Gene A. Nelson

Forest impact estimated with NOAA AVHRR and Landsat TM data related to an empirical hurricane wind-field distribution

Abstract An empirical model was used to relate forest type and hurricane-impact distribution with wind speed and duration to explain the variation of hurricane damage among forest types along the Atchafalaya River basin of coastal Louisiana. Forest-type distribution was derived from Landsat Thematic Mapper image data, hurricane-impact distribution from a suite of transformed advanced very high resolution radiometer images, and wind speed and duration from a wind-field model. The empirical model explained 73%, 84%, and 87% of the impact variances for open, hardwood, and cypress–tupelo forests, respectively. These results showed that the estimated impact for each forest type was highly related to the duration and speed of extreme winds associated with Hurricane Andrew in 1992. The wind-field model projected that the highest wind speeds were in the southern basin, dominated by cypress–tupelo and open forests, while lower wind speeds were in the northern basin, dominated by hardwood forests. This evidence could explain why, on average, the impact to cypress–tupelos was more severe than to hardwoods, even though cypress–tupelos are less susceptible to wind damage. Further, examination of the relative importance of wind speed in explaining the impact severity to each forest type showed that the impact to hardwood forests was mainly related to tropical-depression to tropical-storm force wind speeds. Impacts to cypress–tupelo and open forests (a mixture of willows and cypress–tupelo) were broadly related to tropical-storm force wind speeds and by wind speeds near and somewhat in excess of hurricane force. Decoupling the importance of duration from speed in explaining the impact severity to the forests could not be fully realized. Most evidence, however, hinted that impact severity was positively related to higher durations at critical wind speeds. Wind-speed intervals, which were important in explaining the impact severity on hardwoods, showed that higher durations, but not the highest wind speeds, were concentrated in the northern basin, dominated by hardwoods. The extreme impacts associated with the cypress–tupelo forests in the southeast corner of the basin intersected the highest durations as well as the highest wind speeds.

Generation and validation of characteristic spectra from EO1 Hyperion image data for detecting the occurrence of the invasive species, Chinese tallow

Chinese tallow (Triadica sebifera) is an invasive tree that is spreading throughout the south‐eastern United States and now into the west, and in many places causing extensive change to native habitat and associated wildlife. Detecting and mapping the relative distribution of this species is important to its control and eradication. To map the relative distribution of Chinese tallow within a south‐western Louisiana coastal wetland to upland environment, Earth Observing 1 (EO1) satellite Hyperion sensor hyperspectral image data were combined with a subpixel extraction method that modelled characteristic spectra from the image data without requiring a priori characteristic spectra. Because of the low percentage occurrences of Chinese tallow and high spectral covariation in the environment, unique validation and verification methods were implemented, relying on simultaneous collection of field canopy reflectance spectra and subsequent classification of canopy compositions. The subpixel extraction method produced five characteristic spectra, which we further refined to four that adequately represented the field spectra, as well as the Hyperion imaged canopy reflectance datasets. Characteristic spectra were designated as senescing foliage, cypress‐tupelo trees, and trees without leaves; shadows and green vegetation; senescing Chinese tallow with yellow leaves and yellowing foliage; and senescing Chinese tallow with red leaves (‘red tallow’). About 81% (n = 34) of the field and 78% (n = 33) of the Hyperion imaged characteristic spectra associated with ‘red tallow’ were explained by the compositions generated in the field slide classifications.

Using multiple-polarization L-band radar to monitor marsh burn recovery

Aircraft L-band VV-, HH-, and VH-polarizations were examined as tools for monitoring burn recovery in a coastal marsh. Significant relationships were observed between time-since-burn (difference between burn and image collection dates; 550-900 days after burn) and returns related to all polarizations. As marsh burn recovery progressed, VV returns decreased while HH and VH returns increased. Radar returns extracted from control sites adjacent to each burn-simulated nonburn marsh and were not individually or in combination significantly related to the time-since-burn. Normalized by the control data, VH-polarization explained up to 83% of the total variations. Overall, the L-band multipolarization radars estimated time-since-burn within /spl plusmn/59 to /spl plusmn/92 days.

