TY  - BK
AU  - Alder,D.
TI  - Growth modelling for mixed tropical forests
T2  - Tropical Forestry Papers
SN  - 0-85074-135-1
SN  - 0141-9668
U1  - 578.73 ALD 
PY  - 1995///
CY  - Oxford (United Kingdom)
PB  - Oxford Forestry Institute
KW  - AGROVOC
KW  - Tropical forests
KW  - Forestry
KW  - Forest conservation
KW  - Forest growth
KW  - Forest genetic resources
KW  - Forest ecosystems
N2  - This manual is directed at the analysis of permanent sample plot (PSP) data from mixed tropical forests to produce growth and yield models. It assumes some knowledge of matrix algebra, statistical methods, and computer programming. Its emphasis is on the practical technique of data analysis and model building. It is divided into five sections. The first covers definition of modelling terms and selection of strategy. It recommends a deterministic empirical strategy as most appropriate forforestplanningapplications fromconventional PSPsand Inventory data, and describes two basic approaches: Diameter class projection and cohort modelling. The second section covers thebasic analysis ofPSP data to the pointofcalculating increments and competition indices. using XBASE examples, methods are described for data entry, error checking, listing data, drawing plot maps, converting incompatible measurements to a common basis, merging mUlti-year data sets, and transformation of year-per-record and mUlti-year record data formats. The compilation of stand-level statistics, increment, and competition indices in a record format suitable for regression analysis Is described. Detailed competition index calculation methods are given for spatial and non-spatial measures, including competition influence overlap and overtopplng basal area. Reduction of large data sets by cross-tabulation is exemplified. The third section covers diameter class projection In both Its classical and matrix algebra formulations. The homologies between usher, Markov and other matrix model formulations are explored. The use and relevance ofthe de Llocourt ratio is discussed. Spreadsheet examples (using Lotus 1-2-3) are given of classical stand projection and matrix models. ABASIC program forstand projection Is compared With thespreadsheet examples. The transformation of data structures from raw PSP data to transition matriX are shown. The weaknesses of diameter class prOjection methodsare discussed, including insensitivityto stand density, and the assumption of uniform increments within a class. The GHAFOSIM program is described as a case studyincluding automatic updating oftransition matrices and initial stand tables from PSP data and Inventory plots, and the refinement ofthe method to include different crown classes and stand density Interactions. Diameter class methods are considered suitable for aggregated projections as initial approximations in the early stages of forest management or for a broad sectoral overview of potential yield. The fourth section covers cohort modelling methods and the derivation of increment, mortality, and recrUitment functions. cohort modelling Is considered to be the optimal strategy for forest planning applications. An outline structure of a cohort model is discussed and exemplified using the author'S CAFOGROM model for Brazil. The analysis of diameter Increment Is described. Log transformation of Increment data Is demonstrated graphically and considered appropriate. Basal area increment versus diameter Increment are explored, and several possible robust equations Indicated. The methods of Including competition Index are shown, and the problem of the weak residual covarlance with competition once diameter has been accounted for discussed. Data sets from forests of limited treatment range may not permit the evolution of adequate competition models. Site effects are treated as categorical variables, and methods of analysis exemplified using general linear models. Mortality calculation and adjustment for interval measurement period are demonstrated using compounding of the survival rate. An interactive XBASE dialog to calculate mortality rates from PSP data is given. Mortality is a binomial proportion, and treatment of confidence Intervals is described. The logistic transformation is appropriate for analysis. Methods of comparing species differences in mortality rates are shown. Recruitment and regeneration are defined. usually only recruitment data is available; regeneration modelling is more complex and requires specialized data. calculation from raw PSP data is demonstrated. prediction of recrUitment is recommended as atwo-part process. The Quantity of recruits can be related to alevel of stand disturbance or basal area loss in a previous time period, with the lag depending on defined recrUitment diameter. The distribution of recruits by species is best not handled by regression due to problems of additIvity and covariance, but can be simply managed by creating look-up tables of species proportions for different disturbance classes. The simulation of harvesting, logging damage, and silvicultural treatment is discussed, with empirical methods and examples for estimating logging damage functions in relation to logging Intensity. A case study of specific increment, mortality, recrUitment, and logging damage functions is presented using the CAFOGROM model. The final part of the manual deals with model validation. procedures for residual analysis of individual functions are discussed, with examples of common pathologles cheteroscadacity, bias, lack-of-fitl. Residual analysis applied to complete models is discussed. Methods for testing models by analysing their behaviour in relation to expected forest dynamics and yields are considered and exemplified usIng CAFOGROM. The application of a model as part of a forest management Information system is described briefly, including linkages to data from temporary inventory plots, a species database, and forest stand locations and previous treatment history
ER  -