MSc Thesis topic: Sample locations for calibration of a near infrared soil scanner
AgroCares is an agritech company based in Wageningen. The main product of AgroCares is a Near Infrared soil scanner that can be used to scan soils and predict soil properties. The AgroCares scanner is calibrated with soil samples that are collected in the areas where the scanner is used: In these areas, a number of soil samples are collected and transported to the laboratory of AgroCares. Here, the soil samples are scanned with the scanner and analyzed for several soil properties such as, for example, pH, organic matter, soil carbon content, N, P, K. The combination of the scans, together with the laboratoy measurements are used to create prediction models. With these models, clients of AgroCares can use the scanner in the areas where the calibration samples have been collected. At the moment, the calibration database of AgroCares consist of approximately 20,000 samples, collected in 30 different countries around the world (see Fig.).
Currently, the scanner can only be used in areas where calibration points have been collected. Based on the current calibration database, AgroCares want to make an inventory of where else, besides the country where the calibration samples have been collected, the scanner could potentially be used, by following, for example, the homosoil approach (Mallavan et al., 2010) which assumes homology of soil-forming factors between a reference area and a area of interest, or the work of Meyer and Pebesma (2021), who define an Area of Applicability (AOA) where a model trained with data from a specific refence area can be applied to other similar areas. Moreover, AgroCares want to identify areas that are very different from the current database and could therefore contribute to the increase of the area where the scanner can be used.
In this thesis, you will work on (1) evaluating where the AgroCares soil scanner currently can be used and (2) exploring sampling approaches for collecting reference data that will increase the potential area that where the scanner can be used. The ultimate question that AgroCares hopes to get answered is: how many calibration samples need to be collected – and where – before the scanner can be used anywhere in the world with further calibration?
- Identify geographical regions that are (not) well covered by the current ArgoCares calibration database.
- Determining soil variability based on spatial data.
- Developing a sampling design to capture the soil variability with the least amount of calibration points.
- Mallavan, B., Minasny, B., and Mcbratney, A. (2010). Homosoil, a Methodology for Quantitative Extrapolation of Soil Information Across the Globe, in: Digital Soil Mapping, Progress in Soil Science, edited by: Boettinger, J. L., Howell, D. W., Moore, A. C., Hartemink, A. E., and Kienast-Brown, S., Springer, Dordrecht, The Netherlands, 137–150.
- Meyer, H. and Pebesma, E. (2021). Predicting into unknown space? Estimating the area of applicability of spatial prediction models. Methods in Ecology and Evolution, Vol 12 (9), pp. 1620-1633.
- We look for a motivated student with strong analytical skills and research interests in (spatial) statistics and modelling.
- Student should have programming skills is R or Python and should preferably have affinity in working with large spatial datasets.
- Ideally, the student should have followed the course Spatial Modelling & Statistics.
Theme(s): Sensing & measuring; Modelling & visualisation