Classifying coastal resources by integrating optical and radar imagery and color infrared photography

A progressive classification of a marsh and forest system using Landsat Thematic Mapper (TM), color infrared (CIR) photograph, and ERS-1 synthetic aperture radar (SAR) data improved classification accuracy when compared to classification using solely TM reflective band data. The classification resulted in a detailed identification of differences within a nearly monotypic black needlerush marsh. Accuracy percentages of these classes were surprisingly high given the complexities of classification. The detailed classification resulted in a more accurate portrayal of the marsh transgressive sequence than was obtainable with TM data alone. Individual sensor contribution to the improved classification was compared to that using only the six reflective TM bands. Individually, the green reflective CIR and SAR data identified broad categories of water, marsh, and forest. In combination with TM, SAR and the green CIR band each improved overall accuracy by about 3% and 15% respectively. The SAR data improved the TM classification accuracy mostly in the marsh classes. The green CIR data also improved the marsh classification accuracy and accuracies in some water classes. The final combination of all sensor data improved almost all class accuracies from 2% to 70% with an overall improvement of about 20% over TM data alone. Not only was the identification of vegetation types improved, but the spatial detail of the classification approached 10 m in some areas.

Mapping the invasive species, Chinese tallow, with EO1 satellite Hyperion hyperspectral image data and relating tallow occurrences to a classified Landsat Thematic Mapper land cover map

Our objective was to provide a realistic and accurate representation of the spatial distribution of Chinese tallow (Triadica sebifera) in the Earth Observing 1 (EO1) Hyperion hyperspectral image coverage by using methods designed and tested in previous studies. We transformed, corrected, and normalized Hyperion reflectance image data into composition images with a subpixel extraction model. Composition images were related to green vegetation, senescent foliage and senescing cypress‐tupelo forest, senescing Chinese tallow with red leaves (‘red tallow’), and a composition image that only corresponded slightly to yellowing vegetation. These statistical and visual comparisons confirmed a successful portrayal of landscape features at the time of the Hyperion image collection. These landscape features were amalgamated in the Landsat Thematic Mapper (TM) pixel, thereby preventing the detection of Chinese tallow occurrences in the Landsat TM classification. With the occurrence in percentage of red tallow (as a surrogate for Chinese tallow) per pixel mapped, we were able to link dominant land covers generated with Landsat TM image data to Chinese tallow occurrences as a first step toward determining the sensitivity and susceptibility of various land covers to tallow establishment. Results suggested that the highest occurrences and widest distribution of red tallow were (1) apparent in disturbed or more open canopy woody wetland deciduous forests (including cypress‐tupelo forests), upland woody land evergreen forests (dominantly pines and seedling plantations), and upland woody land deciduous and mixed forests; (2) scattered throughout the fallow fields or located along fence rows separating active and non‐active cultivated and grazing fields, (3) found along levees lining the ubiquitous canals within the marsh and on the cheniers near the coastline; and (4) present within the coastal marsh located on the numerous topographic highs.

Monitoring the recovery of Juncus roemerianus marsh burns with the normalized difference vegetation index and Landsat Thematic Mapper data

Nine atmospherically corrected Landsat Thematic Mapper images were usedto generate mean normalized difference vegetation indices (NDVI) at 11burn sites throughout a coastal Juncus roemerianus marsh in St. MarksNational Wildlife Refuge, Florida. Time-since-burn, the time lapse from thedate of burn to the date of image collection, was related to variation inmean NDVI over time. Regression analysis showed that NDVI increasedfor about 300 to 400 days immediately after the burn, overshooting thetypical mean NDVI of a nonburned marsh. For about another 500 to 600days NDVI decreased until reaching a nearly constant NDVI of about0.40. During the phase of increasing NDVI the ability to predicttime-since-burn was within about ±60 days. Within the decreasingphase this dropped to about ±88 days.Examination of each burn site revealed some nonburn related influences onNDVI (e.g., seasonality). Normalization of burn NDVI by site-specificnonburn control NDVI eliminated most influences. However, differentialresponses at the site-specific level remained related to either storm impactsor secondary burning. At these sites, collateral data helped clarify theabnormal changes in NDVI. Accounting for these abnormalities,site-specific burn recovery trends could be broadly standardized into fourgeneral phases: Phase 1 – preburn, Phase 2 – initial recovery (increasingNDVI), Phase 3 – late recovery (decreasing NDVI), and Phase 4 – finalcoalescence (unchanging NDVI). Phase 2 tended to last about 300 to 500days, Phase 3 an additional 500 to 600 days, and finally reaching Phase 4,900 to 1,000 days after burn.

