Soil Calibration Mendoza
Description
Create a local soil calibration for Mendoza soils. About 30 soils were collected from farms in the area, and they were sampled on 3 devices. These devices are using the PETG calibration rather than the BaSO4 calibration. If we are going to use any previous data or larger models from other devices we would need to adjust the data between the calibration types.
The in-lab measurement which has been performed here is using Walkley Black. Generally, the soil carbon levels locally are very low (around .5%) and it's possible that our model, especially with so few samples, won't be accurate enough to distinguish between locally high and low examples.
As such, we're going to just create a local calibration and see how accurate a model we can predict. Ideally, after creating the model, we could test it on some additional soil types.
Data
The location of the data https://app.surveystack.io/surveys/60e8938270e81500018811d6. Items are marked 'regosh' or 'soilab'
Samples, Equipment, Processes
- 3 Our Sci Reflectometers with PETG calibrations
- 29 bags of soil w/ some management / related data.
- Walkley Black testing (available via local University Lab)
Outcomes
- Create an 'intake' survey to populate the soil carbon model
- Run samples + generate a model
- Create a 'prediction' survey to generate carbon values using the model.
Tasks
-
(nano) check local soil lab to see if they have additional soils we can either add to the model, or test against the model to confirm how accurate it is. -
(greg) create a survey for uploading the soil data. It's possible we could use an existing survey as is used in the lab... - We have an existing survey for general soil organic carbon estimation projects through Our Sci here: https://app.surveystack.io/surveys/60e8938270e81500018811d6. For simplicity in this example, we'll use the existing survey.
-
(greg) run the samples (calibrate one time at the beginning of the process) -
(octavio) generate a model (decide on the model type) and present the results in the 'prediction' survey -
(greg) identify additional local samples (possibly from the local soil lab) which can be run to test the accuracy of the model. -
(greg/octavio) display the results in the final reGOSH event.