A step back is a move forward

2017-10-16T13:35:25Z (GMT) by Thomas Lavergne
Illustration to discuss that model-observation comparison would best be conducted at a middle-ground "location" (along the x axis) somewhere in between the model world (left) and the direct (satellite) observations at top-of-atmosphere (right).<div><br></div><div>The classical approach is that either 1) Earth Observation (EO) algorithms transform the satellite signal all the way to EO retrieval products that are directly in the modelling world (right to left), or 2) the variables in the model world are transformed all the way into direct satellite measurements using Observation Operators (Obs Ops, left to right).</div><div><br></div><div>Starting from a very common case where EO retrievals are produced to be directly comparable to models, the author advocates that the EO scientists should not be afraid to "take a step back" and rather deliver EO products that are maybe less polished and processed, but are closer to what the physics of EO measurement allows. At the same time, the EO scientists must provide modelers with Observation Operators that bridge the gap between model variables and the new, less processed EO retrieval.</div><div><br></div><div>This is illustrated here with the case of Sea Ice Concentration (SIC) EO products from Passive Microwave (PMW) imagery. These are actually more similar to an Ice Surface Fraction (ISF) (thus including both ocean water in leads and melt-pond water, Kern et al. 2016), and in addition are known to underestimate very thin ice (Ivanova et al., 2015).</div><div><br></div><div>By not pretending the EO retrieved SIC products are the exact same as the physical modeled SIC, EO scientists take a step back, give an observation operator to the modelers, that allow for more accurate model-observation comparison.</div><div><br></div><div>A step back is a move forward. CQFD.</div><div><br></div><div>Additional remarks:</div><div>1) it is from time to time more important to pin-point "what" an EO product is (its position along the x-axis) than improving its value or uncertainty (along the y-axis of the plot) by few percents.<br></div><div><br></div><div>2) To design accurate Observation Operators require deep understanding and knowledge of the EO signal, and often the processing of additional EO data. It can take substantial effort to take the step back.</div><div><br></div><div>References:</div><div>Kern, S., Rösel, A., Pedersen, L. T., Ivanova, N., Saldo, R., and Tonboe, R. T.: The impact of melt ponds on summertime microwave brightness temperatures and sea-ice concentrations, The Cryosphere, 10, 2217-2239, https://doi.org/10.5194/tc-10-2217-2016, 2016.<br></div><div><br></div><div>Ivanova, N., Pedersen, L. T., Tonboe, R. T., Kern, S., Heygster, G., Lavergne, T., Sørensen, A., Saldo, R., Dybkjær, G., Brucker, L., and Shokr, M.: Inter-comparison and evaluation of sea ice algorithms: towards further identification of challenges and optimal approach using passive microwave observations, The Cryosphere, 9, 1797-1817, https://doi.org/10.5194/tc-9-1797-2015, 2015.<br></div>