A whole image approach using field measurements for transforming EO1 Hyperion hyperspectral data into canopy reflectance spectra

To maximize the spectral distinctiveness (information) of the canopy reflectance, an atmospheric correction strategy was implemented to provide accurate estimates of the intrinsic reflectance from the Earth Observing 1 (EO1) satellite Hyperion sensor signal. In rendering the canopy reflectance, an estimate of optical depth derived from a measurement of downwelling irradiance was used to drive a radiative transfer simulation of atmospheric scattering and attenuation. During the atmospheric model simulation, the input whole‐terrain background reflectance estimate was changed to minimize the differences between the model predicted and the observed canopy reflectance spectra at 34 sites. Lacking appropriate spectrally invariant scene targets, inclusion of the field and predicted comparison maximized the model accuracy and, thereby, the detail and precision in the canopy reflectance necessary to detect low percentage occurrences of invasive plants. After accounting for artifacts surrounding prominent absorption features from about 400 nm to 1000 nm, the atmospheric adjustment strategy correctly explained 99% of the observed canopy reflectance spectra variance. Separately, model simulation explained an average of 88%±9% of the observed variance in the visible and 98%±1% in the near‐infrared wavelengths. In the 34 model simulations, maximum differences between the observed and predicted reflectances were typically less than ±1% in the visible; however, maximum reflectance differences higher than ±1.6% (<±2.3%) at more than a few wavelengths were observed at three sites. In the near‐infrared wavelengths, maximum reflectance differences remained less than ±3% for 68% of the comparisons (±1 standard deviation) and less than ±6% for 95% of the comparisons (±2 standard deviation). Higher reflectance differences in the visible and near‐infrared wavelengths were most likely associated with problems in the comparison, not in the model generation.

Mapping and improving frequency, accuracy, and interpretation of land cover change: classifying coastal louisiana with 1990, 1993, 1996, and 1999 Landsat thematic mapper image data

Landsat Thematic Mapper images and collateral data sources were used to classify the land cover of the Mermentau River Basin within the chenier coastal plain and the adjacent uplands of Louisiana, USA. Landcover classes followed that of the National Oceanic and Atmospheric Administrations Coastal Change Analysis Program; however, classification methods needed to be developed to meet these national standards. Our first classification was limited to the Mermentau River Basin (MRB) in southcentral Louisiana, and the years of 1990, 1993, and 1996. To overcome problems due to class spectral inseparable, spatial and spectra continuums, mixed landcovers, and abnormal transitions, we separated the coastal area into regions of commonality and applying masks to specific land mixtures. Over the three years and 14 landcover classes (aggregating the cultivated land and grassland, and water and floating vegetation classes), overall accuracies ranged from 82% to 90%. To enhance landcover change interpretation, three indicators were introduced as Location Stability, Residence stability, and Turnover. Implementing methods substantiated in the multiple date MRB classification, we spatially extended the classification to the entire Louisiana coast and temporally extended the original 1990, 1993, 1996 classifications to 1999 (Figure 1). We also advanced the operational functionality of the classification and increased the credibility of change detection results. Increased operational functionality that resulted in diminished user input was for the most part gained by implementing a classification logic based on forbidden transitions. The logic detected and corrected misclassifications and mostly alleviated the necessity of subregion separation prior to the classification. The new methods provided an improved ability for more timely detection and response to landcover impact.