Hyperlocal Weather Forecasting via Sky Image Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current weather forecasting technologies do not provide sufficient accuracy and resolution for farmers with smaller land holdings, as they rely on broader geographical averages, leading to inaccuracies in weather predictions crucial for agronomic decisions.

Innovation Solution

A method and system for downscaling weather forecasts to a hyperlocal level by using a mobile device to obtain real-time sky images and location data, which are then used to generate more accurate local weather forecasts through cloud cover and type analysis, either on the device or a remote server, without relying on historical data, utilizing artificial neural networks and Bayesian inference for improved precision.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If weather forecasts are provided at kilometre resolution using existing forecast models, then global coverage and broad area forecasting are achieved, but accuracy and resolution for small land holdings are insufficient

Engineering Contradiction:

Improveweather forecast accuracyVSAvoidgeographical coverage area

Core Design Contradiction:

Measurement precisionVSArea of stationary object

Solution Approach 1:

The patent segments the broad geographical area into smaller hyperlocal zones by dividing the sky field of view into multiple regions corresponding to different directions. Each segment is analyzed independently for cloud cover and weather conditions, enabling precise forecasting for small land holdings while maintaining broader coverage through aggregation of multiple segments.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by capturing and analyzing sky images from multiple specific directions (north, south, east, west) rather than using a single averaged view. Each directional sky image provides localized weather information for specific portions of the landscape, allowing farmers to receive tailored forecasts for their particular field orientations and microclimates.

Inventive Principle:

Principle #3Local quality

2Ease of manufacture

If weather forecasts rely on broader geographical averages, then computational simplicity and data availability are maintained, but accuracy for specific farm locations deteriorates

Engineering Contradiction:

Improveforecast generation simplicityVSAvoidlocation-specific forecast accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The system enables self-service by using the farmer's existing mobile device camera to capture sky images, automatically processing these images through machine learning algorithms to generate hyperlocal forecasts without requiring external weather stations or complex infrastructure at the farm location.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent changes the fundamental parameter of weather data collection from ground-based meteorological measurements to sky-based visual imagery. By capturing cloud patterns, cover, and movement from multiple sky directions, the system transforms weather forecasting from relying on averaged geographical data to using direct visual observations of atmospheric conditions specific to each location.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If historical cloud information is required for local weather forecasting, then forecast accuracy is improved, but real-time responsiveness and speed of deployment are reduced

Engineering Contradiction:

Improveforecast accuracyVSAvoiddata collection time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-training machine learning models with extensive historical sky image data and cloud classification information before deployment. Once trained, the models can instantly classify new sky images and generate forecasts in real-time without requiring additional historical data collection, achieving both accuracy and speed.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces the mechanical system of collecting and processing historical weather data with an artificial intelligence-based image recognition system. The machine learning model directly analyzes current sky images to infer weather patterns and predict future conditions, eliminating the need for time-consuming historical data aggregation while maintaining or improving forecast accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11500126B2Downscaling weather forecasts

Publication Date: 2022.11.15 YARA INTERNATIONAL ASA

AI summary

A method for downscaling a weather forecast including obtaining a sky image and location data indicative of a user location via a mobile computing device; sending the location data to a weather forecast provider; generating, by the weather forecast provider, a local weather forecast for the user location; sending the local weather forecast to a server; determining, by the mobile computing device, cloud cover data and cloud type data based on the sky image, and sending the cloud cover data and cloud type data to the server or sending the sky image to the server and determining cloud cover data and cloud type data based on the sky image; increasing the resolution of the local weather forecast based on the cloud cover data and the cloud type data, thus obtaining a downscaled local weather forecast; and, sending the downscaled local weather forecast to the mobile computing device.