Integrated yield forecasting and harvest scheduling in a tropical pine plantation in Fiji
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
This thesis reports on enhancements of two planning method components aimed at improving management and planning of forest plantations in the tropics. The two modular planning models subjected to detailed study are growth and yield modelling and harvest scheduling. A case study relating to Caribbean pine in Fiji is used to demonstrate the refined capability. Growth and yield modelling has been improved by applying modern statistical and computer techniques to solve non-linear equations that describe growth of stands appropriately. Further improvements have been achieved by developing diameter distribution growth and yield models solved by a combination of parameter recovery and prediction method thereby ensuring compatibility between average stand values and diameter distribution values. In conducting improvements in growth and yield models, data manipulation and data validation procedures are described and reviewed in detail to emphasize their importance, particularly for non-linear regression fitting of equations, growth, yield and diameter distribution projection modelling. Various growth projection equations were tested before final stand average functional forms for basal area per hectare, standard deviation of diameter at breast height outside bark, maximum diameter at breast height outside bark and survival per hectare were identified and then integrated into the growth and yield model. The precision of the equations was assessed through graphs and statistics relating to residuals. The stand simultaneous growth and yield equations solved and used in the model consist of modified forms of different growth projection functions, such as the Gompertz, exponential and Schumacher, which were then used to derive a diameter distribution based on the Reverse Weibull probability density function. The diameter distribution growth and yield model was prepared as a simulation model to predict stand average values then, in conjunction with existing stem volume and taper equations, to derive stand and stock tables that allow disaggregation of diameter classes into log types. Three simulation models were created, one in Vax Fortran, one in PC Fortran and the other in spreadsheet format to enhance the models's portability. The harvest scheduling model developed is a spreadsheet based LP model which is able to schedule harvests from a number of stands within a medium-term planning horizon using different logging methods with the log harvest to be delivered to different ports or utilization plants. A Fiji case study provided a demonstration of the modelling capability for fifteen stands, seven years, four logging methods and two ports. This new kind of LP harvest scheduling model was developed with a deliberate intention to facilitate the running of it with the input from the improved growth and yield model. In developing this harvest scheduling model, the nature of LP in general was first reviewed and compared to other tools of harvest scheduling like binary search and simulation. LP harvest scheduling was found in this review to be a widely used tool and solution algorithms for which abound. A major problem with most solutions was the need to cater for sophisticated report writing and matrix generation. These two concerns were specifically addressed in the model developed as part of this study. The use of a spreadsheet as input to the LP was seen to be an efficient way of overcoming some of the major criticisms levelled at LP by potential users. The methodology developed was also advantageous because of its capability to facilitate the integration of growth and yield outputs with harvest scheduling. It was concluded that forest planning models can be readily improved with software and hardware that developing countries can easily afford. The models reported here harness the capabilities of the now commonly employed spreadsheet as a powerful tool for easier routine input, output and sensitivity analysis, to assist decision making for harvest scheduling and to simplify managerial planning and control.