Development of a Machine Learning-Based Radiometric Bias Correction for NOAA’s Microwave Integrated Retrieval System (MiRS)

Abstract

We present the development of a dynamic over-ocean radiometric bias correction for the Microwave Integrated Retrieval System (MiRS) which accounts for spatial, temporal, spectral, and angular dependence of the systematic differences between observed and forward model-simulated radiances. The dynamic bias correction, which utilizes a deep neural network approach, is designed to incorporate dependence on the atmospheric and surface conditions that impact forward model biases. The approach utilizes collocations of observed Suomi National Polar-orbiting Partnership/Advanced Technology Microwave Sounder (SNPP/ATMS) radiances and European Centre for Medium-Range Weather Forecasts (ECMWF) model analyses which are used as input to the Community Radiative Transfer Model (CRTM) forward model to develop training data of radiometric biases. Analysis of the neural network performance indicates that in many channels, the dynamic bias is able to reproduce realistically both the spatial patterns of the original bias and its probability distribution function. Furthermore, retrieval impact experiments on independent data show that, compared with the baseline static bias correction, using the dynamic bias correction can improve temperature and water vapor profile retrievals, particularly in regions with higher Cloud Liquid Water (CLW) amounts. Ocean surface emissivity retrievals are also improved, for example at 23.8 GHz, showing an increase in correlation from 0.59 to 0.67 and a reduction of standard deviation from 0.035 to 0.026.

In‐Depth Evaluation of MiRS Total Precipitable Water From NOAA‐20 ATMS Using Multiple Reference Data Sets

“…Schulte and Kummerow (2019) showed that much of the bias was reduced by accounting for sea surface temperature in a bias correction used within a 1DVAR algorithm applied to AMSU‐A/MHS data. More recently, Zhou and Grassotti (2020) applied a neural network method for ATMS radiometric bias correction that was dynamic and varied according to the characteristics of the observed scene. Their results showed that the scan dependence of the TPW bias was largely eliminated, with the largest improvement at near‐nadir zenith angles.…”

Enhance Low Level Temperature and Moisture Profiles Through Combining NUCAPS, ABI Observations, and RTMA Analysis

“…This number is determined with reference to previous works by Tao et al. (2016, 2018) and Zhou and Grassotti (2020), after a few tests, it showed that the improvement of adding more layers for this task is not substantial enough when compared with the increased cost of computation and risk of overfitting. The input layer consists of all the predictors for the training, including the NUCAPS temperature/moisture values for all levels 200 hPa and below, ABI variables, and RTMA variables mentioned in Table 1, with each predictor serving as a neuron.…”

“…In recent years, DNN‐ based algorithms have gained success in bias correction for satellite‐derived products, as Tao et al. (2016) who successfully applied DNN in reducing bias and false alarms of satellite precipitation products, and Zhou & Grassotti (2020) who implemented DNN as a new approach of radiometric bias correction in MiRS system. DNN in these works have shown good capability on addressing nonlinear relationships, and high efficiency in generating predictive models with good accuracy.…”

Application of the Deep Neural Network in Retrieving the Atmospheric Temperature and Humidity Profiles from the Microwave Humidity and Temperature Sounder Onboard the Feng-Yun-3 Satellite

“…It adjusts the initial value of the atmospheric parameter through an iterative process, with the aim that the bias between the observed brightness temperature and the simulated brightness temperature calculated by the radiative transfer model using the initial value (observation bias) satisfies a certain threshold, at which point the adjusted initial value is the retrieved value of the atmospheric parameter [ 19 ]. However, the 1DVAR requires the observation bias to satisfy the unbiased and Gaussian characteristics, so the observation bias must be quantified and removed, and the removal of this bias can be performed by using a bias-correction method based on neural networks (NNs) [ 20 , 21 ]. The second retrieval scheme is based on the statistical relationship between the observed brightness temperature and the atmospheric temperature and humidity profiles.…”

“…Although the global reanalysis can provide a variety of atmospheric parameters with high spatial resolution and high accuracy, it suffers from a long time delay (one month or more) compared with the satellite observations, which cannot meet the requirements for real-time in atmospheric applications, such as numerical weather forecasting, extreme weather monitoring, etc. Microwave radiometer, which takes Sensors 2021, 21, 4673 2 of 20 a passive microwave remote sensing approach, is an important instrument to monitor the Earth-Atmosphere system, and its observation is an important data source to obtain information about atmospheric temperature and humidity in atmospheric science [5,6]. The retrieval algorithm can be used to convert microwave remote sensing measurements into atmospheric temperature and humidity parameters [7].…”