Some refinements for global IOPs products Zhong Ping
Some refinements for global IOPs products Zhong. Ping Lee IOPs Workshop, Anchorage, AK, Oct 25, 2010.
Outline: 1. Two not-so-minor issues: a) Phase function of molecule vs particle scattering b) Angular variation 2. QAA vs Optimization
1 a). Phase function of molecule vs particle scattering Why a 0. 0012 m-1 bias for bbp? (Behrenfeld et al 2005)
Rrs model commonly used: G(0+) (sr-1) g or {g 0 & g 1} depends on phase function shape bb/(a+bb)
G or g is NOT a monotonic function of bb/(a+bb)! Facts: 1. G or g depends on phase function shape. 2. Molecular & particle scattering have significantly different phase function shapes. 3. We can have the same bb/(a+bb) for both molecular and particle scattering phase functions, but different G or g values. 4. For oceanic waters, blue band is dominated by molecular scattering; green/red bands by particle scattering. Conclusions: Need to account for the phase function shape (molecular vs particle) effects, especially for oceanic waters.
Explicitly separate the effects of molecule and particle phase function effects G from formula (sr-1) Lee et al (2004) G from Hydrolight (sr-1) Caveats: Complex function; cannot invert a&bb algebraically.
A practical model for algebraic/analytical inversion: gw&gp model g model (conventional, widely used)
Effects of separating molecular/particle phase function: MODIS Aqua
Comparison of results a 443 bbp 555 bbp is reduced by 40%; even more for adg 443 Consistent in general with the results (Werdell & Franz) that compare f/Q LUT vs Gordon 88 formula. aph 443 adg 443
Conclusion: Significantly different IOPs (in clear oceans) retrieved when the phase function shape effects are considered.
1 b). Angular variation θS θv ψ Ω(10, 20, 30) measured photons going further away from Sun (~forward scatter) Ω(10, 20, 150) measured photons going closer to Sun (~backscatter)
400 nm 640 nm Water-leaving radiance in the Sun plane, zenith dependence (arrow length indicates radiance value) Bottom line: Water-leaving radiance, or reflectance, is a function of angles.
Current, “standard”, approach to deal with angular variation: empirical Rrs(Ω) Chla Rrs[0] IOPs f/Q based on Case-1 The two semi-empirical steps could be omitted with an Rrs model accounts for the angular effects.
Candidate models for angular Rrs: Albert and Mobley (2003) Lee et al (2004) Park and Ruddick (2005) Van Der Woerd and Pasterkamp (2008) Caveats: 1. Some are not visible/transparent about the physics 2. Some are not easily invertible algebraically
A practical choice for algebraic/analytical inversion: G from model [sr-1] Global distribution of Rrs(443) 1: 1 (Ω: 60, 40, 90) G from HL simulation [sr-1] Rrs 443 [sr-1] G ~ 0. 07
Angular-dependent model coefficients for Rrs(Ω): Table ((7 x 13+1)x 4 x 6) array, 2208 elements) of {G(Ω)} (if based on Chl, it is 6 x 13 x 7 = 546 elements per band per Chl)
IOP retrieval from angular Rrs: G[Ω] Rrs(Ω) IOPs QAA, optimization, linear matrix, etc. Now we get IOPs straightforwardly (one step) from Rrs(Ω)! empirical Rrs(Ω) Chla Rrs[0] IOPs f/Q based on Case-1
2. QAA vs spectral optimization All start with: “essential” difference: 1) “Philosophic” difference 2) Difference in measuring “signal” vs “noise” 3) Impact of data and/or model 4) Processing efficiency 5) Stability
1) “Philosophic” difference QAA: Total first, then individuals Optimization: Individuals first, then total or simultaneously Under QAA, the error propagation is visual and easy to quantify Δbbp, Δa will be 0 if ∆a(λ 0) and ∆η are 0.
(Lee and Carder 2004) Spectral shape of aph(λ) is a property we want to obtain from each measured Rrs(λ). Spectral optimization assumes a spectral shape before its derivation. aph(λ) [m-1] In addition: Wavelength [nm]
2) Difference in measuring “signal” vs “noise” Algebraic algorithm (QAA, LMI) (Lee et al. 2002, Hoge and Lyon 1996) Every measurement is perceived as signal. Optimization algorithm (e. g. , GSM 01, HOPE) (Roesler and Perry 1996, Lee et al. 1996, Maritorena et al. 2001, Doerffer 1999) Mis-match is perceived as measurement noise.
3) Impact of data and/or model The increase trend of bbp from Aqua, by GSM 01, is considered due to “error” of Aqua Rrs 412. QAA will have different patterns, at least for bbp, as it does not use Rrs 412 for bbp derivation. (Maritorena et al 2010)
4) Processing efficiency (Lee et al 2002) ‘Resolved’ for multi-band data; hyperspectral data? 5) Stability Optimization: software impact; optimized or closed to be optimized?
The pathway for global IOPs products by GIOP: As QAA and optimization schemes (and other semianalytical algorithms) initiate from the same physics, then IOPs from both retrievals can serve as a consistence/reliability check: Agreeable results highly reliable! Different results need further diagnose …
Thank you!
- Slides: 25