Alexander Koltunov

Detection of salt marsh vegetation stress and recovery after the Deepwater Horizon Oil Spill in Barataria Bay, Gulf of Mexico using AVIRIS data

The British Petroleum Deepwater Horizon Oil Spill in the Gulf of Mexico was the biggest oil spill in US history. To assess the impact of the oil spill on the saltmarsh plant community, we examined Advanced Visible Infrared Imaging Spectrometer (AVIRIS) data flown over Barataria Bay, Louisiana in September 2010 and August 2011. Oil contamination was mapped using oil absorption features in pixel spectra and used to examine impact of oil along the oiled shorelines. Results showed that vegetation stress was restricted to the tidal zone extending 14 m inland from the shoreline in September 2010. Four indexes of plant stress and three indexes of canopy water content all consistently showed that stress was highest in pixels next to the shoreline and decreased with increasing distance from the shoreline. Index values along the oiled shoreline were significantly lower than those along the oil-free shoreline. Regression of index values with respect to distance from oil showed that in 2011, index values were no longer correlated with proximity to oil suggesting that the marsh was on its way to recovery. Change detection between the two dates showed that areas denuded of vegetation after the oil impact experienced varying degrees of re-vegetation in the following year. This recovery was poorest in the first three pixels adjacent to the shoreline. This study illustrates the usefulness of high spatial resolution airborne imaging spectroscopy to map actual locations where oil from the spill reached the shore and then to assess its impacts on the plant community. We demonstrate that post-oiling trends in terms of plant health and mortality could be detected and monitored, including recovery of these saltmarsh meadows one year after the oil spill.

Image construction using multitemporal observations and Dynamic Detection Models

This paper systematically derives and analyses the generic phenomenon of space‐invariant predictability of spatio‐temporal observation fields using past multitemporal observations. We focus on thermal infrared remote sensing as a non‐trivial example illustrating the predictability concept. The phenomenon and the systematic analysis thereof are experimentally demonstrated to be productive for developing effective automated anomaly detection and classification methods operating under the assumption of dynamic environment and sensor response. Using a simple preliminary experiment involving uncalibrated tower‐based high‐resolution thermal infrared surveillance, we test the conceptual validity of the space‐invariant multitemporal prediction and exemplify its potential applications. In addition, we use a MODIS thermal image sequence and the task of hot anomaly detection to demonstrate the applicability of the approach for monitoring the status of large territories from space‐borne platforms.

A new approach for spectral feature extraction and for unsupervised classification of hyperspectral data based on the Gaussian mixture model

This paper considers the task of unsupervised classification of hyperspectral data within a Gaussian mixture modeling framework. Different from the traditional clustering techniques that face serious conceptual problems in complex situations, the Gaussian mixture modeling approach provides a means of solving both simple and complex classification tasks as well as a way to substantiate results. Despite its theoretical advantages, in practice, the approach is rarely used for remote sensing because of objective difficulties in its implementation, especially for hyperspectral data. To show how these difficulties can be successfully overcome we present here a method for Gaussian mixture‐based unsupervised classification of hyperspectral data. The method contains a new approach for spectral feature extraction that reduces the data dimensionality while quantitatively guaranteeing the safety of the relevant information. The unsupervised learning procedure we propose is aimed at finding the number of classes that delivers a local maximum to the users confidence in subsequent labeling the pixels. The class parameters are determined via the Expectation‐Maximization algorithm, starting from the initial values selected systematically by the learning procedure. Decisions on class membership for the pixels are proposed to be made on the basis of the lower confidence bounds of the posterior probability estimates. This allows users to provide a probabilistic guarantee of the results for each separate pixel. The potential capabilities of the method are illustrated by its application to the GER 63 channel scanner hyperspectral data of the Naan area, central Israel. A comparison of the method with one of the traditional techniques we have performed confirms the advantages of the presented method, even though just a simplified version of the method has been employed.

