Research On Aerial Biomass Changes Based On Weather Radar Monitorin

With the in-depth research of information technologies such as big data,artificial intelligence and machine learning,technical support is provided to extract the potential value of large amounts of data.With the powerful computing power of modern computers,carrying out cross research between weather radar and big data technologies is a highly promising research direction.Therefore,this paper takes machine learning technology as the research background to study the potential biological information in dual-polarisation weather radar detection data.The main research components include the following three points:Firstly,the extraction of dual-polarimetric radar bio-echoes is addressed.By parsing the secondary radar-based data,the polarisation variables of the dual-polarisation radar are obtained,and these polarisation variables are converted to a Cartesian coordinate system to facilitate oneto-one correspondence across the detection space.A manual annotation method is used to produce the bio-echo dataset.Three machine learning algorithms,Random Forest,LGBM and XGBoost,are designed to learn the obtained datasets and compare the performance of the models.The best performing model was selected for further optimisation,resulting in a further improvement of 0.28% in the overall performance of the model.A multi-supination angle-based hierarchical VPR inversion model is then developed based on radar beam propagation theory and the geometric observation principle of weather radar.Under the premise that the scatterers are uniformly distributed in the radar beam,the correspondence between the scattering cross section of the radar beam and the sampling distance is established with reference to the refraction model of the radar beam in the standard atmosphere.The overlapping areas of different height layers and scattering cross sections are derived to obtain the reflectivity factor contribution coefficients for different elevation angle and distance combinations,so as to obtain high-resolution vertical profiles of the reflectivity factor.Finally,the statistical analysis of the biomass in the Poyang Lake and Bohai Rim regions is carried out by combining the above methods.The biological echoes in the radar data were extracted using the XGBoost algorithm,while the vertical distribution of aerial fauna was inverted using the VPR algorithm;the annual,seasonal,daily and temporal changes in the biomass of the calculated regions were compared and analysed.The inversion of the historical data further validates the reliability of the above methods and explains some of the airborne animal activity patterns.