Multisource sea surface temperature remote sensing product fusion method based on robust fixed order filtering model

Abstract

The invention discloses a multisource sea surface temperature (SST) remote sensing product fusion method based on a robust fixed order filtering model. The method synthesizes a scale transformation problem in a time-space fusion process of SST remote sensing products, an uncertainty expression problem of the products and complicated space structure and time structure of an SST time-space process, fully utilizes the complementary characteristics of the remote sensing products in the aspects of spatial resolution, time-space integrity and accuracy characteristics and adopts hierarchical Bayes frame based on the robust fixed order filtering model to fuse infrared SST and microwave SST remote sensing products to obtain fine scale fusion data with the high accuracy, the complete space and the rich local spatial models. The method has more reasonable simulation of SST space trend, the seamless spatial-temporal scale transformation between the multisource remote sensing products and the quantification of the uncertainty of the remote sensing products are realized, and the method is suitable for high efficient calculation of remote sensing large data volume.

Description

technical field [0001] The invention relates to the technical field of marine remote sensing, in particular to a fusion method of multi-source ocean surface temperature remote sensing products based on a robust fixed-order filtering model. Background technique [0002] Sea surface temperature is one of the important parameters in the global air-sea coupling system. It plays the most basic role in the exchange of energy and water vapor between the ocean and the atmosphere. It has been widely used in upper ocean processes, climate change detection, In the research of sea-air heat exchange, ocean-atmosphere numerical simulation and forecast, and ocean-atmosphere assimilation model. [0003] At present, the means of obtaining SST mainly include traditional on-site observation and modern remote sensing observation. Traditional on-site observations are mainly made through conventional observation systems such as ships, fixed and drifting buoys at sea, and offshore observation pla...

AI Technical Summary

Problems solved by technology

In the time series reconstruction algorithm of remote sensing products based on time filtering, there are mainly the following problems: lack of utilization of the dynamic process model of remote sensing parameters; insufficient use of historical prior knowledge; no definition and definition of data uncertainty during time series reconstruction. Quantification; spatial heterogeneity is not considered; in the calculation process of most methods, the determination of the time window is affected by subjective experience, and there is a large uncertainty in the choice of noise and real information, so the accuracy of the filtered product is affected by certain influence

Although the spatial filtering method fully considers the spatial heterogeneity, there are still the following problems: the spatial filtering performs interpolation according to the spatial relationship of the pixel neighborhood, without considering the dependence of the time dynamic process between the data, and due to the optimal unbiased estimation condition The variance of the spatial interpolation is lower than that of the original data, which has an obvious smoothing effect, and the local spatial detail information is not easy to maintain; affected by the density of the observed data, the spatial interpolation error is large in large areas of missing data; limited by the covariance inversion , it is difficult to achieve efficient calculation of large amounts of data, although the FRK and FRF methods achieve spatial dimensionality reduction through multi-resolution wavelet basis functions, solve the calculation problem of large data volume covariance inversion in traditional geostatistical methods, and have high computational efficiency , but the problem that the data after spatiotemporal interpolation is too smooth has not been solved, and these two methods are for spatiotemporal interpolation in irregular areas such as SST, and the calculation of multi-variable wavelet basis functions is unstable, which needs to be improved

However, this method uses the space-time covariance model to realize the space-time time difference, which has a large calculation rate problem, and it is difficult to combine with the space-time process model

Images