As with all AI tools, these models are not guaranteed to provide accurate inferences.
Predictions from the model are not endorsements or recommendations.
For all questions, please contact jordan.ubbens@nrc-cnrc.gc.ca.
For models with "weekly" in the name, you are able to specify weather in your query on a weekly basis. For those with "monthly" in the name, weather is specified on a monthly basis. Use a weekly model if you require specifying weather at week-level granularity, but if not a monthly model will run significantly faster.
Variables added under the Add Control Variables section will be held constant while performing inference. If there are variables you would like to control for, it is important that you include them here. Otherwise, the model will infer their value. For example, early seeding occurs more often in marginal soils. Therefore, if you specify an early seeding date, the model will also assume that the soil is marginal. This is also true when using seeding date as one of the axes. If you do not wish for the model to infer a value, then you can add it as a control variable in order to hold it constant.
If you are using only the x axis, you can add as many comparisons as you want under Add Comparisons (Optional). Each value entered here will result in a new line on the plot. A baseline will also appear, representing the result using only the constant variables.
The resolution specifies the number of points sampled along the axes. Higher resolutions result in smoother plots but take longer to run.
A Large Yield Model (LYM) predicts the yield performance of crops under any combination of environmental and management conditions. It is trained on a very large amount of data from multiple sources, in order to form a robust and generalizable concept of the factors which influence crop performance.
The models use data from the Daymet weather service, the Saskatchewan Crop Insurance Corporation, and the Canada Land Inventory from Agriculture and Agri-Food Canada. All data is from Saskatchewan, but may generalize to other locations in Canada.
The models have no concept of location. They are never provided with any location or other identifying information during training and, if your farm is represented in the data, it cannot be associated to you or your land.
Axes are shown in unitless values between -0.5 and 0.5 if a weather value is used as an axis (since the weather changes across timepoints), or if no crop type is specified. This is just the range of values present in the data, and zero does not necessarily represent the average as the distribution may be skewed.
