Transcript Document

(Towards) A New AVHRR Cloud Climatology

Andrew Heidinger, Mitch Goldberg, Dan Tarpley

NOAA/NESDIS Office of Research and Applications

Michael Pavolonis

UW/CIMSS, Madison, Wisconsin

Aleksandar Jelenak

IMSG, Inc This work is funded by ORA and leveraged heavily off IPO and OSD funding

SPIE Asia-Pacific Symposium, 8-11 November 2004, Honolulu Hawaii

Outline

• ORA’s AVHRR Reprocessing Project •The AVHRR Cloud Climatology (Pathfinder Atmospheres Extended Project – PATMOS-x) •Comparison of PATMOS-x Cloud Climatology with other climatologies (ISCCP, NOAA + UW HIRS) •Comparison with new observing systems (GLAS LIDAR data) •Conclusions •Future Work

Why use the The AVHRR for Climate Studies The Advanced Very High Resolution Radiometer (AVHRR) was launched in the 1979 for non-quantitative cloud imagery and SST. It flies on the NOAA Polar Orbiting Satellites (POES) 1. AVHRR Provides enough spectral information for several applications 2. AVHRR provides enough spatial resolution (1 or 4 km) to resolve many atmospheric and surface features 3. Combined with its long data record (1982-2012) make the AVHRR data set appealing for decadal climate studies

AVHRR Data Stewardship Project

For these reasons, the NOAA/NESDIS Office of Research and Applications has embarked a pilot Data Stewardship Initiative Project to reprocess the entire AVHRR GAC data set (1982-2004)

The activities within this project include:

•Improving the level 1b data (navigation/calibration) •Reprocessing for a new AVHRR SST climatology •Regeneration of Global Vegetation Index Climatology (GVI) at higher spatial resolution •Polar Winds •Regeneration of a new PATMOS-x Cloud Climatology

AVHRR Data Improvement Activities

Due to clock errors and other effects, the AVHRR GAC navigation can be off by many kilometers. We have been provided the clock corrections from the University of Miami and are re-navigating the data.

These corrections will be made available to the public as ancillary data before after Note, this is a worst-case example

The AVHRR Pathfinder Atmospheres Extended (PATMOS-x) The PATMOS-x Cloud Climatology is our attempt to provide a new cloud climatology to the satellite climate research community. This builds upon our previous effort (PATMOS) completed in 1998.

PATMOS-x is an extension of the original PATMOS into • the NOAA-klm AVHRR/3 series (1998-2004) • AVHRR from morning orbits (higher temporal resolution) • higher spatial resolution (0.5

o or 0.25

o versus 1.0

o ) • many more products with new algorithms and data files in HDF (

with metadata

) PATMOS-x also differs from PATMOS in that we are now able to reprocess from the level 1b data multiple times. PATMOS allowed only one pass through the data.

Reasons for Deriving a New Cloud Climatology from AVHRR •Uncertainties in climate predictions often attributed to cloud uncertainties.

• In some ways, the existing cloud climatologies show little consensus (i.e. The trend and magnitude of the July Total Cloud Amount time series) • • PATMOS-x will provide another data-set that is mostly independent. PATMOS-x will provide more information to better understand observed cloud amount climatologies.

• PATMOS-x is being developed to be

physically consistent

and NPOESS/VIIRS with MODIS – this is critical for time series continuity beyond 2010.

PATMOS-x Cloud Products

• Pixel level cloud detection and cloud type* • Pixel level cloud optical depth, particle size, lwp, iwp • Pixel level cloud top temperature and emissivity* • Gridded means and standard deviations of all pixel level products • Cloud amounts: total, high, mid, low, ice, water, multilayer • Mean and standard deviations of the radiances for each cloud mask category •

Products are similar to those from MODIS and VIIRS, lack of spectral information does hamper AVHRR’s performance in stressing conditions such as terminator conditions and polar regions. For most regions for most clouds, AVHRR can do well for many important climate data records.

