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.

Satellite earth observation is being increasingly used to monitor forests across the world. Freely available Landsat data stretching back four decades, coupled with advances in computer processing capabilities, has enabled new time-series techniques for analyzing forest change. Typically, these methods track individual pixel values over time, through the use of various spectral indices. This study examines the utility of eight spectral indices for characterizing fire disturbance and recovery in sclerophyll forests, in order to determine their relative merits in the context of Landsat time-series. Although existing research into Landsat indices is comprehensive, this study presents a new approach, by comparing the distributions of pre and post-fire pixels using Glass’s delta, for evaluating indices without the need of detailed field information. Our results show that in the sclerophyll forests of southeast Australia, common indices, such as the Normalized Difference Vegetation Index (NDVI) and the Normalized Burn Ratio (NBR), both accurately capture wildfire disturbance in a pixel-based time-series approach, especially if images from soon after the disturbance are available. However, for tracking forest regrowth and recovery, indices, such as NDVI, which typically capture chlorophyll concentration or canopy ‘greenness’, are not as reliable, with values returning to pre-fire levels in 3–5 years. In comparison, indices that are more sensitive to forest moisture and structure, such as NBR, indicate much longer (8–10 years) recovery timeframes. This finding is consistent with studies that were conducted in other forest types. We also demonstrate that additional information regarding forest condition, particularly in relation to recovery, can be extracted from less well known indices, such as NBR2, as well as textural indices incorporating spatial variance. With Landsat time-series gaining in popularity in recent years, it is critical to understand the advantages and limitations of the various indices that these methods rely on.

The free open access data policy instituted for the Landsat archive since 2008 has revolutionised the use of Landsat data for forest monitoring, especially for estimating forest aboveground biomass (AGB). This paper provides a comprehensive review of recent approaches utilising Landsat time-series (LTS) for estimating AGB and its dynamics across space and time. In particular, we focus on reviewing: (1) how LTS has been utilised to improve the estimation of AGB (for both single-date and over time) and (2) recent LTS-based approaches used for estimating AGB and its dynamics across space and time. In contrast to using single-date images, the use of LTS can benefit forest AGB estimation in two broad areas. First, using LTS allows for the filling of spatial and temporal data gaps in AGB predictions, improving the quality of AGB products and enabling the estimation of AGB across large areas and long time-periods. Second, studies have demonstrated that spectral information extracted from LTS analysis, including forest disturbance and recovery metrics, can significantly improve the accuracy of AGB models. Throughout the last decade, many innovative LTS-based approaches for estimating forest AGB dynamics across space and time have been demonstrated. A general trend is that methods have evolved as demonstrated through recent studies, becoming more advanced and robust. However, most of these methods have been developed and tested in areas that are either supported by established forest inventory programs and/or can rely on Lidar data across large forest areas. Further investigations should focus on tropical forest areas where inventory data are often not systematically available and/or out-of-date.

Hyperspectral remote sensing serves as an effective tool for estimating foliar nitrogen using a variety of techniques. Vegetation indices (VIs) are a simple means of retrieving foliar nitrogen. Despite their popularity, few studies have been conducted to examine the utility of VIs for mapping canopy foliar nitrogen in a mixed forest context. In this study, we assessed the performance of 32 vegetation indices derived from HySpex airborne hyperspectral images for estimating canopy mass-based foliar nitrogen concentration (%N) in the Bavarian Forest National Park. The partial least squares regression (PLSR) was performed for comparison. These vegetation indices were classified into three categories that are mostly correlated to nitrogen, chlorophyll, and structural properties such as leaf area index (LAI). %N was destructively measured in 26 broadleaf, needle leaf, and mixed stand plots to represent the different species and canopy structure. The canopy foliar %N is defined as the plot-level mean foliar %N of all species weighted by species canopy foliar mass fraction. Our results showed that the variance of canopy foliar %N is mainly explained by functional type and species composition. The normalized difference nitrogen index (NDNI) produced the most accurate estimation of %N (R²_CV = 0.79, RMSE_CV = 0.26). A comparable estimation of %N was obtained by the chlorophyll index Boochs2 (R²_CV = 0.76, RMSE_CV = 0.27). In addition, the mean NIR reflectance (800–850 nm), representing canopy structural properties, also achieved a good accuracy in %N estimation (R²_CV = 0.73, RMSE_CV = 0.30). The PLSR model provided a less accurate estimation of %N (R²_CV = 0.69, RMSE_CV = 0.32). We argue that the good performance of all three categories of vegetation indices in %N estimation can be attributed to the synergy among plant traits (i.e., canopy structure, leaf chemical and optical properties) while these traits may converge across plant species for evolutionary reasons. Our findings demonstrated the feasibility of using hyperspectral vegetation indices to estimate %N in a mixed temperate forest which may relate to the effect of the physical basis of nitrogen absorption features on canopy reflectance, or the biological links between nitrogen, chlorophyll, and canopy structure.

