Humidity, clouds and snow in the Arctic

The Arctic climate is changing and the temperatures in the region are predicted to increase much more rapidly than the global average.

Water is here key; the surface consists mainly of seawater, ice, or snow, and the weather is largely determined by the humidity and clouds in the atmosphere.

These water reservoirs are not independent, atmospheric water and circulation are strongly influenced by the distribution of sea ice, which in its turn is affected by e.g. clouds´ influence on heat fluxes.

This project will improve the use of satellite microwave radiometry to follow this intricate interplay inside the climate system.

The primary objective is to provide better information on humidity, but water in all forms is involved and contributions regarding liquid water content, falling snow, and sea ice variability are also foreseen.

Sensors supported by the project include SSMIS, MHS, ATMS, MWI, and AWS.To overcome the limitations in present retrievals, sophisticated 3D radiative transfer and machine learning (ML) will be combined.

In short, SAR imagery and high-resolution reanalysis data give input for simulating scenes of measured brightness temperatures, that are used to train the ML algorithm. We are here applying a methodology introduced by us, providing robust uncertainty estimates in contrast to standard ML.

By using input having high spatial resolution and simulating scenes of satellite footprints, we avoid simplifications applied in standard retrievals as well as in data assimilation.

Most importantly, no remapping of data is required and we can make use of the spatial information contained in the overlap between the footprints.

We can also handle the case of an inhomogeneous surface, that is required to provide accurate data over sea ice leads and along coastlines. To our best knowledge, all these aspects are fully novel.

As a consequence, while other data users must reject large fractions of the observations, our approach needs no filtering with respect to the interference of clouds and the surface.

Albeit, some retrievals will have poor precision, but the ML approach will capture this, and low weight can be given to those cases in a statistically sane manner.Both conical (SSMIS) and cross-track (such as ATMS) sensors will be considered.

All data, for several years, of one instrument of each type, will be processed, to provide retrieval datasets to be explored in e.g. climate studies.

The overall procedure requires a deep understanding of the physics governing the observed radiances, knowledge that can be transferred to others and we expect a significant impact of the project on the assimilation of the same data for weather forecasting.