loading page

Machine Learning Approach to Classify Precipitation Type from A Passive Microwave Sensor
  • +3
  • Spandan Das,
  • Jie Gong,
  • Chenxi Wang,
  • Dong Wu,
  • Stephen Munchak,
  • William Olson
Spandan Das
Thomas Jefferson High School for Science and Technology

Corresponding Author:[email protected]

Author Profile
Jie Gong
NASA Goddard Space Flight Center
Author Profile
Chenxi Wang
University of Maryland College Park
Author Profile
Dong Wu
NASA/Goddard Space Flight Cent
Author Profile
Stephen Munchak
NASA Goddard Space Flight Center
Author Profile
William Olson
Joint Center for Earth Systems Technology
Author Profile


Precipitation flag (precipitating or not; stratiform or convective) is a key parameter for us to make betterretrieval of precipitation characteristics as well as to understand the cloud-precipitation physicalprocesses. The Global Precipitation Measurement (GPM) Core Observatory’s Microwave Imager (GMI)and Dual-Frequency Precipitation Radar (DPR) together provide ample information on globalprecipitation characteristics. As an active sensor in particular, DPR provides an accurate precipitationflag assignment, while passive sensors like GMI were traditionally believed not to be able to tell apartprecipitation types. Using collocated precipitation flag assignment from DPR as the “truth”, this project employs machinelearning models to train and test the predictability and accuracy of using passive GMI-only observationstogether with ancillary atmosphere information from reanalysis. Precipitation types are classified intothe following classes: convective, stratiform, convective-stratiform mixed, no precipitation, and otherprecipitation. Sub-sampling with different probabilities is employed to construct a balanced trainingdataset. A variety of classification algorithms are tested, including Support Vector Machines, NaiveBayes, Random Forests, Gradient Boosting, and Neural Networks (Multilayer Perceptron Network), andtheir results are evaluated and compared. The trained model has ~ 85% of prediction accuracy for everytype of precipitation. High-frequency channels (166 GHz and 183 GHz channels) and 166 GHzpolarization difference are found among the most important factors that contribute to the modelperformance, which shed light on future instrument channel selection.