Improving GPS RO Stratospheric Retrieval for Climate Benchmarking Chi O. Ao 1, Anthony J. Mannucci 1, E. Robert Kursinski 2 1 Jet Propulsion Laboratory, Caltech, Pasadena, CA, USA 2 Broad Reach Engineering, Golden, CO, USA Submitted to Geophys. Res. Lett. (March 2012) April 10-12, 2012 CLARREO SDT Meeting, Hampton, VA © 2012 California Institute of Technology. Government sponsorship acknowledged.

Introduction GPS RO refractivity (N) is a key component of CLARREO and has been shown to be useful for climate monitoring [Leroy et al., 2006; Huang et al. 2010]. Accuracy is well proven in the “sweet spot” of 5-20 km alt (CLARREO baseline), but considerable uncertainty in the lower troposphere ( 20 km) exists. For climate, we are interested in the mean N averaged over many profiles ( ). This motivated us to develop a new method that reduces the stratospheric bias for. 2

3 ROTrend Project: Intercomparison of 7-yr CHAMP RO refractivity among multiple processing centers [Ho et al., JGR, 2009] Disagreement over 25 km is largely due to different ways of dealing with noisy bending angles at high altitudes dT/T ~ dN/N 0.2% dN/N ~ 0.4 K

4 Phase delay time series (L1, L2) Bending angle profiles (L1, L2) Iono-corrected bending angle profile Refractivity profile Abel inversion ionosphere atmosphere

5 ~ 2 km vertical smoothing stdev km Iono-corrected bending angle from COSMIC Jan 2008 median 10 %ile 90 %ile Systematic error starts to dominate at ~ 80 km Mean

6 Abel Upper Boundary Condition: High altitude noisy bending angles are replaced by a “model”. Refractive Index = 1+N Impact parameter a = n(r) r Bending angle Blending of obs and model (Statistical Optimization): Fixed height transition from obs to model: Bending angle errors from high altitudes will propagate downward through the Abel integral:

Which “Model”? MSIS Climatology NCAR Climatology (exponentially extrapolated) Climatology model with bias adjustment Our preference for climate dataset: – Exponential extrapolation based on measurements: No dependence on climatology models. – Fixed height transition: “model” influence more transparent. All “models” will introduce systematic bias at lower altitudes if the max obs height (h m ) is too low. 7

To reduce the bias, the cutoff height h m should be set as high as possible. 8 Truncation Exp Extrap

Strategies for Reducing Random Noise Better antenna gain (increase SNR); Less calibrating links (zero differencing); More vertical smoothing (decrease vertical resolution); Averaging over large number of bending angle profiles (e.g., monthly zonal means) 9 Can we compute the mean refractivity profile by Abel inversion of the mean bending angle profile?

10 In general, they are not the same. However, for n ~ 1, the difference can be neglected. The neglected higher order term has relative error ~ 0.5 N 10 -6, which at 10 km (N~100), is only 0.005%. ?

11 Simulations using MSIS profiles with COSMIC sampling for one zonal band (45-50 deg latitude, Jan 2008, ~ 2000 profiles). Random Gaussian noise of 1 μ-rad sigma and zero mean were added. Averaged bending angle ~ 1/50 reduction in noise

12 Averaged bending angle with exponential extrapolation at different h m Significantly lower bending angle error with hm=70 & 80 km

13 AvgBend: from inversion of averaged bending angle AvgRef: from averaging individually retrieved refractivity profiles AvgBend70 & AvgBend80 give much less biased results

14 Results from actual COSMIC data (Bending) Simulation

15 Results from actual COSMIC data (Refractivity) Simulation

Conclusions For climate averaging, stratospheric refractivity can be improved by averaging the bending angles first before Abel inversion. (It’s also more computationally efficient.) By reducing the random noise, the bending angle measurements are useful up to higher altitudes, leading to substantial improvement in the Abel upper boundary condition and strato refractivity (esp. above 30 km). 16

Ongoing Work Reduction of systematic error from residual ionospheric effect. – Phase - Range: minimize L1, L2 path splitting but range measurements are noisier. – Utilizing the third GPS frequency (L5) that will be available from next generation GPS RO: remove 1/freq^4 ionospheric term at the expense of increasing noise. Comparison of ~ 10 yrs GPS RO measurements (geopotential height, tropopause) with CMIP5. 17

TriG Receiver Development at JPL 18 TriG continues on a development path to meet the technical requirements for CLARREO. TriG is scheduled to fly on COSMIC-2 in 2016 and is proposed for several other NASA missions including ICESat-2 and GRACE-FO. Courtesy of Jeff Tien and Tom Meehan