AgroCares proposes products and services to analyse soil samples faster, easier and cheaper without any chemical reagents, based on sensor technology. The purpose of this article is to illustrate how we ensure the quality of our results.
The Calibration Database
All AgroCares services rely on a global calibration database, in which the electromagnetic spectrum of soil samples from all over the world are stored alongside with their wet-chemistry characteristics. When a sensor measurement comes in from the Scanner or the Lab-in-a-box, it is converted into wet-chemistry parameters using this calibration database. Therefore, to measure the accuracy of the sensor measurements what is really required is to test the quality of the calibration database.
To expand and strengthen the calibration database, new samples are added each month and its quality is thoroughly tested on a regular basis by our team of experts.
The term ‘quality’ is not limited to a single definition. At AgroCares we therefore use different indicators and statistical parameters to reflect the different aspects of quality of our measurements and predictions.
The deviation in measurement values when a single sample is repeatedly measured.
The variation coefficient is the relative standard deviation and is used to express the repeatability of an assay. A low VC indicates high precision and repeatability. AgroCares strives to a VC of 10 or less to secure sufficient precision.
The correlation between the results of sensor and wet-chemistry measurements of the same samples.
The R² reflects the correlation between wet-chemistry and sensor measurements over a large number of samples in the database. A high R² indicates good correlation. AgroCares aims at a R² of 0.7 or higher.
The expected impact of the fertiliser recommendation give on current yield levels.
The prediction error is the expected difference between the predicted value and the true value. It reflects the chance the measurement results in its correct classification, used to determine the fertilizer recommendation.
Dry chemistry vs. Wet chemistry
Quality is about more than just comparing a soil status predicted with sensor technology (also called dry chemistry) to those obtained with wet chemistry. For example, the process before the soil sample is actually measured is equally important and wet-chemistry has its own challenges that affect the accuracy and precision of the measurement (human factor, work conditions, differences amongst labs...)
At AgroCares, we therefore consider that the quality of the results is the sum of four subsequent elements:
The quality aspects cannot be easily summed. It depends on factors such as the variability of the soil, the transport distance and the quality of the wet-chemistry lab facility, etc that determine the overall quality of the analysis. With dry chemistry, the first three steps outperform traditional wet chemistry methods: the samples are analysed on the spot and can therefore nor be altered, because of its low price, more samples can be collected and analysed giving a better overview, the results are obtained using mathematical predictions that involve no human factor or reagents giving less room to mistake.
Lab-in-a-box vs. Scanner
The main difference between the Lab-in-a-Box and the Scanner lies in the sensor used in the equipment. The Lab-in-a-box contains a MID infrared spectrometer and an XRF allowing it to produce detailed spectra from which a lot of information can be extracted easily. The Scanner is composed of a NIR Infrared spectrometer, cheaper and easier to transport, but producing spectra with less detailed and informative information. For this reason, the Lab-in-a-box can measure more soil parameters and has a better performance compared to the Scanner.