The water year ended on September 30, so the Land IQ team has spent some time looking back at our efforts over the last year to collect data points for our crop mapping. Over the last twelve months, our team drove the equivalent of 8.5 trips across the country, that is over 23,000 miles! And this doesn’t even consider all the miles traveled in Australia.
Land IQ has a fundamental belief that the accuracy of any product is a result of the high-quality data collected. Land IQ collects ground truth data for numerous projects, but the most vigorous set of data collected is for our Statewide Land Use Mapping project in cooperation with the Department of Water Resources. For the most recently released data set (Water Year 2021) over 15.5 million acres were mapped with an accuracy of 97%, which was the result of the collection of over 88,238 data points across 46 crop classes. Look for the provisional release of 2022 coming soon!
Ground truth data is a key component in determining the accuracy of remote sensing and machine learning. Ground truthing allows the user to check the results of the remote sensing analysis against the real world by collecting real features and materials on the ground. The collection of this data enables the user to both calibrate the algorithms used, as well as validate the accuracy of the results.
In determining accuracy in our mapping, error results vary among crop types. In particular some less extensive crops that have very few validation samples may have a skewed error result depending on the number and nature of validation sample points. This helps our team in future mapping events develop the best routes for collecting the most meaningful data, which means thousands of miles traveled across California.
Collecting ground truth data is one of the first steps in our crop mapping timeline. Field boundaries have been updated, and the team is starting the classification process for Water Year 2023. It won’t be long before the team is “on the road again,” collecting points for the 2024 mapping event.