Registering the exact GPS coordinated of your field is important because it can influence the accuracy of your results but let us explain to you why.
Do you know what exactly a GPS coordinate is? GPS is an abbreviation of the words “global positioning systems” and GPS coordinates are a unique identifier of a precise geographic location on the earth.
AgroCares Scanners rely on a near-infrared sensor and a connection to AgroCares unique soil database to determine the chemical composition of soils, leaf, etc. The soil database is constituted by calibration samples which are a combination of wet chemistry analysis on representative soil samples of a specific target area, including the analysis of Lab in a Box and Scanner.
Thanks to empirical models, AgroCares is able to produce powerful deep learning prediction models. These prediction models get more accurate every time we add new calibration samples to our database.
To produce even more accurate results our deep learning models are fed with additional information deriving from covariates. These covariates include information from soil grids that are location based (e.g., opensource ISRIC tables).
Meaning, to produce the most accurate results it is crucial that that the GPS location is added correctly as the combination of analysis & location-based reference information can improve the overall analysis results provided.
What if the samples are brought to a general collection point and are not being analyzed in the field directly?
In the case that the samples are brought by the farmers to a central location and no specific GPS location is known, it is important to choose the closest knows point e.g., village and register this point as field location instead. You can select the location of a field through the app following the those 2 steps. See Picture 1 and 2 below.
How inaccurate is the Scanner measurement when the GPS location is displaced?
The level of inaccuracy depends on many factors, such as: displacement distance, country, change in terrain, specific parameters but an estimated RMSE of approximately 5% at average.
If you want to learn more about our deep learning models, please have a look at this article: