Jan Verbesselt

Evaluating satellite and climate data-derived indices as fire risk indicators in savanna ecosystems

The repeated occurrence of severe wildfires has highlighted the need for development of effective vegetation monitoring tools. We compared the performance of indices derived from satellite and climate data as a first step toward an operational tool for fire risk assessment in savanna ecosystems. Field collected fire activity data were used to evaluate the potential of the normalized difference vegetation index (NDVI), normalized difference water index (NDWI), and the meteorological Keetch-Byram drought idex (KBDI) to assess fire risk. Performance measures extracted from the binary logistic regression model fit were used to quantitatively rank indices in terms of their effectiveness as fire risk indicators. NDWI performed better when compared to NDVI and KBDI based on the results from the ranking method. The c-index, a measure of predictive ability, indicated that the NDWI can be used to predict seasonal fire activity (c=0.78). The time lag at the start of the fire season between time-series of fire activity data and the selected indices also was studied to evaluate the ability to predict the start of the fire season. The results showed that NDVI, NDWI, and KBDI can be used to predict the start of the fire season. NDWI consequently had the highest capacity to monitor fire activity and was able to detect the start of the fire season in savanna ecosystems. It is shown that the evaluation of satellite- and meteorological fire risk indices is essential before the indices are used for operational purposes to obtain more accurate maps of fire risk for the temporal and spatial allocation of fire prevention or fire management.

Spectral mixture analysis to monitor defoliation in mixed-aged Eucalyptus globulus Labill plantations in southern Australia using Landsat 5-TM and EO-1 Hyperion data

Defoliation is a key parameter of forest health and is associated with reduced productivity and tree mortality. Assessing the health of forests requires regular observations over large areas. Satellite remote sensing provides a cost-effective alternative to traditional ground-based assessment of forest health, but assessing defoliation can be difficult due to mixed pixels where vegetation cover is low or fragmented. In this study we apply a novel spectral unmixing technique, referred to as weighted Multiple Endmember Spectral Mixture Analysis (wMESMA), to Landsat 5-TM and EO-1 Hyperion data acquired over a Eucalyptus globulus (Labill.) plantation in southern Australia. This technique combines an iterative mixture analysis cycle allowing endmembers to vary on a per pixel basis (MESMA) and a weighting algorithm that prioritizes wavebands based on their robustness against endmember variability. Spectral mixture analysis provides an estimate of the physically interpretable canopy cover, which is not necessarily correlated with defoliation in mixed-aged plantations due to natural variation in canopy cover as stands age. There is considerable variability in the degree of defoliation as well as in stand age among sites and in this study we found that results were significantly improved by the inclusion of an age correction algorithm for both the multi-spectral (R2no age correction = 0.55 vs R2age correction = 0.73 for Landsat) and hyperspectral (R2no age correction = 0.12 vs R2age correction = 0.50 for Hyperion) image data. The improved accuracy obtained from Landsat compared to the Hyperion data illustrates the potential of applying SMA techniques for analysis of multi-spectral datasets such as MODIS and SPOT-VEGETATION.

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.

Penalized regression techniques for prediction: a case study for predicting tree mortality using remotely sensed vegetation indices

Constructing models can be complicated when the available fitting data are highly correlated and of high dimension. However, the complications depend on whether the goal is prediction instead of estimation. We focus on predicting tree mortality (measured as the number of dead trees) from change metrics derived from moderate-resolution imaging spectroradiometer satellite images. The high dimensionality and multicollinearity inherent in such data are of particular concern. Standard regression techniques perform poorly for such data, so we examine shrinkage regression techniques such as ridge regression, the LASSO, and partial least squares, which yield more robust predictions. We also suggest efficient strategies that can be used to select optimal models such as 0.632+ bootstrap and generalized cross validation. The techniques are compared using simulations. The techniques are then used to predict insect-induced tree mortality severity for a Pinus radiata D. Don plantation in southern New South Wales, Austr...

Spatio-Temporal Segmentation Based on Subsequences of Satellite Image Time Series

Hierarchical image segmentation methodologies have the potential to integrate temporal information, spatial context and the hierarchical complexity of satellite image time series. The current methods, however, fail to identify the distinction between subsequences of time series, which can be essential for the interpretation of ecosystem processes. Therefore a novel conceptual methodology is introduced that allows an enhanced multi-temporal hierarchical image segmentation (EMTHIS) based on subsequences of time series (i.e., time series over a specified time window). The effect of using these subsequence windows is illustrated in an accuracy assessment approach that determines the accuracy of a classification based on subsequence windows versus the existent methodologies that do not take subsequence windows into account. Analysis of the accuracy assessment approach demonstrated the importance of considering image time series subsequences when the percentage of pixels that shows a land cover / land use change between consecutive years is above 0.5%.