* - our primary focus for algorithm validation

Examples of PATMOS-x Cloud Products Cloud Detection PATMOS-x uses the CLAVR-x cloud detection algorithm which serves as NOAA’s operational AVHRR cloud mask. CLAVR-x is well suited for climate work because it requires little ancillary data and is stable over time. CLAVR-x has been designed to minimize day/night/terminator artifacts. Spatial uniformity plays a big role.

Cloud mask similar to MODIS: clear, prob. clear, prob. cloudy, cloudy

Examples of PATMOS-x Cloud Climate Records (Cloud Type) Each cloudy pixel is classified into one of following cloud types: (0) clear, (1) fog, (2) water , (3) super-cooled water (4) opaque ice, (5) cirrus, (6) multilayer cirrus Validated against ARM cloud type products (Tanneil Uttal/NOAA) and a version developed for VIIRS with IPO support Knowledge of Multilayer Cloud may help resolve discrepancies between climatologies which do not contain this information (ISCCP and HIRS)

Cloud Temperature/Emissivity • PATMOS-x uses a split-window optimal estimation approach to estimate both cloud temperature and emissivity. Pure IR approach helps achieve consistent performance for all orbits/solar geometries •Results being validated against MODIS. Goal is to achieve

consistency with MODIS and VIIRS

for thin cirrus – difficult from AVHRR.

MODIS (MOD06) cloud temp.

AVHRR cloud temp AVHRR cloud emissivity

Comparison with Other Cloud Climatologies • Two existing global satellite derived climatologies are: 1.

International Satellite Cloud Climatology Project 2.

(ISCCP) (GEO imagers+AVHRR) University of Wisconsin / NOAA HIRS • We can learn a lot about PATMOS-x from these comparisons. For example, we can test the inter-annual variability, annual and diurnal cycles of PATMOS-x compared to the others.

• Though the MODIS time series is short, it offers superior performance in some situations that can provide point out PATMOS-x deficiencies and guide future development.

• New active cloud observing systems (GLAS, CloudSat, CALIPSO) offer direct validation of many PATMOS-x parameters and allow us to characterize what clouds under what conditions we are missing.

Comparison of Total Cloud Amount Trends

The figure shows Total Cloud Amount time series from 60S to 60N for July •PATMOS-x does not exhibit the downward trend scene in ISCCP •

Differences in magnitude are likely due to PATMOS-x weighting of partly cloudy pixels. ISCCP and HIRS do no weighting of partly cloudy pixels

PATMOS-x trends are preliminary until calibration work is finished

Spatial Pattern in the Total Cloud Amount Trends

The following images show the linear trend from the time series based on the July High cloud amounts (roughly 20 years) ISCCP PATMOS-x HIRS •More agreement in patterns in PATMOS-x and ISCCP than HIRS •Agreement in the positive trends in Philippine and Arabian Seas •Agreement in the negative trends in the Caribbean Sea •HIRS show more wide spread distribution of positive trends (consistent with trend in 20S to 20N result)

July High Cloud Amount Trends in Tropics

• HIRS High Cloud Amounts are much greater than PATMOS-x and ISCCP •HIRS shows a slight positive trend while PATMOS-x shows little trend and ISCCP shows a trend opposite to HIRS.

July High Cloud Amount Trends in Tropics (Continued)

• If we compare the time series of the ratio of the High to Total Cloud Amount, HIRS and PATMOS-x are very similar •This ratioing may remove effects in cloud amount due to partly cloud pixels

Correlation of the PATMOS-x Cloud Climatology with ENSO Our goal in this analysis to verify that PATMOS-x cloud climatology behaves as expected where expected under ENSO The following images show to mean July High Cloud Amount and its Anomaly correlation with ENSO

Positive correlation of Tropical Pacific Cloud is expected as is the distribution over North America. Similar correlation seen in Total Cloud Amount. These anomalies probably explain spatial pattern of trends .

