EO in Orbit: Scientific webinars – Croplands
Overview
- Type: Webinar
- Theme: Croplands
- Date:
- Time: 11:00 a.m. to noon ET
- Duration: 1 hour
- Cost: Free
- Location: Virtual
- Language: English
- Target audience: Industry, academic institutions, media, not-for-profit organizations, scientists, government.
Summary
Presentations will focus on scientific developments in the field of Earth observation (EO) and croplands.
Detailed description
Presentations:
- Leveraging Machine Learning and Earth Observation Data for Official Statistics: Tracking Woody Biomass Change on Canadian Croplands
- Tracking Daily Changes in Crop Condition using Synthetic Aperture Radar
Leveraging Machine Learning and Earth Observation Data for Official Statistics: Tracking Woody Biomass Change on Canadian Croplands
Presentation 1 (in English)
From 11:00 to 11:30 a.m.
Olivier Godard
Statistics Canada
In this session, the Statistics Canada Data Science Division will present an application of a convolutional neural network (CNN) model in utilizing EO data to support official statistics. Specifically, we will present a case study focusing on tracking woody biomass change on Canadian croplands. Currently, there is a lack of accurate methods to estimate carbon stocks in woody vegetation species on farmlands. However, with the availability of high-resolution optical imagery, we demonstrate the feasibility of monitoring changes in vegetation areas down to the level of individual trees.
The Statistics Canada Data Science Division, in collaboration with Agriculture and Agri-Food Canada (AAFC), is actively working on implementing this methodology on a national scale. Nevertheless, a major challenge faced is the high acquisition cost associated with obtaining high-resolution imagery. Through this webinar, we aim to make the community aware of the value of this data input for enhancing official statistics and facilitating informed decision-making for Canadians.
Furthermore, we highlight that the image segmentation model employed in this study was originally developed for detecting greenhouses. We will demonstrate its transferability and potential for other similar applications. These applications include the detection of buildings, construction sites, and various human-made objects. By showcasing the versatility of this deep-learning architecture, we emphasize its broader utility beyond the specific use case of tracking woody biomass change.
Tracking Daily Changes in Crop Condition using Synthetic Aperture Radar
Presentation 2 (in English)
From 11:30 a.m. to noon
Heather McNairn
AAFC
Operations to track crop condition at national and regional scales most often rely on coarser-resolution optical satellite data. AAFC is developing a method to estimate crop condition using synthetic aperture radar (SAR) data from satellites that include RADARSAT and Sentinel-1. AAFC calibrates polarimetric parameters from these satellites, to estimate the Normalized Difference Vegetation Index (NDVI) using machine learning algorithms. These SAR-calibrated estimates are then integrated into a Crop Structure Dynamics Model. The output provides a daily estimate of crop condition, at sub-field scales. Models have been developed for canola, soybeans, corn and wheat. The next step in this process is to test over wider geographies and to assess if these products can improve estimates of crop yield.