There is no clear agreement yet on what is the systematic part of the uncertainty, and on what the random part of the uncertainty in FTIR measurements is and how to evaluate each part. Random and systematic uncertainty sources are defined differently for the two main retrieval software distributions within the FTIR NDACC working group (PROFFIT and SFIT). To harmonize the uncertainty computation, a recipe should be developed as to how a random and systematic uncertainty should be determined for each of the leading uncertainty contributions in the target retrieval uncertainty budget. The distinction between systematic and random uncertainties is important for determining accuracy and precision, e.g. when comparing to satellite data, and uncertainty of an average of data.
The uncertainty calculation routines of the SFIT4 retrieval software package has been adapted so that the uncertainty budgets between both PROFFIT and SFIT4 are comparable.
Specific remedy proposed
Comparison and tuning of the uncertainty modules of the retrieval software packages. Write down a manual of how to estimate the uncertainties for all parameters that are part of the forward model in the retrieval packages.
Measurable outcome of success
Comparable and consistent errors for all the different sites.
Achievable outcomes
Technological / organizational viability: high
The development is currently ongoing.
Indicative cost: medium (>1 million)
Relevance
The agreement on the input data for the uncertainty calculations will assure that the error estimations are comparable between different sites.
Timebound
We are now woeking on reprocessing FTIR CO for QA4ECV, 3 years.
Identified future risk / impact |
Probability of occurrence if gap not remedied |
Downstream impacts on ability to deliver high quality services to science / industry / society |
Incomparable uncertainty budgets for different sites on NDACC |
High, it occurs right now |
Difficulty of a network-wide and consistent data usage |