HIRS Observations of Clouds since 1978 Donald P. Wylie & W. Paul Menzel Cooperative Institute for Meteorological Satellite Studies NOAA/NESDIS University of Wisconsin-Madison Madison, Wisconsin,USA Darren Jackson Environmental Technology Laboratory, NOAA/OAR Boulder, Colorado, USA John Bates National Climate Data Center Asheville, North Caroline USA CO2 Slicing Method 22 year stats Effects of orbit drift, CO2 increase, and sensor changes 16 year trends Comparison with ISCCP and GLAS October 2004

Climate System Energy Balance

Rationale for Cloud Investigations clouds are a strong modulator of shortwave and longwave; their effect on global radiative processes is large (1% change in global cloud cover equivalent to about 4% change in CO2 concentration) accurate determination of global cloud cover has been elusive (semi transparent clouds often underestimated by 10%) global climate change models need accurate estimation of cloud cover, height, emissivity, thermodynamic state, particle size (high/low clouds give positive/negative feedback to greenhouse effect, and higher albedo from anthropogenic aerosols may be negative feedback) there is a need for consistent long term observation records to enable better characterization of weather and climate variability (ISSCP is a good start)

Why are clouds so tough? Aerosols 1000 km Cloud particles grow in seconds: climate is centuries Cloud growth can be explosive: 1 thunderstorm packs the energy of an H-bomb. Cloud properties can vary a factor of 1000 in hours. Few percent cloud changes drive climate sensitivity Best current climate models are 250 km scale Cloud updrafts are a 100 m to a few km.

Cirrus detection has been elusive in the visible bands Depending on view angle GOES sees or misses Texas cirrus

IR window sees cirrus but cannot place height correctly

Two unknowns, N  and P c, require two measurements Radiance from a partly cloudy FOV R=[1- N  ]R clear air + N  R opq cld (P c )

CO2 slicing corrects for semi-transparency of cirrus

RTE in Cloudy Conditions I λ = η I cd + (1 - η) I clr where cd = cloud, clr = clear, η = cloud fraction λ λ o I clr = B λ (T s )  λ (p s ) +  B λ (T(p)) d  λ. λ p s p c I cd = (1-ε λ ) B λ (T s )  λ (p s ) + (1-ε λ )  B λ (T(p)) d  λ λ p s o + ε λ B λ (T(p c ))  λ (p c ) +  B λ (T(p)) d  λ p c ε λ is emittance of cloud. First two terms are from below cloud, third term is cloud contribution, and fourth term is from above cloud. After rearranging p c dB λ I λ - I λ clr = ηε λ   (p) dp. p s dp

Cloud Properties from CO2 Slicing RTE for cloudy conditions indicates dependence of cloud forcing (observed minus clear sky radiance) on cloud amount (  ) and cloud top pressure (p c ) p c (I - I clr ) =    dB. p s Higher colder cloud or greater cloud amount produces greater cloud forcing; dense low cloud can be confused for high thin cloud. Two unknowns require two equations. p c can be inferred from radiance measurements in two spectral bands where cloud emissivity is the same.  is derived from the infrared window, once p c is known.

CO2 channels see to different levels in the atmosphere 14.2 um 13.9 um 13.6 um 13.3 um

Different ratios reveal cloud properties at different levels hi /13.9 mid /13.6 low /13.3 Meas Calc p c (I 1 -I 1 clr )  1   1 dB 1 p s = p c (I 2 -I 2 clr )  2   2 dB 2 p s

Determining Cloud Presence and Properties Detect clouds where (I - I clr ) > 1 mW/m2/ster/cm-1 in IRW or CO2 channels Use CO2 Slicing Method to estimate p c p c selected best satisfies RTE for all bands Estimate  IRW using IRW radiances If no CO2 bands qualify, IRW estimates opaque cld p c If too low in atmosphere, declare FOV clear

Ratio of measured cloud signal for spectrally close bands yields Pc

All Clouds Thin Clouds NE<0.5 Thick Clouds Opaque Clouds NE>0.95 Vis Optical Depth High (<400 hPa) 0.1< 33% <3 15% <6 15% >6 3% Mid (400  700 hPa) 18% 5% 7%6% Low (>700 hPa)24% -1%23% All Clouds75%20%23%32% UW NOAA Pathfinder HIRS global cloud statistics from December 1978 through December 2001

