A critical look at procedures for validating growth and yield models.

Abstract

This chapter reviews the general procedures and methodologies used for validating growth and yield models. More specifically, it addresses: (i) the optimism principle and model validation; (ii) model validation procedures, problems and potential areas of needed research; (iii) data considerations and data-splitting schemes in model validation; and (iv) operational thresholds for accepting or rejecting a model. The roles of visual or graphical validation, dynamic validation, as well as statistical and biological validations are discussed in more detail. The emphasis in this chapter is placed on the understanding of the validation process rather than the validation of a specific model. The limitations and the pitfalls of model validation procedures, as well as some of the frequent misuses of these procedures are discussed. Several technical and practical recommendations concerning the validation of growth and yield model are made.

Systematic Evaluation of Kriging and Inverse Distance Weighting Methods for Spatial Analysis of Soil Bulk Density.

“…The MPE values from kriging and IDW methods to 95% confidence intervals for observed mean were: -4% and -6% for 1999; and -18% and -15% for 2000, respectively. Using the first criterion suggested by Huang et al (2003), MPE is less than ±10% of the observed mean at 95% confidence intervals; therefore, both methods are acceptable for interpolation for 1999. For 2000, both methods are not acceptable because the MPE from both models is greater than ±10% of the observed mean at 95% confidence intervals.…”

“…Average systematic errors were lower while unsystematic errors were higher again proving the superiority of the IDW model. Mean prediction errors to observed mean Values of MPE to the 95% confidence intervals for observed mean from kriging and IDW methods for bulk density were calculated using approach suggested by Huang et al (2003). The MPE values from kriging and IDW methods to 95% confidence intervals for observed mean were: -4% and -6% for 1999; and -18% and -15% for 2000, respectively.…”

“…A few validation indices include cross validation (Carr 2002), mean prediction error (Vos et al 2005;Huang et al 2003), mean absolute error, (Brounder et al 2005;Naoum and Tsanis 2004), mean squared prediction error, (Schloeder et al 2001;Voltz and Webster 1990), and root-mean-squared prediction error (Vos et al 2005;Bishop and McBratney 2001). Besides commonly used scatter-plot, cross validation and coefficient of determination, many accuracy, precision, and consistency indices have been reported in the literature.…”

“…Willmott (1981;1984) partitioned total errors (RMSE2) into average "systematic" errors (RMSEs) and "unsystematic" errors (RMSEu). Huang et al (2003) tested models based on the percentage of mean prediction errors (MPE) to the observed mean; however, there is no consistent rule of thumb to assess if prediction models' results are accurate enough to be acceptable. For example, Freese (1960) took an allowable error of ±7%.…”

“…In Alberta, a ±10% prediction error at 95% of confidence interval was used in several studies. Huang et al (2003) reported that most researchers used thresholds for model acceptance between 10 and 20% of the observed mean at 95% of confidence interval. For instance, Prevost (2004) accepted a relative prediction error (SDPE2/MSPE) of 20% for bulk density.…”