Statistics methods for the assessment of coffee yield losses caused by pests and diseases under different production situations

Costa Rica and Central America, in general, recelty knew epidemic on coffee plant which has an impact on losses of production. The need to qualify and quantify yield losses caused by pests and diseases become urgent. Factors which influence production loss are numerous and their interactions are complex. In the contact of the doctoral research of Rolando Cerda which aimed at assessing coffee plantation yield losses, my objective was to propose statistical methods which permit to explain the influence of production situation on pests and diseases indidence and on yield. A conceptual model was created, to illustrate this link and bring to light factors impacting yield losses of the coffee plant. This study is using data from 69 coffee plots in rural part of the area of Turrialba, Costa Rica where we can find different production situations. A qualitative and quantitative approach was conducted to provide strong understanding of all the relation. A first step was the treatment of data, and creates variables. For the qualitative approach, the use of PLS (Partial Least Squares) approach to select information before clustering was very important. Finally Fisher's exact test and correspondence analysis permit to resume all the interaction between profiles of combination of pests and diseases incidences with typologies such as soil, management, climate, coffee plant production characteristics, etc. It permits to synthesize the impact of production situation on injury profiles, which them self have an impact on yield losses.

Bibliographic Details

Main Author:	Clement, Eugénie
Format:	thesis biblioteca
Language:	eng
Published:	INSA
Subjects:	F01 - Culture des plantes, U10 - Informatique, mathématiques et statistiques, H01 - Protection des végétaux - Considérations générales,
Online Access:	http://agritrop.cirad.fr/577252/ http://agritrop.cirad.fr/577252/1/RapportStage5A_Eug%C3%A9nieCLEMENT.pdf
Tags:	Add Tag No Tags, Be the first to tag this record!