• News

Old meets new … Enhancing traditional growth and yield forecasts with the latest tech innovations

A team of researchers has developed a prototype system that combines remote sensing data and process-based modelling with the traditional method of empirical modelling. The system will help forest managers improve planning and forecasting systems for the optimal maintenance and development of their forest growth and yield.

The team has also developed a series of recommendations for industry to enable individual forestry companies to adopt and fully operationalise this new hybrid approach to modelling within the next five years.

“Australian plantation forest managers face ongoing challenges around maintaining and developing their forest growth and yield planning and forecasting systems,” research leader Dr Philip Smethurst said.

Dr Smethurst is a soil and water scientist, and plant nutritionist at CSIRO Land and Water.

Remote sensing and process-based modelling

There are many opportunities (and complexities) associated with new data sources for the prediction of current and future attributes of plantation resources. This data, which will guide forest management practices, can be sourced using the likes of satellite and airborne or ground-based sensors, as well as process-based modelling.

Process-based modelling in this context refers to resource forecast modelling that includes the specific mathematical representation of ecosystem processes that lead to wood production, e.g. the use of light, water and nitrogen.

Traditional empirical modelling that relies on simple statistically-based correlations tends to lose reliability where forecasts are required to consider future conditions not yet felt. While the underlying physiological relationships between CO2, temperature and rainfall in the plant might remain the same, predictions resulting from unexperienced conditions can be quite uncertain when using these traditional models. As climate variables change, so too must growth and model predictions.

Process-based modelling has the advantage of relying on specific representations of biophysical processes that are maintained while new conditions are experienced. It offers a more reliable method of yield prediction under a changing climate.

“Taking full advantages of new opportunities such as process-based modelling and remote sensing technologies often requires specialist skills not always available within forest companies,” Dr Smethurst said.

“As a result, many companies rely on a combination of legacy technologies and systems, alongside various modern additions, limiting the potential of their capabilities.”

The research

In 2020, the two-year FWPA-supported project Next Generation Resource Assessment and Forecasting for Australian Plantation Forestry began to help address these concerns.

This work aimed to explore industry needs in this field and evaluate, investigate and develop these technologies to create a prototype system for resource assessment and forecasting that included process-based modelling and remote sensing.

“This research set out to offer the potential for growers to design and experience elements of a modern system akin to those available and successfully applied by grain growers and in other agricultural sectors,” Dr Smethurst said.

The goal was for the new system to enhance, rather than replace, existing growth and yield forecasting systems by incorporating process-based modelling capabilities and data from sensors, alongside traditional empirical modelling.

The project also aimed to evaluate and develop cooperative business models that could sustain delivery of these advanced technical services to the forest plantation industry.

The project demonstrated elements of a proposed workflow for merging remotely sensed data with process-based modelling, and current inventory and empirical modelling.

The Agricultural Production Systems sIMulator (APSIM) modelling framework – well-established in the agricultural sector nationally and internationally for research and commercial uses – was chosen as the process-based modelling framework for this project.

Consultants and researchers completed basic training in the use of APSIM for plantation forestry. APSIM simulates plant growth at the plot scale, and before this project it had been calibrated for Eucalyptus grandis plantations and related tropical and sub-tropical genotypes.

Model development added genotypes required for temperate eucalypt plantations in Australia (E. globulus and E. nitens) and pines in tropical-to-temperate regions (Pinus radiata, P. elliottii, P. caribaea and hybrids).

These models use climate variables as inputs as well as management variables, including initial stocking, thinning, mortality, weeds, nitrogen fertilisation and irrigation. New genotypes can also be calibrated. These models are freely available for public-good use and licensing for commercial use is also readily available.

For current inventory at a specific location, historical empirical modelling can be used until a user wants to include process-based modelling, or when new data becomes available via sensors, such as average tree height, stocking or leaf area index, leaf greenness, water use, and/or soil water content.

Hindcasting – testing the accuracy of computer models by comparing them with actual historical observations – can then be conducted. From here, forecasting can be used that considers future climate or management scenarios.

In addition, by combining remote sensing and process-based modelling, hindcasting and forecasting can be conducted on virtual plots established anywhere on a plantation estate.

The team found that process-based modelling would benefit from additional plantation studies to help quantify growth and biomass allocation patterns for particular genotypes in relation to contrasting growing conditions.

“Remote sensing options can provide extra useful sources of data to complement process-based modelling, with many companies already implementing airborne LiDAR scanning programs across their estates. These scans can provide stand metrics for direct input to the overall modelling,” Dr Smethurst explained.

“Major changes to leaf area in a plantation can be caused by, for example, insect browsing, drought or fire. It would be useful to know when and where this is occurring across a plantation estate without resorting to labour-intensive field inspections, followed by the incorporation of the outcomes of those inspections into yield forecasting, and alerting managers of any required actions on the ground.”

The tool was developed and tested predominantly in pine plantations in south-east Queensland and the Green Triangle region of South Australia and Victoria. It was found to be useful in detecting monthly and seasonal atypical changes.

The symptomatic tool was tested at desktop level, migrated to a cloud computing environment and is currently at the prototype stage.

A collaborative approach for easy and affordable industry access

While in-house expertise or consultancy services are theoretically the best options for providing advanced technical services to the industry, the results of an industry survey by the team indicated an overwhelming belief that individual companies do not have the range of technical expertise required and that consulting services can be expensive and lack technical flexibility.

“The project explored opportunities for an industry cooperative model that could potentially provide such services with more technical flexibility, and at a reasonable cost,” Dr Smethurst said.

Tree Breeding Australia was profiled as an example of a long-standing model for delivering advanced technical services to the Australian forestry plantation sector.

The main elements of a similar business model were explored for providing remote sensing and process-based modelling services. These elements included governing documents, membership codes of conduct, member exit mechanisms and strategies, risk management strategies, and a strategy around management of intellectual property.

As a result of this work, the research team strongly recommends that the industry: ·

  • further explores the technical aspects of integrating remote sensing and process-based modelling into workflows for forecasting short-to-long-term wood supply, and inputs to natural capital accounting
  • rigorously evaluates the developed low-cost prototype tool and compares its cost and effectiveness with the costs of existing paid-for services from higher-resolution satellites
  • forms a working group to fully explore and develop a collaborative business model for evolving and delivering these services.

Final Report  VNC519-1920

Access the final report by clicking here.

 

Posted Date: July 17, 2023

Related Resources