All Clouds Thin Clouds NE<0.5 Thick Clouds Opaque Clouds NE>0.95 Vis Optical Depth High (<400 hPa) 0.1< 33% <3 15% <6 15% >6 3% Mid (400  700 hPa) 26% 7% 10%9% Low (>700 hPa)49% -2%47% All Clouds75%20%23%32% UW NOAA Pathfinder HIRS global cloud statistics from December 1978 through December 2001 (corrected for higher cloud obstruction of lower clouds using random overlap assumption)

How Cloudy is the Earth? GLAS 22 Feb – 28 Mar 2003, HIRS 1979 – 2001, ISCCP 1983 – 2001, SAGE , Surface Reports , CLAVR ISCCP reports 7-15% less cloud than HIRS because it misses thin cirrus. HIRS and GLAS report nearly the same high cloud frequencies. HIRS reports more clouds over land than GLAS probably because GLAS sees holes in low cumulus below the resolution of HIRS. CLAVR 60

GLAS

All Cloud Observations from GLAS vs HIRS GLASHIRS

HIRS minus GLAS All Cloud Difference HIRS Frequency of All Clouds during the period of GLAS GLAS finds more tropical clouds over oceans where HIRS reports <40%. GLAS finds less clouds in polar regions and western tropical Pacific.

HIRS minus GLAS High Cloud Difference HIRS Frequency of High CloudHIRS – GLAS Difference GLAS > HIRS HIRS > GLAS HIRS reports more high clouds in parts of tropics and southern hemisphere, but areas of differences are scattered and not meteorologically organized.

Looking at animation of monthly means for 1997

HIRS-GLAS by latitude HIRS under detection is mainly over oceans.

Inferring Decadal HIRS Cloud Trends requires corrections for (1) anomalous satellite data or gaps (2) orbit drift (3) CO2 increase constant CO2 concentration was assumed in analysis

Satellite by satellite analysis Gap in 8am/pm orbit coverage between NOAA-8 and -10 HIRS cloud trends show unexplained dip with NOAA-7 in 2 am/pm orbit. Used only 2 am/pm orbit data after 1985 in cloud trend analysis for continuity of data and satellite to satellite consistency

morning (8 am LST)afternoon (2 pm LST) NOAA 6 HIRS/2NOAA 5 HIRS NOAA 8 HIRS/2NOAA 7 HIRS/2 NOAA 10 HIRS/2NOAA 9 HIRS/2 NOAA 12 HIRS/2NOAA 11 HIRS/2I * NOAA 14 HIRS/2I * HIRS/2I ch 10 at 12.5 um instead of prior HIRS/2 8.6 um. Asterisk indicates orbit drift from 14 UTC to 18 UTC over 5 years of operation Measurements from 9 sensors used in 22 year study of clouds Some sensors experienced significant orbit drift

all 2 am/pm satellites adjusted linearly to represent data for ascending node at 1400 hrs local time

(From Engelen et al., Geophys Res Lett, 2001) Atmospheric CO2 has not been constant

SARTA calculations: BT with 360 ppmv minus BT with 340,345,…380 ppmv

HIRS cloud trends have been calculated with CO 2 concentration assumed constant at 380 ppm. Lower CO 2 concentrations increase the atmospheric transmission, so radiation is detected from lower altitudes in the atmosphere. For January and June 2001 the clouds detected by NOAA 14 in the more transparent atmosphere (CO2 at 335 ppm) are found to be lower by hPa  dry(335,p,ch) =  dry(380,p,ch)**{335/380)  (p,ch) =  dry(p,ch)*  H2O(p,ch)*  O3(p,ch). More transparent atmosphere (CO2 at 335 ppm) results in HIRS reporting clouds lower by hPa with 2% less high clouds than in the more opaque atmosphere (CO2 at 380 ppm); this implies that the frequency of high cloud detection in the early 1980s should be adjusted down. Cloud time series was adjusted to represent a linear increase of CO2 from 335 ppm in 1979 to 375 ppm in 2001

Wielicki et al (2002) CERES deviation of reflected shortwave flux wrt mean for 20N-20S HIRS deviation of hi cloud detection wrt mean

Conclusions clouds were found in 75% of HIRS observations since 1978 (hi clouds in 33%) good agreement with GLAS ISCCP finds % fewer high and all clouds loop of monthly means shows latitudinal cloud cover follows the sun 16 yr trends in HIRS high cloud statistics reveal modest 2% increase during last decade compared with previous decade orbit drift, CO2 increase, and satellite to satellite differences were mitigated ISCCP shows decreasing trends in total cloud cover of 3 to 4 % per decade but little high cloud trend