Meta-analysis of California Spotted Owl (Strix occidentalis occidentalis) territory occupancy in the Sierra Nevada: Habitat associations and their implications for forest management

ABSTRACT We assessed the occupancy dynamics of 275 California Spotted Owl (Strix occidentalis occidentalis) territories in 4 study areas in the Sierra Nevada, California, USA, from 1993 to 2011. We used Landsat data to develop maps of canopy cover for each study area, which we then used to quantify annual territory-specific habitat covariates. We modeled the relationships between territory extinction and colonization using predictor variables of habitat, disturbance (logging, fire), climate, and elevation. We found that forests with medium (40–69%) and high (≥70%) canopy cover were the most important predictors of territory occupancy in all study areas, and that both canopy cover categories were positively correlated with occupancy. We used analysis of deviance to estimate the amount of variation explained by the habitat covariates (primarily medium and high canopy cover) and found that these covariates explained from 35% to 67% of the variation in occupancy. Climatic covariates were not correlated with occupancy dynamics and explained little of the variation in occupancy. We also conducted a post hoc analysis in which we partitioned canopy cover into 10% classes, because our original partitioning into 3 classes may have lacked sufficient resolution to identify canopy cover levels where occupancy changed abruptly. In this post hoc analysis, occupancy declined sharply when territories contained more area with <40% canopy cover, and the amount of 50–59% and 60–69% canopy cover had a more positive association with occupancy than did 40–49% canopy cover. Our results suggest that some fuels treatments intended to reduce fire risk and improve forest resilience could be located within Spotted Owl territories without adversely impacting territory occupancy if such treatments do not consistently reduce canopy cover below 50%. We suggest that future work quantify components of forest structure (e.g., large tree density, vertical complexity) known to be selected by owls and relate these characteristics to occupancy and fitness metrics.

Quantifying biomass consumption and carbon release from the California Rim fire by integrating airborne LiDAR and Landsat OLI data

Abstract Quantifying biomass consumption and carbon release is critical to understanding the role of fires in the carbon cycle and air quality. We present a methodology to estimate the biomass consumed and the carbon released by the California Rim fire by integrating postfire airborne LiDAR and multitemporal Landsat Operational Land Imager (OLI) imagery. First, a support vector regression (SVR) model was trained to estimate the aboveground biomass (AGB) from LiDAR‐derived metrics over the unburned area. The selected model estimated AGB with an R 2 of 0.82 and RMSE of 59.98 Mg/ha. Second, LiDAR‐based biomass estimates were extrapolated to the entire area before and after the fire, using Landsat OLI reflectance bands, Normalized Difference Infrared Index, and the elevation derived from LiDAR data. The extrapolation was performed using SVR models that resulted in R 2 of 0.73 and 0.79 and RMSE of 87.18 (Mg/ha) and 75.43 (Mg/ha) for the postfire and prefire images, respectively. After removing bias from the AGB extrapolations using a linear relationship between estimated and observed values, we estimated the biomass consumption from postfire LiDAR and prefire Landsat maps to be 6.58 ± 0.03 Tg (1012 g), which translate into 12.06 ± 0.06 Tg CO2e released to the atmosphere, equivalent to the annual emissions of 2.57 million cars.

Wavelet-Compressed Representation of Landscapes for Hydrologic and Geomorphologic Applications

The availability of high-resolution digital elevation data (submeter resolution) from LiDAR has increased dramatically over the past few years. As a result, the efficient storage and transmission of those large data sets and their use for geomorphic feature extraction and hydrologic/environmental modeling are becoming a scientific challenge. This letter explores the use of multiresolution wavelet analysis for compression of LiDAR digital elevation data sets. The compression takes advantage of the fact that, in most landscapes, neighboring pixels are correlated and thus contain some redundant information. The space-frequency localization of the wavelet filters allows one to preserve detailed high-resolution features where needed while representing the rest of the landscape at lower resolution. We explore a lossy compression methodology based on biorthogonal wavelets and demonstrate that, by keeping only approximately 10% of the original information (data compression ratio ~94%), the reconstructed landscapes retain most of the information of relevance to geomorphologic applications, such as the ability to accurately extract channel networks for environmental flux routing, as well as to identify geomorphic process transition from the curvature-slope and slope-distance relationships.

