Christine Stone

The Performance of Random Forests in an Operational Setting for Large Area Sclerophyll Forest Classification

Mapping and monitoring forest extent is a common requirement of regional forest inventories and public land natural resource management, including in Australia. The state of Victoria, Australia, has approximately 7.2 million hectares of mostly forested public land, comprising ecosystems that present a diverse range of forest structures, composition and condition. In this paper, we evaluate the performance of the Random Forest (RF) classifier, an ensemble learning algorithm that has recently shown promise using multi-spectral satellite sensor imagery for large area feature classification. The RF algorithm was applied using selected Landsat Thematic Mapper (TM) imagery metrics and auxiliary terrain and climatic variables, while the reference data was manually extracted from systematically distributed plots of sample aerial photography and used for training (75%) and accuracy (25%) assessment. The RF algorithm yielded an overall accuracy of 96% and a Kappa statistic of 0.91 (confidence interval (CI) 0.909–0.919) for the forest/non-forest classification model, given a Kappa maximised binary threshold value of 0.5. The area under the receiver operating characteristic plot produced a score of 0.91, also indicating high model performance. The framework described in this study contributes to the operational deployment of a robust, but affordable, program, able to collate and process large volumes of multi-sourced data using open-source software for the production of consistent and accurate forest cover maps across the full spectrum of Victorian sclerophyll forest types.

Crown-scale evaluation of spectral indices for defoliated and discoloured eucalypts

Remote sensing for evaluation of canopy health in plantation eucalypts is a realistic option for forest managers in the near future if reliable and robust methods of spectral analysis can be developed. Pot‐grown eucalypts of three species important to the Australian plantation industry were used for crown‐scale spectral (400–1000 nm) evaluations of vegetation indices as indicators of common symptoms of stress. When defoliation treatments (in E. globulus) or exposure to cold and nutrient deprivation (in E. pilularis) resulted in large differences in leaf cover, the red edge position and slope indices, two normalized difference vegetation indices (NDVIs), modified chlorophyll absorption ratio index 2 (MCARI2) or modified triangular vegetation index 2 (MTVI2) were most strongly correlated to leaf cover. However the NDVIs were significantly affected by soil background in a study with E. globulus. The percentage of red leaves resulting from stress treatment was most strongly correlated with the anthocyanin reflectance index (ARI) and red‐green index (RGI) in both E. grandis and E. pilularis, however the RGI was affected by background type in the E. globulus study while the ARI was not. Exposure to cold and nutrient deprivation led to marked changes in leaf cover for E. pilularis but not in E. grandis and a much more reduced level of chlorophyll in E. pilularis than is suspected in E. grandis. In E. globulus, defoliation from the upper crown was easier to detect with spectral data than from the lower crown. Results were generally comparable to studies of eucalypt crown condition from native forests.

Assessment and monitoring of decline and dieback of forest eucalypts in relation to ecologically sustainable forest management: a review with a case study

Summary In this paper techniques for assessing and monitoring the decline and dieback of forest eucalypts are reviewed in relation to ecologically sustainable forest management. The concept of ecologically sustainable management is a guiding philosophy being adopted for the management of all Australian native forests. An integral component of this strategy will be the ability to monitor forest health. This requires assessment of both the appearance of trees (canopy condition) and their underlying physiological status. However, visual assessment of eucalypt canopies can be misleading because of their opportunistic leafing phenology and epicormic regrowth characteristics. Several forest health assessment programs outside Australia now use multispectral imagery obtained by remote sensing. Such programs depend on relating spectral features with plant physiological properties such as chlorophyll content of the foliage. More recently chlorophyll fluorescence spectra have been related to the functioning of the p...

A Review of Eucalypt Dieback Associated With Bell Miner Habitat in South-Eastern Australia

Summary We aim to assess current knowledge, and identify gaps in knowledge concerning bell-miner-associated dieback (BMAD) in south-eastern Australia. We review BMAD as a form of forest dieback, and bell miner and psyllid interrelations. We then consider indirect and direct causal factors associated with local functional scales (tree crown), and finally, indirect and direct causal factors associated with broader functional scales (forest stand and landscape processes). This paper emphasises published literature and is a summary of a more detailed report prepared for the BMAD working group which explicitly included personal communications with many researchers, managers and members of conservation groups and the timber industry. We conclude that BMAD is a significant threat to the sustainability of the moist eucalypt forests of north-eastern NSW and south-eastern Queensland, and to biodiversity conservation at a national scale. There are serious deficiencies in the information base for most BMAD issues. While there are clear interactions between bell miners and psyllids. there are many other, less well quantified interactions that may be of greater significance to the development of the problem. We suggest that management and research efforts towards solutions urgently target disturbances that lead to changes in forest canopy structure, but there is unlikely to be a single or simple management solution. An integrated management program will be necessary as concentration on particular management regimes in isolation is unlikely to resolve the BMAD problem because BMAD is associated with interacting disturbances.

