Multi-model Ensemble Downscaling for Temperature Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current weather-climate forecasting models face limitations in spatial resolution, particularly for short- to seasonal-term forecasts, which affects the accuracy of temperature prediction and is inadequate for managing energy resources and industrial logistics effectively.
Innovation Solution
A unified approach combining dynamic and statistical systems for seamless prediction, using a multi-model ensemble down-scaling method that integrates global and regional models to provide high-resolution temperature forecasts, incorporating empirical-statistical models for local conditions and surface interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If global climate models with coarse horizontal resolution (50-250 km) are used, then they can provide seasonal forecasts on a global or continental scale, but they cannot provide detailed temperature predictions on regional or local scales
Solution Approach 1:
The patent divides the forecasting system into multiple nested models operating at different spatial scales. A global model provides boundary conditions for regional models, which in turn provide conditions for local models. This segmentation allows each model to operate at its optimal resolution, with the global model covering large areas and local models providing detailed temperature predictions for specific regions.
Solution Approach 2:
The patent implements a nested modeling approach where regional and local models are embedded within the global model framework. The global model domain contains regional model domains, which contain local model domains. This nesting allows information to be passed from coarse to fine scales, with each level providing boundary conditions for the next, thereby achieving both broad spatial coverage and high local resolution.
2Measurement precision
If higher resolution regional models are used to improve local temperature prediction accuracy, then detailed local forecasts can be obtained, but the computational complexity and data requirements increase significantly
Solution Approach 1:
The patent uses global and regional models to establish boundary conditions and large-scale atmospheric patterns in advance before running local high-resolution models. This preliminary action provides the local models with preprocessed information about synoptic-scale weather patterns, reducing the computational burden and complexity of running standalone high-resolution models while maintaining prediction accuracy.
3Device complexity
If statistical models based on historical data are used, then they can provide simple seasonal forecasts, but they cannot capture dynamic weather processes and climate interactions
Solution Approach 1:
The patent merges statistical downscaling techniques with dynamic regional climate models. The statistical component identifies relationships between large-scale atmospheric patterns and local temperature variations from historical data, while the dynamic model simulates the physical processes. This combination allows the system to capture both historical statistical patterns and dynamic weather processes, improving forecast reliability while maintaining reasonable complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method is described for the weather-climatic temperature forecasting, from short to long term, comprising the phases of acquiring meteorological parameters of a large-scale (SG) geographical area having a pre-determined amplitude; decomposing the large-scale geographical area (SG) into a base area and a regional area (SR); determining the temperature close to the surface of the base area, starting from the parameters available on the large-scale geographical area (SG), using an empirical-statistical model (statistical down-scaling); determining the tendencies of the meteorological parameters in the regional area (SR), starting from the meteorological parameters available on the large-scale geographical area (SG), using a dynamic numerical model (dynamic down-scaling); performing the combination (ensemble down-scaling), through an applicative model, of the empirical- statistical model (statistical down-scaling) and the dynamic numerical model (dynamic down-scaling) to obtain in continuous, from short term to seasonal term, the temperature forecast close to the surface. The applicative model adds a statistical scaling of the data for temperatures close to the surface which are therefore re-assimilated in the regional area (SR) as new temperature values close to the physical boundary of the regional area (SR), introducing, during the dynamic down-scaling phase, a range of pseudo- observations, which properly act on the regional area (SR). In this way, during the dynamic down-scaling phase, a reconstructed range of observations is introduced, de facto, which are on a spatial scale compatible with that of the regional model.