
-0, I
-0,15
-0,2
Western Pacific Warm Pool contraction cloud pattern displacement
7/17/2009 19Gregory Leptoukh, IGARSS '09
MODIS Cloud Fraction Standardized Anomaly TrendWestern Pacific Warm Pool case
60 ~~
40 ''''-'''0
o '-"
.g 0 B . ~ ... Ol
....l -20
-40
-150 -100 -50 0 50 100 150 Longitude (0)
-150 -100
0.2
0.15
0.1
0.05
o -0.05
-0.1
-0.15
-0.2
7/17/2009 Gregory Leptoukh, IGARSS '09 20
0
Displacement leads to artificial trend
Negative “trend”
Positive “trend”
This explains spatial coherence, similarity in shape of adjacent areas with opposite trends, and magnitude of trend
NAO and SO cause precipitation trends
7/17/2009 Gregory Leptoukh, IGARSS '09 21
From: Kyte, E.A., G.D. Quartly, M.A. Srokosz and M.N. Tsimplis, 2006, 'Interannual variations in precipitation: The effect of the North Atlantic and Southern Oscillations as seen in a satellite precipitation data set and in models', J. Geophys. Res., 111, art. no. D24113
a) 80~~~-'C-~~~--~ "\ "''''~~~~-.~'-----~r-~"
70
60
50
40
60
50
40
30~~~:=~~~~~~~~~ C) 80; 70
60
50
40
- 20 - IS - 10 - S 0 5 10 15 20 - 20 - IS - 10 -s 0 S 10 15 20
mm month-' / NAO unit mm month-' / SO unit
Conclusions (trends)• There is a broad spatial inhomogenueity in AOT trends
over 6 years of MODIS Terra and Aqua• Some of the areas demonstrate clear positive trends
related to increase of emission (e.g., Eastern China)• Strong trends in some other areas are superficial and
might be attributed, in part, to:– Least squares linear trend sensitivity to outliers (need to use
more robust linear fitting method)– Spatial and temporal shifts or trends in meteorological
conditions, especially in wind patterns responsible for aerosol transport
• Aerosol trends should be studied together with changes in meteorology patterns as they might closely linked together
7/17/2009 22Gregory Leptoukh, IGARSS '09