Investigation of the potential for precision soil and crop growth mapping to improve potato (Solanum tuberosum L.) tuber size distribution at harvest
Mhango, J.K. (2022) Investigation of the potential for precision soil and crop growth mapping to improve potato (Solanum tuberosum L.) tuber size distribution at harvest. Doctoral thesis, Harper Adams University.
|
Text
Kanthu Mhango PhD upload.pdf Download (6MB) | Preview |
Abstract
Control of tuber size distribution (TSD) in potatoes (Solanum tuberosum L) is desired for farmers seeking to maximize profit in a market environment that is sensitive to tuber size. The TSD and its spatial variability are related to stem density variation. Throughput improvements in the methods of quantifying stem density will unlock adoption of more precise methods of managing TSD. Understanding the variability of soil nutrients and their effects on TSD can also help in the delineation of management zones for precision applications like variable rate fertilization. In this study, a method for quantifying TSD based on the Weibull distribution was proposed, with consistently lower Root Mean Square Error than currently prevalent methods. With this method, negative relationships between TSD and excess soil nutrients were uncovered. In above-ground canopy studies, a novel potato stem detector was developed using deep convolutional neural network (CNN) and aerial imagery. Novel colour indices were also developed for elucidating the locations of potato stems from aerial imagery. For the first time, this study demonstrated the potential to map stem density (a key determinant of TSD) in a field using high throughput methods. The potential of satellite image time series in modelling stem density and yield was also examined. Sentinel-2 satellite data was used to create spectral signatures of potato plants and their temporal evolution. Features engineered from this data were able to model potato Marketable yield and stem density. Temporal evolution of specific wavelengths (e.g. 559nm), integrated with manually determined stem density, was found to have highly significant relationships with marketable yield. As a conclusion, the study uncovered high potential for crop growth mapping to predict TSD and aid in decision-support systems. Furthermore, the study proposed a unitless Weibull shape parameter as a means of quantifying TSD to enable inter-study comparisons in TSD work.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Divisions: | Agriculture and Environment (from 1.08.20) |
| Depositing User: | Ms Kath Osborn |
| Date Deposited: | 15 Mar 2022 14:42 |
| Last Modified: | 15 Mar 2022 14:42 |
| URI: | https://hau.repository.guildhe.ac.uk/id/eprint/17820 |
Actions (login required)
 |
Edit Item |