Biophysical drought metrics extraction by time series analysis of SPOT Vegetation data

The repeated occurrence of severe wildfires has highlighted the need for development of effective vegetation moisture monitoring tools. The normalized difference infrared index (NDII) derived from SPOT Vegetation satellite data and the Keetch-Byram drought index derived from temperature and rainfall data are both related to vegetation moisture dynamics. Autocorrelation of time series is a major issue when time series derived from remote sensing and meteorological variables are analyzed. Autocorrelation affects cross-correlation between variables measured in time, and violates the basic regression assumption of independence. Therefore, this study focuses on the extraction of independent drought metrics from seasonal time series to define quantitive relationships between remote sensing and meteorological time series. First, the correlation between time series of satellite- and climate-data based indices is investigated by cross-correlation analysis. Secondly, a novel method for extraction of drought metrics is optimized for satellite- and in-situ derived time series. The method is based on a nonlinear least squares fit of asymmetric Gaussian model functions. The smooth model functions are then used for defining key seasonality parameters. The hypothesis is that the `seasonal shapes of satellite- and in-situ derived time series are correlated. Based on this hypothesis, the performance for parameter extraction from time series is explored

Integration of Magnitude and Shape Related Features in Hyperspectral Mixture Analysis to Monitor Weeds In Citrus Orchards

Traditionally, Spectral Mixture Analysis (SMA) fails to fully account for highly similar ground components or endmembers. The high similarity between weed and crop spectra therefore hampers the implementation of SMA for steering weed control management practices. To address this problem, the current study presents an alternative SMA technique, referred to as Integrated Spectral Mixture Analysis (iSMA). iSMA combines both magnitude (~reflectance) and shape (~derivatives) related features in an automated waveband selection protocol and allows for an optimal separation between weed and crops, irrespective of the scenario considered. Compared to traditional approaches iSMA significantly improved weed cover fraction estimations (~17% increase). Analysis was performed on different simulated mixed pixel spectra sets compiled from in situ measured weed, Citrus canopy and soil spectra.

Estimating vegetation dryness to optimize fire risk assessment with spot vegetation satellite data in savanna ecosystems

The lack of information on vegetation dryness prior to the use of fire as a management tool often leads to a significant deterioration of the savanna ecosystem. This paper therefore evaluated the capacity of SPOT VEGETATION time-series to monitor the vegetation dryness (i.e., vegetation moisture content per vegetation amount) in order to optimize fire risk assessment in the savanna ecosystem of Kruger National Park in South Africa. The integrated Relative Vegetation Index approach (iRVI) to quantify the amount of herbaceous biomass at the end of the rain season and the Accumulated Relative Normalized Difference vegetation index decrement (ARND) related to vegetation moisture content were selected. The iRVI and ARND related to vegetation amount and moisture content, respectively, were combined in order to monitor vegetation dryness and optimize fire risk assessment in the savanna ecosystems. In situ fire activity data was used to evaluate the significance of the iRVI and ARND to monitor vegetation dryness for fire risk assessment. Results from the binary logistic regression analysis confirmed that the assessment of fire risk was optimized by integration of both the vegetation quantity (iRVI) and vegetation moisture content (ARND) as statistically significant explanatory variables. Consequently, the integrated use of both iRVI and ARND to monitor vegetation dryness provides a more suitable tool for fire management and suppression compared to other traditional satellite-based fire risk assessment methods, only related to vegetation moisture content.

Development of indicators of vegetation recovery based on time series analysis of SPOT VEGETATION data

Large-scale wild fires have direct impacts on natural ecosystems and play a major role in the vegetation ecology and carbon budget. Accurate methods for describing post-fire development of vegetation are therefore essential for the understanding and monitoring of terrestrial ecosystems. Time series analysis of satellite imagery offers the potential to quantify these parameters with spatial and temporal accuracy. Current research focuses on the potential of time series analysis of SPOT Vegetation S10 data (1999-2001) to quantify the vegetation recovery of large-scale burns detected in the framework of GBA2000. The objective of this study was to provide quantitative estimates of the spatio-temporal variation of vegetation recovery based on remote sensing indicators. Southern Africa was used as a pilot study area, given the availability of ground and satellite data. An automated technique was developed to extract consistent indicators of vegetation recovery from the SPOT-VGT time series. Reference areas were used to quantify the vegetation regrowth by means of Regeneration Indices (RI). Two kinds of recovery indicators (time and value- based) were tested for RIs of NDVI, SR, SAVI, NDWI, and pure band information. The effects of vegetation structure and temporal fire regime features on the recovery indicators were subsequently analyzed. Statistical analyses were conducted to assess whether the recovery indicators were different for different vegetation types and dependent on timing of the burning season. Results highlighted the importance of appropriate reference areas and the importance of correct normalization of the SPOT-VGT data.