Prediction From Archived Vegetation and Environmental Monitoring Data

David Scott

doi:10.5539/jps.v3n2p91

Abstract

A comparison was made between analysis methods of archived monitoring data for extrapolating quantitative values or trends from additional partial measurements. These were done from both the ‘species from environment and management’ or ‘environment and management from species’ predictions. The sample data used was from two 19 year annually monitored grazing trials in a tussock grassland environment under 90 different controlled inputs, but treated as if from a random ecological survey. Quantitative estimation of trends from such a reference data set was best by a direct data search for closest matching observation, followed by response surface regression, simple multiple regression, multiple imputation, transformation regression, and least from principal component analysis.

Highlights

In recent decades there has been increasing interest in establishing the medium and long-term sustainability of different land-use practices and the desirability of establishing monitoring systems so that effects and trends can be quantified
A comparison was made between analysis methods of archived monitoring data for extrapolating quantitative values or trends from additional partial measurements
Quantification of trends is dependent on both the methods of initial monitoring of vegetation and environment, and analysis methods used to determine the relationships within those data

Summary

Introduction

In recent decades there has been increasing interest in establishing the medium and long-term sustainability of different land-use practices and the desirability of establishing monitoring systems so that effects and trends can be quantified. This has been true for conservation management and rangeland pastoral farming systems. Quantification of trends is dependent on both the methods of initial monitoring of vegetation and environment, and analysis methods used to determine the relationships within those data. Only one method of analysis has been used in each investigation and there has been little comparison of the effectiveness of different methods on the same data set

Objectives

Methods

Results

Conclusion