Satellite remote sensing is regularly used for wildfire detection, fire severity mapping and burnt area mapping. Applications in the surveillance of wildfire using geostationary-based sensors have been limited by low spatial resolutions. With the launch in 2015 of the AHI (Advanced Himawari Imaginer) sensor on board Himawari-8, ten-minute interval imagery is available covering an entire earth hemisphere across East Asia and Australasia. Existing active fire detection algorithms depend on middle infrared (MIR) and thermal infrared (TIR) channels to detect fire. Even though sub-pixel fire detection algorithms can detect much smaller fires, the location of the fire within the AHI 2 × 2 km (400 ha) MIR/TIR pixel is unknown. This limits the application of AHI as a wildfire surveillance and tracking sensor. A new multi-spatial resolution approach is presented in this paper that utilizes the available medium resolution channels in AHI. The proposed algorithm is able to map firelines at a 500 m resolution. This is achieved using near infrared (NIR) (1 km) and RED (500 m) data to detect burnt area and smoke within the flagged MIR (2 km) pixel. Initial results based on three case studies carried out in Western Australia shows that the algorithm was able to continuously track fires during the day at 500 m resolution. The results also demonstrate the utility for wildfire management activities.

Operational assessment of forest structure is an on-going challenge for land managers, particularly over large, remote or inaccessible areas. Here, we present an easily adopted method for generating a continuous map of canopy height at a 30 m resolution, demonstrated over 2.9 million hectares of highly heterogeneous forest (canopy height 0–70 m) in Victoria, Australia. A two-stage approach was utilized where Airborne Laser Scanning (ALS) derived canopy height, captured over ~18% of the study area, was used to train a regression tree ensemble method; random forest. Predictor variables, which have a global coverage and are freely available, included Landsat Thematic Mapper (Tasselled Cap transformed), Moderate Resolution Imaging Spectroradiometer Normalized Difference Vegetation Index time series, Shuttle Radar Topography Mission elevation data and other ancillary datasets. Reflectance variables were further processed to extract additional spatial and temporal contextual and textural variables. Modeled canopy height was validated following two approaches; (i) random sample cross validation; and (ii) with 108 inventory plots from outside the ALS capture extent. Both the cross validation and comparison with inventory data indicate canopy height can be estimated with a Root Mean Square Error (RMSE) of ≤ 31% (~5.6 m) at the 95th percentile confidence interval. Subtraction of the systematic component of model error, estimated from training data error residuals, rescaled canopy height values to more accurately represent the response variable distribution tails e.g., tall and short forest. Two further experiments were carried out to test the applicability and scalability of the presented method. Results suggest that (a) no improvement in canopy height estimation is achieved when models were constructed and validated for smaller geographic areas, suggesting there is no upper limit to model scalability; and (b) training data can be captured over a small percentage of the study area (~6%) if response and predictor variable variance is captured within the training cohort, however RMSE is higher than when compared to a stratified random sample.