Comparison of the Annual Cycle in High Cloud

Other checks of the PATMOS-x cloud climatologies are if they reproduce the expected diurnal and annual cycles • All show the same annual cycle • ISCCP shows a very large diurnal cycle that is likely an algorithm artifact •PATMOS-x shows little diurnal cycle in tropical high cloud. •PATMOS-x does show more cloud over land during day

data provided by H. Zhang and D. Wylie

Using New Cloud Observing Systems (GLAS) to estimate optical depth sensitivity of cloud climatologies By filtering out GLAS results with optical depths below some minimum, we can estimate the sensitivity of our passive cloud climatologies: Minimum GLAS optical depth to match observed High Cloud Amount :

AVHRR Day – 0.23

AVHRR Night – 0.1

MODIS/TERRA – 0.12

ISCCP Day – 0.27

ISCCP Night – 0.40

HIRS Day/Night – 0.04

AVHRR can identify thinner high cloud at night than day – opposite of ISCCP. If we can’t get reduce day/night differences, we want to characterize them.

Conclusions

• NOAA/NESDIS ORA is working on improving the AVHRR GAC data record (

Please contact us is you want to participate)

•PATMOS-x will use this improved data and provide a new global cloud climatology (along with radiances and selected non-cloud products) •PATMOS-x cloud amounts show many similarities in spatial distribution but not much consensus in trends with ISCCP and HIRS •PATMOS-x annual cycles appear consistent with MODIS/HIRS/ISCCP •GLAS observations are useful in estimating the cloud missed by passive sensors (derive errors bars for passive cloud climatologies) •PATMOS-x cloud amount shows some expected variations with ENSO.

Future Work • Finish navigation and calibration improvements to AVHRR GAC •Validate remaining cloud algorithms (Cloud Temperature, Cloud LWP) •Make preliminary data-set available for testing •Finish intercomparison with other climatologies •Extend GLAS analysis to CALIPSO and CloudSat •Continue to explore AVHRR/EOS/NPOESS climate continuity

We welcome any feedback and seek collaboration with others [email protected]

Goals for PATMOS-x

• Provide a low resolution radiance data-set that allows the climate community access to ORA’s Level 1b improvements.

• Provide critical climate records derived from accepted algorithms, on a consistent grid in an easy to read format (HDF). We focus on products that we (ORA) have experience with (NDVI,aerosol, SST, OLR, Clouds) •Serve as a new cloud climatology with information that compliments and builds upon the existing climatologies (ISCCP and UW/HIRS).

Using Cloud New Observing Systems to Validate Cloud Climatologies

NASA’s GLAS LIDAR optical depth / cloud layer data has been made publicly available for one month (Oct/Nov 2003). LIDAR provides a direct measurement of the presence of cloud and the vertical profile for thin clouds.

In comparison to PATMOS-x, GLAS detects many more clouds.

Previous Cloud Climatologies • International Satellite Cloud Climatology Project (ISCCP) •NOAA + University of Wisconsin HIRS (UWHIRS) •Original AVHRR Pathfinder Atmospheres (PATMOS) •Cloud climatologies exist for most missions (NIMBUS-7, SAGE, EOS) •Other AVHRR climatologies includes several region studies (APP-x – Polar Regions, SCANDIA - European)

Cloud Amounts

• Using the Pixel level cloud mask, cloud type and cloud temperature, the following cloud amounts are produced: Total, Low, Mid, High, Ice and Water •Being compared to ISCCP and UW/HIRS •NCEP comparing them against other real-time products (USAF)

Total Cloud Water Cloud High Cloud (P < 440 Hpa)

Spatial Pattern in the High Cloud Amount Trends

The following images show the linear trend from the time series based on the July High cloud amounts (roughly 20 years) ISCCP PATMOS-x HIRS • Agreement in the positive trends in Philippine and Arabian Seas •HIRS show more wide distribution of positive trends (consistent with trends in 20S to 20N result)

Using GLAS to estimate sensitivity of the AVHRR cloud detection By filtering out clouds with optical depths lower than some threshold, we can determine the lower limit of the optical depth of clouds observed in the AVHRR/PATMOS-x and other cloud climatologies Using this analysis applied to the

Total Cloud Amounts: AVHRR = 0.33

MODIS/TERRA = 0.05

ISCCP = 0.08 HIRS = 0.015

This does not take into account the differences in cloud amount that are not due to cloud detection (partly cloudy pixels)