Bell-miner-associated dieback at the tree crown scale: a multi-trophic process

Summary This paper examines some of the factors and processes which interact within eucalypt stands affected by bell-miner-associated dieback. A key symptom of this form of dieback is sustained foliar damage from herbivorous insects, in particular psyllids. Repeated cycles of defoliation/refoliation result in branch death and crown contraction. Weakened trees become more susceptible to secondary stressors such as wood borers, soil fungal pathogens and abnormal levels of soil moisture. Psyllid outbreaks can occur through a reduction in the efficacy of their natural enemy complex, and the provision of sufficient favourable foliage to sustain high insect populations. The young foliage preferred for feeding and oviposition by most eucalypt insect herbivores can occur when soil moisture and nutrients and canopy irradiation are non-limiting. Bell miner colonies require high insect densities in their food source and they reinforce this situation through their feeding behaviour and territorial defence. The study suggests the effect of bell miners on the populations of herbivorous insects, and hence the amount of insect-damaged foliage, is density dependent. Bell miners also require dense understorey, about 2–5 m high, for nesting sites, and the presence of surface water. Any management practice that reduces bell miner density will in turn reduce the density of insect herbivores and hence foliar damage in the eucalypt tree crowns. Then, if the trees have not become too debilitated, crown recovery is likely to occur.

A comparison of field-based and modelled reflectance spectra from damaged Pinus radiata foliage

Accurate and cost-effective monitoring of the health and condition of Australian Pinus radiata D.Don plantations is crucial to predicting the impact of damaging agents on wood yield and, where appropriate, targeting timely intervention. Stressful agents can induce changes in the biochemical, physiological and structural integrity of pine needles and subsequently reduce tree growth and ultimately cause plant death. Three important stressful agents occurring within Australian P. radiata plantations are the aphid Essigella californica, soil nitrogen deficiency and Sphaeropsis sapinea, a fungal pathogen. Within a study site in southern New South Wales, needles were sampled from crowns exhibiting key symptoms at three levels of crown severity. Needle level spectra were measured with a field spectroradiometer and foliage samples taken to extract needle chlorophyll a and b and to determine needle moisture content. A radiative transfer model (LIBERTY) was also used to estimate theoretical needle reflectance, given changes in two of its five input parameters (needle chlorophyll and moisture content). Two specific questions were posed. First, given that most spectral indices are based on a reference or stable wavelength as well as sensitive wavelengths, what is the most effective suite of stable wavelengths for predicting of needle chlorophyll and moisture? Second, which published spectral indices best discriminated the three categories of crown-damage severity for each damaging agent? Analysis of needle samples indicated that the needles affected by E. californica were the least chlorotic compared with the other damaging agents. For all damaging agents, needles showed an increase in reflectance with a lowering of chlorophyll content in the visible region (400–700 nm), associated with increasing severity. Changes in the shape of the spectral curve in the red-edge region of the electromagnetic spectrum were minor for E. californica-affected and nitrogen-deficient needles; however, changes were significant when comparing the S. sapinea severity classes. Correlations with published vegetation indices indicated that needle chlorophyll content was most highly correlated with a number of the recently proposed indices, including the structurally insensitive simple ratio. In general, the best results were obtained with 705 nm as the chlorophyll sensitive wavelength and either 750 or 445 nm as the insensitive wavelengths to account for needle reflectance and surface properties. By varying two of the input parameters of the LIBERTY model, the estimated spectra generally matched the trends and magnitude of actual spectra. This suggests that the application of radiative transfer models, correctly parameterised, can provide important information when estimating discrimination categories of needle damage.

Integrating plantation health surveillance and wood resource inventory systems using remote sensing

Summary Commercial softwood growers in Australia are keen to improve the efficiency and precision of resource inventory underpinning their timber supply commitments. At the same time, they also need to implement forest health strategies which contribute to their environmental management systems and certification process. For example, the Australian Forestry Standard requires forest managers to identify, assess and prioritise any potential damage agents that may impact on forest ecosystem health and vitality. These health programs, however, are often run parallel with, and independently of, resource inventory programs. While most large growers maintain a health surveillance program, their capacity to quantify the impact of damaging agents on stand productivity and wood volume is often limited. Quantification of productivity losses due to biotic and abiotic agents would significantly improve decisions associated with resource scheduling and allocation of resources for pest control and stand amelioration. This paper discusses how remote sensing technologies can provide spatially-explicit data that permit the integration of plantation inventory and health assessments. The emerging diversity of sensor capabilities on both satellite and airborne platforms enables the development of hierarchical monitoring programs that can be customised for individual regions. For example, the coarse-scale sensor MODIS can provide very cheap coverage suitable for frequent temporal condition monitoring (thus identifying areas requiring more detailed attention in a timely manner), whereas the new generation of high-resolution sensors are facilitating a shift from manually mapped stand polygons (e.g. those from aerial sketchmapping and aerial photographic interpretation — API) to pixel and object-based digital analysis techniques suitable for both crown and stand-level inventory and canopy health assessment on a continuous, broad-scale basis. The application of these new technologies and associated spatial analyses permits the integration of plantation inventory and health assessment, thus providing forest managers with a holistic and cost-effective approach to timber production.

