Process-based models integrated with G.I.S.

Hi,

My name is Georgios Xenakis and I am currently completing my Ph.D. with the University of Edinburgh. I am a forester and I hold a master degree in Resource Management by the University of Edinburgh. I got interested in Geographical Information Systems and Spatial analysis during my M.Sc. thesis. My first experience with G.I.S. was when I worked for the Forest Research Institute of Greece. There I saw for the first time the real application of G.I.S. in forest management. During my M.Sc. thesis I’ve try to introduce the “new” for me technology with more traditional forestry such as the predicting the sustainable yield based on yield tables. I’ve customised ArcView using Aveniue to introduce automated ways of predicting potential growth of Black Pine for Taygetos forest in Greece. With the application I was able to demonstrate the use of G.I.S. in forest management planning by providing maps of potential productivity and maps of the removable yield so as to ensure sustainability of resources.

However, the application was based on statistical empirical models, as yield tables normally are. The way of construction of yield tables and also their inability to introduce any possible environmental effects restrict their use, especially for a Geographic Information System framework. The solution to this is process-based modelling.

Process-based models are either detailed or simplified mathematical models, which are based on representation of basic physiological principals of tree growth. Photosynthesis, respiration, senescence and regeneration are some of the basic parts of the process that such models include utilizing differential equations. Introducing those into a G.I.S. framework and you have a management tool, which is capable for adaptation under different management scenarios.

However, such models have also their limitations. Differential equations could become extremely complex when a fine temporal scale is applied. Multiply by a fine spatial scale and you have such a huge calculation matrix that computer calculation capabilities are either impossible or extremely time consuming. When it comes into real fast and reliable solutions for forest management decisions then detailed process-based become prohibited. To overcome such model, ecological modellers developed simplified process-based models. Those models still include all the basic tree physiological principles but their differential equations are more simplified in their representation. This simplification comes either because the temporal scale is increased or because some basic physiological functions were simplified e.g. respiration.

I am using 3-PGN (Physiological Principles of Predicting Growth – Nitrogen), which is a simplified process-based model. Their routines include photosynthesis estimation based on quantum utilisation radiation efficiency, respiration based on a ratio between net and gross primary production and tree mortality. Unfortunately it does not include any regeneration routines however it includes a soil decomposition model, which provides estimates of soil carbon and nitrogen stocks together with an estimation of soil fertility based on nitrogen. Additionally the model is capable to provide estimates of stem, root and foliage biomass through allometric relationships and also an estimation of produced timber, mean annual increment, mean diameter and height, variables all valuable for forest management decisions.

The model was scripted into C++ and compiled. The integration of 3-PGN was done with a loose coupling technique, where model and G.I.S. software share common files but not the same interface. The C++ executable version shares float point grid files with ArcMap and ArcInfo. The model performs all simulations separately and results are produced in the same file format. Then transformation with the float grid routine of ArcMap and ArcInfo provide the spatial grids for representation and further analysis.

Although the version of the model without the soil decomposition models exists fully integrated with ArcMap, programmed with Visual Basic and is available by CSIRO Australia, the full ecosystem model 3-PGN is only available with this loose-coupled version.

I used the model to provide spatial patterns of potential Scots Pine productivity for Scotland. Based on a series of spatially generated environmental inputs such as maximum and minimum temperature, precipitation, incoming solar radiation and frost days the model gave a series of spatially explicit outputs such as potential mean annual volume increment, potential carbon accumulation, potential net ecosystem productivity, ecosystem respiration, carbon and nitrogen stocks and soil available nitrogen.

One part of my Ph.D. project is to provide a site classification based not only on capabilities of a site for potential timber production but also based on site capabilities for potential carbon sequestration. This way, forest managers would be able to provide forest planning based on different management objectives, which could include either timber production, recreation or even carbon sequestration.

G.I.S. provided me the tool to understand which is the combined effect of environmental and topographical effect with its advanced spatial capabilities. Spatial analysis performed on the spatial 3-PGN outputs include principal component analysis for reduction of the climatic input variables into a manageable number of variables, spatial autocorrelation analysis using Moran’s I statistic so as to explore how clustered or random are the produced by the model estimated patterns of carbon and timber production, regression analysis to identify which is the climatic or topographic inputs which spatial patterns match those of the produced timber production and carbon sequestration patterns and correlation analysis to explore thoroughly the most significant effects of climate and topography on patterns of potential timber and carbon productivity. Advanced mapping tools from G.I.S. helped me to produce high quality maps of all outputs of the model such as net ecosystem production, ecosystem respiration, soil carbon and nitrogen tools and also to provide the visual tool for producing potential timber production classes maps and carbon sequestration regions.

The results so far shows that Scots pine potential timber productivity is highly rationalised with almost circular regions of high productivity next to coastal areas leading to low productivity regions within the mountainous regions of Scotland. G.I.S. spatial analysis and map overlay showed that the less productive regions of Scots pine in Scotland is the western parts where precipitation and so potential productivity is larger, where the high productivity areas are placed on the east side of Scotland. Water although in excess in western parts of the country does not contribute into greater carbon or timber production as solar radiation (a key factor for productions) is also extremely limited. Also results of the project based on spatial analysis showed that Scots pine grows better in sandy regions where soils are not water logged. Winter values of maximum and minimum temperature have a greater effect on growth (both in terms of carbon and timber). Significant also appears to be the number of frost days during the winter months which could limit growth. Finally map overlay of model produced maps with maps from 1970’s showed that potential patterns of Scots pine productivity are explained by potential soil water availability with the waterlogged soils having restricting growth.

The potential of G.I.S. for research and forest management combined with process-based models is huge. When remote sensing comes also into alliance then the results is a more powerful tool for sustainable forest management decisions.

The future framework of any national forest service should be the introduction of a G.I.S. web-based tool utilizing spatially extended climatic and soil with remotely sensed inputs of leaf area index, providing simulations with process-based models and creating a database of potential scenarios based on management decisions. Model predictions should include both timber and carbon based predictions, which later should be capable for visualisation through a web-based G.I.S. version freely available to all managers for decision making.

The future of process-based models has two directions. First is the introduction of routines capable to predict regeneration of woodlands based on current environmental and soil conditions, and second the integration of the whole package into a web-based G.I.S. framework.

Because the key for a sustainable management decision for world’s forests is based on reliable and process-based information widely available by modern technologies.

————–

If you are interested to learn more about process-based models, 3-PGN, my spatial analysis with ArcInfo and the results of my project you could either visit my personal web-page at http://www.geos.ed.ac.uk/homes/s0199797 or email me at: G.Xenakis@sms.ed.ac.uk