Methods and systems for plant performance analysis

Details Methods and systems for plant performance analysis Methods and systems for plant performance analysis

2000-10-14

2003-12-09

Corrielus, Jean M. (Department: 2172)

Data processing: database and file management or data structures

Database design

Data structure types

C707S793000, C707S793000, C707S793000, C702S002000, C702S003000, C702S022000, C702S030000

Reexamination Certificate

active

06662185

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a process of assessing the performance of crop varieties based on wide-area performance testing data. A process of the present invention compares varietal performance using spatial estimation and spatial prediction based on a statistical mixed effects model.
2. Description of Related Art
In the development of a new crop variety, performance data are collected on the variety and on other competing varieties. These performance data include measurements on various agronomic traits relevant to the given crop; e.g., for
Zea mays
, measurements taken on grain yield, grain moisture, and plant lodging.
In assessing the potential commercial value of a new crop variety (hereafter referred to as “variety”), its agronomic performance is compared to the agronomic performance of other varieties. The other comparison varieties include commercial and pre-commercial varieties from the company developing the variety and commercial varieties from competitor companies. Note that this same type of assessment is also performed on existing commercial varieties, to determine if they should remain on the market or be replaced by newer varieties in development.
Agronomic performance data for the new variety and for the comparison varieties come from multiple testing locations. The testing locations are usually widely distributed over the area of adaptation of the varieties included in the test. The area of adaptation covered by these testing locations is typically quite large, on the scale of hundreds of square miles. For example, a new
Zea mays
cultivar may be tested from western Iowa to eastern Michigan and from central Wisconsin to southern Illinois.
Due to variation in testing programs, the data for a given variety and its competitors tend to be quite ‘unbalanced’ in the sense that not all of the given set of varieties appear at all testing locations. Considering the testing data for a single pair of varieties, i.e., the new variety and a single competitor, some of the testing locations will have both of the varieties, while the rest will have only one of the two varieties.
These performance data are analyzed in order to determine the geographic regions over which the new variety has large enough performance advantages relative to the comparison varieties to justify its introduction to the market in those regions. Ideally, the variety under consideration will have a significant performance advantage relative to all of the comparison varieties over its entire area of testing. However, in some cases a variety may have performance advantages only on a regional basis, but it could still serve a significant market need within that region. Thus, it is important to characterize the performance of the given variety relative to other comparison varieties not only over the entire area where it was tested, but also within the various regions.
The variations in relative performance (performance difference) of two varieties in different geographic locations or regions arise from what is referred to as ‘genotype by environment interaction’ (Sprague & Eberhart, 1976). Genotype by environment interaction is caused by differential responses of varieties to environmental conditions. These environmental conditions may include, for example, day length, temperature, soil moisture, disease and insect pressure. Note that the term ‘environments’ can refer to different locations in a given year or different years for a given location or some combination of locations and years.
Methods involving traditional statistical analyses for varietal performance assessment are described in Bradley, et al (1988). These traditional methods are usually based on “location-matched” data, i.e., for a given variety relative to a comparison variety, data from only the testing locations where both varieties co-occur are used in the analysis. A paired t-test is used to test the hypothesis of no difference in performance of the two varieties. Moreover, for inferences regarding relative performance in a given geographic region, data from the testing locations only within that region are used in the analysis.
Traditional analysis using the t-test for location-matched data is inefficient for at least five reasons. First, it does not use all of the data; it only uses data from testing locations where both of the varieties co-occur. Second, for regional performance comparisons, it does not use data from nearby areas outside the region of interest. Third, it does not make use of covariates related to the performance trait of interest (e.g., irrigation or soil series) to help explain and predict the differences. Fourth, it only uses within-year data in the analysis model; more robust inferences can be accomplished by having a model that uses data from multiple years. Fifth, it is based on an incorrect assumption that the observations from one testing location are independent of those from other locations.
There are two broad reasons why the inefficiencies, listed above, are limitations in wide-area crop assessment. First is efficient use of the data. It is natural for the experimenter to want to use all of the data when making inferences. In the present scenario, this includes data on a variety at a location where the other variety does not occur, it includes data from areas that are in proximity to the region of interest, it includes data on covariates, and it includes data across years. Statistical methods should strive to make full use of all available information. The second reason that the traditional analysis is limited is that it is based on the classical assumption that data are independent Varietal performance data almost invariably violate the assumption of independence and render the statistical inference invalid, typically causing one to infer that the variety differences exist when the data do not really justify it.
Review of Current Literature
A brief review of current literature also highlights underlying deficiencies in the art, which the present invention strives to solve.
One of the most relevant papers and probably the most noteworthy work in the area of application of spatial statistics to field plot experiment is by Zimmerman and Harville (1991), where the authors have introduced the so-called random field linear model (RFLM) by considering the observations as realization of a random field. In this model the trend is modeled by a mean structure and the small scale dependence is modeled through spatial autocorrelation structures. The parameter estimation is done through a likelihood approach. Through real data analysis, the authors have tried to demonstrate the superiority of their model over nearest neighbor analysis (NNA) approach. Note that their study is exclusively in the context of small area estimation where the range of spatial dependence is confined to a testing location.
Another noteworthy paper in the context of the use of covariate in spatial prediction is by Gotway and Hartford (1996), where the authors have presented the use of auxiliary or secondary variable(s) in spatial prediction by applying cokriging to predict soil nitrate level with data on grain yield as a covariate. Through an application of their method to data from a test site, they have demonstrated the benefit of their method over the more traditional external drift method. Again the scope of their study is limited to intra-site prediction.
One of the recent papers that deal with multi-location yield trials is by Cullis et.al. (1998). In this paper the authors have proposed a method for spatial analysis for multi-environment early generation variety trials. The method uses best linear unbiased prediction (BLUP) for genotype effect and genotype by environment interaction effect and REML for the spatial parameters and variance components. However, the proposed method is based on separately modeling the covariance structure for each trial, i.e., no across-trial correlation is taken into consideration.
Yost, Uehara and Fox (1982a and 1982b) were one of the first researc

Affiliated with

Also associated with

Rating

Methods and systems for plant performance analysis does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Methods and systems for plant performance analysis, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Methods and systems for plant performance analysis will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3178804