1. Radar Remote Sensing Microphysical Processes? Feasibility using In Situ Data from SEAC4RS and TC4 Jay Mace Measurement Provided by: Simone Tanelli, Paul Lawson and Co., Svetla Hristova-Veleva, Steve Durden, Paul Bui Convective Turret Penetrated during SEAC4RS by the DC8

2. Past (passive): Grossly characterize the bulk properties of profiles ? ? ? Present (A-Train): Characterize the basic profile of microphysics Future (ACE?): Characterize the processes that drive changes to particles in the column Evolution of satellite Measurement Strategy

3. By Processes we mean the conversion of one hydrometeor species to another through collection… Autoconversion – growth of cloud mode droplets to precipitation size Accretion – collection of cloud drops by falling precipitation Consider a parameterization of processes by Khairoutdinov and Kogan (2000) The terms in the collection equation are either measured directly N(D) or can be inferred (V). So processes can be estimated from the in situ measurements using numerical solutions of the double integral. Rico Clouds (Rauber et al, 2008) Note: No time derivative on RHS.

4. We can conceive of two approaches whereby remote sensing can “observe” process… 1.Take two measurements of the same volume separated in time and observe changes to the measureables from which process can be inferred (i.e. a train of small satellites) 2.Use the collection kernel and a broad suite of measurements to diagnose (retrieve) the dominant process instantaneously. Fundamentally, we need to know something quantitative about the cloud mode and precip particle modes simultaneously. Are Either of these approaches viable?

5. Method We manipulate the collection kernel to provide the time rate of change of Z due to the collection process. Then we can estimate how far apart two radar measurements would need to be to capture some measureable change. We simulate dual frequency Doppler radars like those that may fly in space to determine the degree to which the simulated measurements are sensitive to the inferred collection process. Is there information in the measurements about process?

6. 9/4 13:53-54 Temp ~ -10C Can Process be inferred by watching radar measureables change? APR2 Doppler Velocity 9/4 13:54 UTC DC8

7. Temp: 263 K. Rh liquid : 95% W: 5 m/s Z Ka: : -3 db Z w :-5 db Precip: 0.8 mm/hr Cloud Mass: 0.7 g/m3 Riming Rate: 1.1 g/m 2 /km/hr Time to 3 db change in Ku Z: 62 Seconds in this convective updraft. Meanwhile: parcel has ascended 300 m and advected relative to the rotating Earth ~2 km (recall slanted updraft) Can Process be inferred by watching radar measureables change?

8. Can Process be inferred from some combination of radar measureables? We will examine Information Content as a function of Instrument Noise and Measurement Error, and the sensitivity (dZ/d(parameter)) using PSD’d directly measured during TC4 and SEAC4RS Consider a Retrieval problem posed as follows: 4x4 Matrix of Sensitivities Prior and S a derived from in situ statistics – TC4 and SEAC4RS

9. Can Process be inferred from some combination of radar measureables? Red histogram shows the fraction of volumes that have non zero information Black shows all volumes Retrieval of accretion rates seems to be within reach for a fairly large fraction of volumes (40%) for this particular retrieval combination… Caveat: Ice-Ice single mode PSD’s are predominant.

10. How would temporal difference approach compare to direct retrieval? Typical time to 3db change is ~200 minutes.

11. Summary: Explored the direct retrieval of process rates in comparison to temporal differencing approach from a train of small satellites. Data used was from TC4 and SEAC4RS Temporal differencing is feasible for strong convective cores. Otherwise, probably not. Dual Frequency Ka-W measurements provide information for retrieval of ice crystal properties, precipitation rates, and accretion rates ~40% of the time in the mostly stratiform anvils during TC4. Interestingly, we are most sensitive to weakest accretion rates because the fractional change in Z and delta Z is largest.

12. “ If you look deeply, you can see the clouds in the rain (& snow)… ” Thich Naht Hanh, Zen Buddhist Monk Photo by Christy Wall from Storm Peak Lab

13. Sensitivity to Graupel/Hail Parameterization: The peak stratiform and convective areas differed by 105% and 150% respectively Accumulated precipitation varied by a 558% (Adams et al., 2013) The Representation of microphysical processes in models is proving to be THE limiting factor in high resolution simulations. Will become a critical issue as models move to global cloud resolving resolution in the next decade. Sensitivity to Riming of ice in Mixed Phase: Surface snowfall rates and totals vary by 200 – 300% due to differences between bin and bulk microphysical riming schemes (Saleeby and Cotton, 2008) Sensitivity to Droplet Breakup in Rain: Small changes to droplet breakup parameters => 500-600% differences in precipitation rates (Morrison et al., 2012) Sensitivity to Microphysical Scheme Complexity (# of moments): 300-400% differences in surface precipitation due to the number of moments predicted => feedbacks to storm dynamics

14. Theory The collection Eqn: Represents the time change of precipitation mass per unit mass of air due to collection of cloud- mode (liquid or ice) We want to know the time change in radar measureable due to collection of cloud-mode hydrometeors. If we multiply the outer integral by, then… We can quantify a time rate of change of radar reflectivity due to collection of cloud mode. Similarly, for Doppler Velocity. Solve Numerically for each measured PSD fitted by gamma functions. Using T-Matrix and Mie theory, we can explore sensitivity to multiple frequencies, sensitivity to differential Doppler, etc

15. Basic Assumption: Everything is uncertain. PSD’s are bimodal – cloud and precip coexist Measurements: Two Frequency Doppler Radar (Ka-W) Can Process be inferred from some combination of radar measureables? OE We will examine Information Content (H) as a function of Instrument Noise and Forward Model Error (Sy), the terms of the Jacobian (dZ/d(parameter)) using PSD’d directly measured during TC4 and SEAC4RS Consider a Retrieval problem posed as follows: 4x4 Matrix of Sensitivities Prior and S a derived from in situ statistics – TC4 and SEAC4RS

16. Sensitivity of Z and dZ (in db) to characteristic changes in geophysical quantities (Ka – W bands)…. Derived from TC4 Data 1% to 10% changes in Z and dZ due to Accretion (small but..) Factor of >2 changes in Z due to M-D params

17. Sensitivity of V d and dV d to characteristic changes in geophysical quantities (Ka – W bands)…. Derived from TC4 Data Meter per second sensitivity of Vd to M-D params centimeter per second sensitivity of delta Vd to M-D params