StanForD as a data source for forest management: a forest stand reconciliation implementation case study

Abstract

The New Zealand forest industry is in a state of change from motor-manual chainsaw processing towards fully mechanised harvesting operations. This is driven predominately by changes in the health and safety legislation and increased efficiency targets. Through the use of advance harvesting machinery with built in computer systems and standardised compatible data collection software (called StanForD), all mechanised processing operations are able to produce near real-time production data. This data stream enables forest management to work with datasets containing detailed information of all harvesting production. StanForD data will therefore enable the development of new ways of forest management. The study objective was to research the use of StanForD data in a forest stand reconciliation scenario. StanForD production volumes were compared against a weight docketing system and inventory yield predictions on four harvesting sites. These studies were conducted in a clearfell harvesting crew with an experienced harvester operator over the duration of approximately one year. The data collection included all relevant production files from the harvester; .PRI (production data), APT (harvester cutting instruction) and KTR (harvester head calibration data) files. The forest management company supplied load delivery dockets, conversion factors and inventory data. The inventory data was processed to estimate the yields of the harvested stands. PLE1 (p ≤ 0.05) boundaries by grade group and total volume were calculated. The estimated yields with its PLE boundaries were compared against the volume recorded by the harvester and the data retrieved via the docketing system. The results show the harvester data, when compared with the inventory data, was within the PLE limits for seven out of 15 grade groups. Small utility was the only grade correctly predicted at all sites. Pulp wood hasn’t been predicted correctly at any site in comparison to the harvester data. The docket data was for five out of 15 grade groups within the PLE limits. For the total volume the harvester data was two out three sites within the PLE limits. The docket data failed on all three sites to be within the PLE boundaries on total volume. These results show both reconciliation methods, docketing system and harvester data based system have failed to confirm the yield predictions repetitively. Comparison of the harvester data against the docketing data, showed the harvester had lower recorded volumes for pulp, export pulp and an systematic over-measurement for the higher grades compared to the docket data at all sites. Subsequent to data collection, the reason for the lower harvester volume measurements on the lower quality grades was identified to be operators not correctly recording harvest data. As possible causes for the over measurement, missing bark function and the use of estate wide conversion factors were identified. The study showed higher grades, despite the schematic differences, were recorded more accurately than lower quality grades. Taking all results in account, using harvester data remains a valuable data source for the future; especially for aspects such as reconciliation. More emphasis on operator training on the harvester computers systems is likely to increase the data quality collected by the harvester.