Marsh Loss Due to Cumulative Impacts of Hurricane Isaac and the Deepwater Horizon Oil Spill in Louisiana

Coastal ecosystems are greatly endangered due to anthropogenic development and climate change. Multiple disturbances may erode the ability of a system to recover from stress if there is little time between disturbance events. We evaluated the ability of the saltmarshes in Barataria Bay, Louisiana, USA, to recover from two successive disturbances, the DeepWater Horizon oil spill in 2010 and Hurricane Isaac in 2012. We measured recovery using vegetation indices and land cover change metrics. We found that after the hurricane, land loss along oiled shorelines was 17.8%, while along oil-free shorelines, it was 13.6% within the first 7 m. At a distance of 7–14 m, land loss from oiled regions was 11.6%, but only 6.3% in oil-free regions. We found no differences in vulnerability to land loss between narrow and wide shorelines; however, vegetation in narrow sites was significantly more stressed, potentially leading to future land loss. Treated oiled regions also lost more land due to the hurricane than untreated regions. These results suggest that ecosystem recovery after the two disturbances is compromised, as the observed high rates of land loss may prevent salt marsh from establishing in the same areas where it existed prior to the oil spill.

Geomorphologic mapping from hyperspectral data, using Gaussian mixtures and lower confidence bounds

Probabilistic classification under the Gaussian mixture model is normally based on posterior probability (p.p.) estimates of class membership. The question, how accurate they are for a given pixel, is traditionally left without attention, which may lead to unreasonable optimism about the classification results obtained. Addressing the issue, Koltunov and Ben‐Dor have proposed an unsupervised, lower confidence bound (l.c.b.)‐based method for thematic interpretation of remote sensing data. This method predicts the sampling properties of the p.p. estimators of a given pixel, to assess reliability of the estimates. The present paper describes a modified version of the method. In particular, instead of defining the l.c. bounds in terms of two first moments of the sampling distribution, as has been suggested previously, we use percentiles. Combining this with a probabilistic model of supervised identification of the mixture components yields the post‐classification uncertainty value for a given pixel and the confidence level, at which this value is proven to be maximal. In the application to an arid landscape in the Southern Negev desert, Israel, the compressed raw hyperspectral data acquired by the Digital Airborne Imaging Spectrometer (DIAS‐7915) was clustered once, whereas two thematic tasks were solved corresponding to different map legends, identification procedures, and the associated requirements to the level of detail and reliability of the thematic maps. The reference data collected in the field have provided evidence for accurate algorithmically estimated confidence bounds of the classification quality. The classification has revealed new information about the geomorphological subunits forming the study area.

Adaptive recognition under static and dynamic environment assumptions

This paper presents two approaches to ATR* by trainable algorithms. The first approach assumes that the measurements coming from the objects remain unchanged for the time passed between the stages of learning and recognition. For outdoor scenes such an approach is viable when both learning and recognition can be completed within minutes, which is difficult to achieve in practice. More realistic is to acquire training image data short before surveying the scene of interest. Then computer-intensive or interactive learning algorithms can be applied. We exemplify this approach qualitatively by detecting buildings and asphalt roads in a typical urban scene from AISA hyperspectral sensor data. The second, new approach we derive takes into account the joint changes of all targets and backgrounds under dynamic external factors. This requires multitemporally surveying an area that is specially selected for training an ATR system. Then at the future recognition stage the system can take advantage of the learning results in the real-time mode. Experimental verification of the new approach was performed using a fixed FLIR-type camera that surveyed the site containing more than 50 thermally different objects, whereas learning and recognition were spaced one week apart. The thermal joint prediction model proved working and was applied for detecting and identifying a scene anomaly -- an intruder.

Remote detection of water stress in orchard canopies using MODIS/ASTER airborne simulator (MASTER) data

Vegetation canopy water content (CWC) is an important parameter for monitoring natural and agricultural ecosystems. Previous studies focused on the observation of annual or monthly variations in CWC but lacked temporal details to study vegetation physiological activities within a diurnal cycle. This study provides an evaluation of detecting vegetation diurnal water stress using airborne data acquired with the MASTER instrument. Concurrent with the morning and afternoon acquisitions of MASTER data, an extensive field campaign was conducted over almond and pistachio orchards in southern San Joaquin Valley of California to collect CWC measurements. Statistical analysis of the field measurements indicated a significant decrease of CWC from morning to afternoon. Field measured CWC was linearly correlated to the normalized difference infrared index (NDII) calculated with atmospherically corrected MASTER reflectance data using either FLAASH or empirical line (EL). Our regression analysis demonstrated that both atmospheric corrections led to a root mean square error (RMSE) of approximately 0.035 kg/m2 for the estimation of CWC (R2=0.42 for FLAASH images and R2=0.45 for EL images). Remote detection of the subtle decline in CWC awaits an improved prediction of CWC. Diurnal CWC maps revealed the spatial patterns of vegetation water status in response to variations in irrigation treatment.