Issues with Describing the Uncertainties in Atmospheric Remote Sensing Measurements

Abstract:

Uncertainty in atmospheric measurements from satellites and other remote sensing platforms comes from several sources. Users are familiar with concepts of accuracy and precision for physical measurements made using instrumentation, but retrieval algorithms also frequently require statistical information since measurements alone may not completely determine the parameter of interest. This statistical information has uncertainty associated with it as well, and it often contributes a sizeable fraction to the total uncertainty. The precise combination of physical and statistical information in remotely sensed data can vary with season, latitude, altitude, and conditions of measurement. While this picture is complex, it is important to clearly define the overall uncertainty for users without oversimplifying so they can interpret the data correctly. Assessment of trends, quantification of radiative forcing and chemical budgets, and comparisons of models with satellite observations all benefit from having adequate uncertainty information. But even today, terminology and interpretation of these uncertainties is a hot topic of discussion among experts. Based on our experience producing a 44 year-long dataset of total ozone and ozone profiles, we discuss our ideas for describing uncertainty in atmospheric datasets for global change research. Assumptions about the atmosphere used in retrievals can also be provided with exact information detailing how the final product depends on these assumptions. As a practical example, we discuss our modifications to the Total Ozone Mapping Spectrometer (TOMS) algorithm in Version 9 to provide robust uncertainties for each measurement and supply as much useful information to users as possible. Finally, we describe how uncertainties in individual measurements combine when the data are aggregated in time and space.