This study used hyperspectral data to determine nitrogen, weed, and water stresses in a corn (Zea mays L.) field in southwestern Quebec, and incorporated these data in crop growth models for better crop growth simulation under stressful conditions. / In 2000, aerial hyperspectral images (72 wavebands, ranging from 407 to 949 nm) were acquired, and analyzed using a stepwise approach to identify wavebands useful in detecting weed and nitrogen stresses. Discriminant analysis (DA) was used to classify different weed and nitrogen treatments and their combinations. This analysis showed greater classification accuracy (nearly 75%) than those obtained with artificial neural networks (58%) or decision tree algorithms (60%), at the initial growth stages, the time when remedial actions are most needed to alleviate weed and nitrogen stresses. / To explore the possibility of improving nitrogen stress detection in corn in the presence of a confounding water stress, ground-based 2151 narrow-waveband reflectance values (350 to 2500 nm), were collected in 2002. Using DA with the chosen subset of narrow-wavebands, a classification accuracy of greater than 95% was obtained. / For crop growth monitoring, the STICS model was evaluated for yield and biomass estimation in cornfields under different stressful growth conditions using the data collected from 2000 to 2002. Measured yield, biomass, and leaf area index (LAI) were used for both calibration and validation of the model. High correlation coefficients between the measured and estimated grain yield (0.96), biomass (0.98), and LAI (0.93) indicated that the model has good potential in the simulation of corn growth. The model was also linked with LAI values estimated from the hyperspectral observations using the Support Vector Machines technique. Coupling STICS with remote sensing resulted in an overall improvement in the simulation of corn yield (6.3%) and biomass (3.7%). / A new approach was developed to apply crop growth models for yield estimation in weedy areas. The proposed method first corrects the measured/estimated LAI values in weed infested fields for weed effect, and then uses the corrected LAI values as input to the crop growth model. The results showed that the crop yield and biomass predictions were correctly simulated by this method.* / *This dissertation is a compound document (contains both a paper copy and a CD as part of the dissertation).
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.85560 |
| Date | January 2005 |
| Creators | Karimi-Zindashty, Yousef |
| Publisher | McGill University |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Format | application/pdf |
| Coverage | Doctor of Philosophy (Department of Bioresource Engineering.) |
| Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
| Relation | alephsysno: 002223557, proquestno: AAINR12867, Theses scanned by McGill Library., Theses scanned by UMI/ProQuest. |
Page generated in 0.0019 seconds