GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center HIRS Upper Tropospheric Humidity and Humidity Profiles Lei Shi NOAA National Climatic Data Center Asheville, NC, U.S.A.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Outline of Upper Tropospheric Humidity Motivation for intersatellite calibration –Time series discontinuity from satellite to satellite, particularly from HIRS/2 to HIRS/3 –Upper tropospheric water vapor (UTWV) is an important fundamental climate data record (CDR) –UTWV is a key component to water vapor feedback Approach –Intersatellite calibration based on overlaps of zonal means Result to achieve –Extended time series of the fundamental CDR to present

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Motivation – Uncorrected Intersatellite Differences of UTWV (Channel 12) Due to the independence of individual HIRS instrument’s calibration, biases exist from satellite to satellite. These intersatellite biases have become a common source of uncertainty faced by long-term studies. Start of HIRS/3 HIRS/2

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Spectral Filter Functions Differences between HIRS/2 and HIRS/3 are expected due to different filter functions. In-orbit performance still has biases unexplained by filter functions. Thus empirical approach is considered.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center More than half of satellites have bias variations larger than 0.5 K. Temperature-dependent Intersatellite Differences from Zonal Mean Approach

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Intersatellite Calibrated to N-12 (showing clear-sky 30S – 30N) Biases minimized. Temperature dependent biases accounted for. Similar overall variances between HIRS/2 and HIRS/3/4. Time series can be extended as variance preserved. PairsT_N- N06 N06- N07 N07- N08 N09- N10 N10- N11 N11- N12 N12- N14 N14- N15 N15- N16 N16- N17 N17- M02 Ave. Diff

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Climatologic Mean for Jan, Apr, Jul, and Oct JanuaryApril JulyOctober

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Climatological Monthly Mean and Variance There are two peaks of mean UTWV brightness temperatures in 30N-30S, one in winter (January) and another one in summer (June). During these two seasons the subtropics are dominated by a belt of strong subsidence (in northern hemisphere during winter and in southern hemisphere during summer). The lows of the UTWV brightness temperatures are found in April and November, indicating weaker descending branch of the general circulation during spring and fall. Large variances occur during the winter and early spring months.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center For an easier comparison of UTWV with other conventional observations, Soden and Bretherton [1993] derived a formula to calculate the upper tropospheric humidity (UTH) based on UTWV brightness temperature as UTH = cos(θ)exp( Tb)(1) In which Tb is the brightness temperature of HIRS channel 12 and θ is the satellite zenith angle. Upper Tropospheric Humidity

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Time Series of High and Low UTH There are usually two peaks of large area coverages in a year, one in summer and another one in winter. The area peaks of large UTH values and small UTH values usually occur in the same month, indicating when there are large organized convections in some parts of tropics, there are enhanced descending areas in other parts of the tropics.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center UTH values Trend (grids/yr) Std Dev (grids)Lag-1 Yrs Sig (yrs) Greater than 40% Less than 20% Trends of high and low UTH There is an increase of 2.6 grids/yr for the area with UTH values greater the 40%, and an increase of 2.2 grids/yr in the area of UTH with values less than 20%. The long-term area increases in both high and low UTH values reveal the likelihood of enhanced convective activities in the tropics.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center MJO ER Monitoring Tropical Waves

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Madden-Julian oscill.Kelvin wavesEq. Rossby waves Monitoring Tropical Waves

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Outline of Humidity Profile Inter-satellite calibration. Neural network scheme for deriving temperature and water vapor profiles. Comparisons with surface observations.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Channel Number Central Wavenumber (cm -1 ) Wavelength (micrometers) , , , , , , , , , (visible channel)14, High-Resolution Infrared Radiation Sounder (HIRS)/3 Spectral Characteristics.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Non-Inter-calibrated Clear-sky HIRS Channel Brightness Temperatures CH 2 CH 6 CH 8 CH 11

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Non-Inter-calibrated Clear-sky HIRS Channel-5 Brightness Temperatures

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Simultaneous Nadir Overpass (SNO)

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Inter-satellite Biases of 12 HIRS Longwave Channels

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Seasonal Variation of Inter-satellite Differences

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center CO 2 Increase From :

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center CO 2 Impact Averaged from Tb’s derived from 13,459 global profiles. The only variable changed in the simulation is CO 2 amount. For mid-tropospheric channels (channels 4-6), this means if there were no temperature change in the past 30 years, one would see decrease in the observed Tb’s. It is important to consider CO 2 impact in the retrieval!

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Statistics of Training Data TemperatureSpecific Humidity (K)(g/kg) Training data: A diverse sample of ECMWF profiles selected from the 1 st and 15 th of each month between January 1992 and December 1993 (Chevallier, 2001)

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Model Simulation RTTOV-9 is used to simulate HIRS channel brightness temperatures. –Advantage: model simulation of CO 2 effect. –Disadvantage: relationship can be affected by errors in pre-launch instrument measurements and model errors. However, the errors can be reduced by adjustment to observations (for example, adjusted to co- located HIRS and homogenized radiosonde samples). The selected profiles and corresponding HIRS brightness temperatures are randomly divided into three data sets: -60% training set -20% testing set -20% independent validation set

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Neural Network Separate neural networks for temperature and water vapor. Input: HIRS channels 2-12 and CO 2 concentration. Output: Tskin, Ta, and temperature profiles from 1000 to 50 hPa and water vapor profiles from 1000 to 300 hPa.

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Retrieval Root Mean Square (RMS) Errors

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Comparison with a Buoy Observation

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Comparisons of Retrievals with and without HIRS Inter-satellite Calibration

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Comparison with Drifting and Moored Buoy Observations DriftingMoored

GlobVapour Frascati, ItalyMarch 8-10, NOAA’s National Climatic Data Center Conclusions Upper Tropospheric Humidity based on almost-all-sky data –Temperature dependent inter-calibration –Extension of time series to current HIRS/2 and HIRS/3 series connected 30 years of global data Humidity Profiles based on clear-sky data –The HIRS data are intersatellite-calibrated using data from simultaneous nadir observations (SNOs). –CO 2 effect needs to be considered in the retrieval scheme. –A temperature and specific humidity retrieval scheme is developed based on neural network technique. –The retrievals are consistent with sea surface observations. Comparisons with land surface observations and with profile observations are being planned.