GEOSTATISTICAL ANALYSES OF SOME AGRONOMICAL OBSERVATIONS

The variability of agronomical measurements in either space or time is the primary concern of this study. The following three chapters contain analyses of different data sets using similar approaches.;Chapter 2 contains the analysis of the variability of some bare surface measurements over time. As measurements were obtained during the winter growing season, which extends through late spring, a strong seasonal drift was found in the measurements. Residuals were calculated by subtracting each measurement from a 29 point moving average. The residuals of air and surface temperature were correlated for approximately 6-10 days, and cross correlation between air and surface temperatures were almost identical for the three years under study.;Chapter 3 contains the application of different tools to find autocorrelation of surface temperature measurements taken along transects during redistribution of water. It was concluded that when estimation is not of concern, the autocorrelation is better than semivariogram in characterizing the length of autocorrelation and comparing for different environmental conditions. This is mainly due to the normalization of the autocorrelation from -1 to +1. However, the second-order stationarity required for the autocorrelation may not be assumed if drift is present. The semivariogram requires weaker assumptions.;It was shown that many advantages can be taken out of the cross correlation between variables obtained under identical conditions, but these cross correlations will need to be the subject of research for more years to verify the results found.;Chapter 1 contains detailed derivation of the kriging and cokriging systems for punctual samples in one- or two-dimensions as well as other aspects involved in geostatistical estimation, such as the estimation error or the estimation neighborhood. The use of these theoretical developments is illustrated with examples using 23 variables from different agronomical fields and a variety of sampling schemes. All computer programs used are listed in the appendix. In general, it is shown that the continuity of data acquisition over many years is essential for the knowledge of semivariograms and cross semivariograms as a variability measure of a given environmental condition.