[Research on the Noise Reduction with Hyper-Resolution Infrared Spectrum Based on Improved PCV Method].

Area of Science:

• Atmospheric Science
• Remote Sensing
• Data Analysis

Background:

• Accurate thermodynamic profile retrieval from infrared radiance is vital.
• Current Principal Component Analysis (PCA) noise filters often use subjective statistical methods or require real-time Noise-Equivalent Spectral Radiance (NESR), which is difficult to obtain.
• Existing methods for determining the optimal number of principal components (k) have limitations in objectivity and applicability.

Purpose of the Study:

• To develop an improved PCA noise filter for high-resolution infrared radiance data.
• To overcome the limitations of arbitrary threshold selection in the Percent Cumulative Variance (PCV) method.
• To eliminate the need for real-time NESR in noise reduction for thermodynamic profile retrieval.

Main Methods:

• A novel PCA noise filter based on an Improved PCV algorithm was developed.
• The threshold for PCV was determined by iteratively comparing simulated and reconstructed spectra.
• The impact of normalization on noise reduction and temperature profile retrieval was analyzed.

Main Results:

• The Improved PCV method objectively determines the optimal number of principal components (k) without requiring real-time NESR.
• Noise reduction using the Improved PCV method showed a 0.1 K Root Mean Square Error (RMSE) improvement in retrieved temperature profiles compared to the factor indicator function method when NESR was unavailable.
• The impact of normalization on retrieval accuracy was found to be less significant than the error in calculating k.

Conclusions:

• The proposed Improved PCV method offers an objective and reasonable approach to reduce noise in ground-based high-resolution infrared radiance data.
• This method enhances the accuracy and stability of thermodynamic profile retrieval, particularly when real-time NESR is not accessible.
• The study highlights the importance of objective k determination for effective PCA-based noise filtering in remote sensing applications.