Effects of leaf age and psyllid damage on the spectral reflectance properties of Eucalyptus saligna foliage

Leaf chlorophyll content is influenced directly by many environmental stress factors. Because leaf pigment absorption is wavelength dependent, numerous narrow-band reflectance-based indices have been proposed as a means of assessing foliar health and condition. Chlorophyll content, however, also varies with leaf developmental stage. In this study, a range of morphological and physiological traits including insect damage, relative chlorophyll content (SPAD values), chlorophyll fluorescence (Fv/Fm) and reflectance spectra was measured of leaves sampled from mature Eucalyptus saligna. Relative differences among three leaf-age cohorts were compared with differences obtained from mature leaves that were either healthy or infested with the psyllid Glycaspis baileyi. Differences in relative chlorophyll content were greater between immature and mature foliage than between damaged and healthy mature leaves. These differences were confirmed in the comparisons of reflectance spectra and indices. As many eucalypt species have opportunistic crown phenology and long-lived leaves, leaf-age composition of crowns needs to be taken into account when applying reflectance-based indices to assess foliar condition of eucalypts.

Forest canopy health and stand structure associated with bell miners (Manorina melanophrys) on the central coast of New South Wales

Summary Forest attributes including canopy health and stand structure were visually assessed to identify features commonly associated with the bell miner (Manorina melanophrys, Meliphagidae), a cooperatively breeding, insectivorous honeyeater endemic to south-eastern Australia. A stratified random sampling methodology, using a combination of SPOT5 satellite imagery and a 25-m pixel digital altitude model, was used to establish 130 circular plots 40 m in diameter within a 12800 ha study area in the Watagan mountain range, central coast of New South Wales (NSW). In this study site, the presence of bell miners was significantly associated with unhealthy eucalypt crowns. This supports the proposition that a dieback syndrome known as bell-miner-associated dieback (BMAD) exists in central coastal forests of NSW. Our results also demonstrate a strong association of a sparse eucalypt canopy over a dense lower midstorey with the presence of bell miners. Altitude and fire frequency were negatively correlated with bell miners while the topographical wetness index was positively correlated with bell miners. Logistic regression was used to quantify the effects of a series of explanatory variables including tree species, forest structure variables and selected topographic variables. Although there were significant differences in the ranking of crown condition between eucalypt species, tree species was not selected in the final model. The final variables identified by the logistic regression for the presence of bell miners included sparse eucalypt overstorey cover, dense lower midstorey cover, the absence of young vigorous regrowth trees and higher values of the topographic wetness index. Based on the results of our survey we discuss practical options for managing the forest dieback syndrome associated with bell miners in the central coast region of NSW and an approach to mapping forests at risk from BMAD.

Semi-automating the Stand Delineation Process in Mapping Natural Eucalypt Forests

Summary For decades, aerial photo interpretation has been, and to a good extent is still, the method of choice for producing fine-scale native forest stand mapping. Recent computer techniques have eased the task of the interpreter, who is now able to delineate polygons through on-screen digitising in a geographical information system (GIS) environment. Even with these advances, a great deal of skill is required in the polygon delineation. In an effort to contribute to the automation of this process, we introduce an open-source object-based solution to the mapping of forest stand boundaries using attributes derived from digital aerial photography and laser scanning data acquired over a study area in the Victorian Central Highlands. This methodology transforms remotely sensed imagery (single or multichannel) and canopy raster layers derived from laser scanning (lidar) into polygon vector layers. It is intended that the resultant polygon layer should resemble the product derived by an aerial interpreter, without any prior knowledge of the scene. The derived product aims to produce a layer comprised of relatively homogeneous polygons all exceeding a minimum size. The derived product is meant to be a preliminary template aimed at reducing time and effort in manual digitisation. The relationship between spectral, texture and laser scanning derived features for forest stand boundary delineation and human interpreted boundaries is not straight forward. The interpreter however, can aggregate and sometimes correct the automated delineated regions by simple drag-and-click operations This approach is relatively cheap and flexible, being a workable compromise between fully automated image interpretation which requires further research for acceptable levels of accuracy and reliability, and manual segmentation and classification. Preliminary results are encouraging, both in regard to automating the process and the delivery of robust delineation of stand boundaries in native forest landscapes. Future research will focus on appropriate input resolution to reduce computation requirements and improved data fusion methods to obtain more accurate forest stand delineation.