The Center for Atmospheric Sciences | » Quantifying the Accuracy of Combined Artificial Intelligence Microwave/Optimal Estimation Infrared Single Footprint All-sky Retrievals in the Planetary Boundary Layer

  • Researchers

  • Research Projects

  • Quantifying the Accuracy of Combined Artificial Intelligence Microwave/Optimal Estimation Infrared Single Footprint All-sky Retrievals in the Planetary Boundary Layer

    This project primary focus is to assess the accuracy of thermodynamic retrievals in the Planetary Boundary Layer (PBL) using the current generation of satellite-based microwave (MW) and infrared (IR) instruments. When compared to measurements from aircraft and continental US ground based lidar/ceilometers (L/C)networks and global sonde database, will allow us to carry out the assessment over all seasons and under a variety of climate conditions. An important parameter, namely the PBL height (PBLH), will be computed from satellite instrument retrievals using the derivative of the relative humidity with respect to pressure.

    The combined MW and hyperspectral IR sounder thermodynamic retrieval will work at the single field of view IR horizontal resolution (currently 15 km) under heterogeneous cloud conditions. The microwave sounder will provide thermodynamic profiles even if the conditions are cloudy; the infrared sounder will improve the vertical resolution of the retrieved products. The core of the retrieval will be a combination of MIIDAPS-AI, an Artificial Intelligence (AI) based algorithm for the microwave (MW/IR) retrievals, fed into a physically based Optimal Estimation Method (OEM) algorithm coupled to a fast scattering radiative transfer algorithm (RTA) for the infrared retrievals. The novelty of the single footprint infrared (SFPIR) retrieval is the cloud initialization (cloud loading, cloud top/bottom, cloud fraction), using a combination of “best initial fit” clouds from a Numerical Weather Prediction (NWP) model and cloud profiles obtained from the AI/MW retrieval.

    The flexibility of the AI retrieval will allow us to switch between generating AI MW only, AI MW+IR or AI IR only profiles as initialization for the physically based SFPIR retrieval. The modular design of the AI(MW/IR)/SFPIR code will allow us to easily switch between sensor pairs such as NOAA’s Cross Track Infrared Sounder (CrIS) and the Advanced Technology Microwave Sounder (ATMS) on the Suomi and JPSS satellites, or the Atmospheric Infrared Sounder (AIRS) and the Advanced Microwave Sounding Unit (AMSU) on board NASA’s Aqua satellite. We anticipate that our methodology will immediately achieve the following improvements over existing IR sounder products (AIRS L2, AIRS and CriS NUCAPS): (a) retrievals at the native resolution of the sounders (15 km) instead of the 3×3 45 km footprint currently achieved using cloud clearing (b) improved yield of retrievals (currently 60% globally) and easier quantification of quality assurance (using the degrees of freedom).

    Validation and quantification of PBL parameters will use a combination of radiosondes, aircraft and ground based (L/C) data. Thermodynamic profile validation will include radiosonde data (from Global Climate Observing System (GCOS) Reference Upper-Air Network (GRUAN)), DOE Atmospheric Radiation Measurements, profiles from ground-based (L/C) and MW radiometer data, lidar profiles from airborne instruments and AIRS and CriS L2 products.

    Importantly, in line with the requirements of this call, we will also have access to water vapor and aerosol extinction profiles from the NASA Langley’s aircraft-borne High Altitude Lidar Observatory (HALO). We have combined experience in using AI based algorithms to detect boundary layer thickness from lidars and will apply this knowledge to studying the thermodynamic retrievals from the combined AI(MW/IR)/SFPIR retrieval. This project is contributing to NASA’s Decadal Survey Incubation program as recommended in the 2017 Earth Science Decadal Survey.