Lawrence W. Lass

A review of remote sensing of invasive weeds and example of the early detection of spotted knapweed (Centaurea maculosa) and babysbreath (Gypsophila paniculata) with a hyperspectral sensor

Abstract Remote sensing technology is a tool for detecting invasive species affecting forest, rangeland, and pasture environments. This article provides a review of the technology, and algorithms used to process remotely sensed data when detecting weeds and a working example of the detection of spotted knapweed and babysbreath with a hyperspectral sensor. Spotted knapweed and babysbreath frequently invade semiarid rangeland and irrigated pastures of the western United States. Ground surveys to identify the extent of invasive species infestations should be more efficient with the use of classified images from remotely sensed data because dispersal of an invasive plant may have occurred before the discovery or treatment of an infestation. Remote sensing data were classified to determine if infestations of spotted knapweed and babysbreath were detectable in Swan Valley near Idaho Falls, ID. Hyperspectral images at 2-m spatial resolution and 400- to 953-nm spectral resolution with 12-nm increments were used to identify locations of spotted knapweed and babysbreath. Images were classified using the spectral angle mapper (SAM) algorithm at 1, 2, 3, 4, 5, and 10° angles. Ground validation of the classified images established that 57% of known spotted knapweed infestations and 97% of known babysbreath infestations were identified through the use of hyperspectral imagery and the SAM algorithm. Nomenclature: Babysbreath, Gypsophila paniculata L. GYPPA; spotted knapweed, Centaurea maculosa Lam. CENMA.

Detecting Spotted Knapweed (Centaurea maculosa) with Hyperspectral Remote Sensing Technology1

Failure to detect noxious weeds with current survey methods prevents their control and has contributed to their ability to establish and spread in remote range and forest sites. Techniques used in remote sensing can classify plant occurrence on maps, offering a method for surveying invasive species in remote locations and across extensive areas. An imaging hyperspectral spectrometer recorded images on July 19, 1998 in Farragut State Park near Bayview, ID, in the reflected solar region of the electromagnetic spectrum ranging from 440 to 2,543 nm to detect spotted knapweed. The sensor records 128 spectral bands in 12- to 16-nm intervals at a spatial resolution of 5 m. A spectral angle mapper (SAM) algorithm was used to classify the data. Infestations in Idaho with 70 to 100% spotted knapweed cover that were 0.1 ha were detected regardless of the classification angle. However, narrow angles (2 to 8°) did not completely define the extent of the infestation, and the widest angle tested (20°) falsely classified some areas as infested. The overall image error for all classes was lowest (3%) when SAM angles ranged from 10 to 11°. Specific errors for the spotted knapweed class for the 10 to 11° angles showed that omissional and commissional errors were less than 3%. Areas with as little as 1 to 40% spotted knapweed cover were detected with an omissional error of 1% and a commissional error of 6%. Further verification sites were established on August 11, 1998 near Bozeman, MT, using the algorithms developed for Idaho. The omissional error for the Montana sites was 0%, and the commissional error was 10%. The hyperspectral sensor, Probe 1, proved an effective detection tool with the ability to detect spotted knapweed infestations. Nomenclature: Spotted knapweed, Centaurea maculosa Lam. #3 CENMA syn C. stoebe L. and C. biebersteinii DC. Additional index words: Hyperspectral sensor, imaging spectrometer, weed detection, whiskbroom scanner. Abbreviations: Ĉi, commissional error; DGPS, differentially corrected global positioning system; GPS, global positioning system without differential correction; L95, lower bounds expressed as 95% probability interval; Ôi, omissional error; SAM, spectral angle mapper; U95, upper bounds expressed as 95% probability interval.

Detecting the Locations of Brazilian Pepper Trees in the Everglades with a Hyperspectral Sensor1

Brazilian pepper is a small evergreen tree that forms dense colonies. It was introduced for horticultural use in the United States in the early 1800s and was widely distributed in Florida in the late 1920s. Previous remote-sensing projects to detect Brazilian pepper achieved moderate success and warranted additional research using a hyperspectral sensor. Detection with remote sensing is desirable because complete access to ground survey crews is not practical. The western half of the Everglades National Park was imaged at a 5-m spatial resolution with a hyperspectral sensor by Earth Search Sciences Inc. of Kalispell, MT, on December 12, 2000, and January 10, 2001. The sensor has 128 channels and spectral resolution between 450 and 2,500 nm. The purpose of this research was to develop spectral reflectance curves for Brazilian pepper and establish the accuracy of classified images. Classified images showed that a hyperspectral sensor could detect a “pure” Brazilian pepper pixel representing the center of an infestation but not “mixed” Brazilian pepper pixels at the sparsely populated edges. To define the sparse populations, images were classified using a spatial buffer (15- to 100-m radius) based on a low–omissional error image. A 25-m buffer reduced the amount of commissional error for Brazilian pepper in mangrove-dominated forest to 8.2% and buttonwood-dominated forest to 0%. Wider buffers did not significantly improve image accuracy when compared with the 25-m buffer distance. Results indicate that removal crews using hyperspectral images will be able to reliably find the colonies of Brazilian pepper but will not be able to use the images to find isolated scattered trees. Nomenclature: Brazilian pepper (Schinus terebinthifolius Raddi) #3 SCITE. Additional index words: Hyperspectral sensor, imaging spectrometer, invasive plant detection, whiskbroom scanner. Abbreviations: GPS, global positioning system without differential correction; PLOS, posteriori least-squares orthogonal subspace projection; SAM, spectral angle mapper; USGS-BRD/SERP, United States Geological Survey—Biological Resource Division/Southeast Environmental Research Program.

Assessing agreement in multispectral images of yellow starthistle (Centaurea solstitialis) with ground truth data using a Bayesian methodology.

Abstract: Digital imagery from satellites and airborne remote sensing offer an opportunity to accurately detect weed infestations. Image resolution and plant growth stage are critical factors for maximum weed detection with low errors. Data analysis in traditional image assessment has relied on agreement measures, such as Cohens kappa and asymptotic procedures, that compare what is on the image but not on the ground and what is on the ground but not on the image. Statistical comparisons of multispectral images, however, require some knowledge of the variability of the image classification results to determine significant differences among agreement measures. Bayesian methods were used to develop probability distributions for an agreement measure, conditional kappa, and were then subsequently applied to assess and compare image resolutions and plant growth stages. Results showed that images of a study site known to have yellow starthistle populations could identify the noninfested areas with greater accuracy than infested areas at spatial resolutions of 0.5, 1.0, 2.0, and 4.0 m. The detection accuracy of yellow starthistle in the images taken either prebloom or at flowering with 4.0-m spatial resolution usually was equal to or better than spatial resolutions of 0.5, 1.0, and 2.0 m for the cover classes that were not, moderately (31 to 70%), and highly (71 to 100%) infested. The 0.5-m resolution was better than 4.0-m spatial resolution when detecting the moderate cover class, but both resolutions had high omissional and commissional errors. Contrasting the best detection resolution for finding yellow starthistle colonies across flight times indicated that flying at flowering stage with the 4.0-m spatial resolution provided the best detection of the yellow starthistle cover classes considered. In the cases where different spatial resolutions resulted in equal detection accuracy, the larger spatial resolution was selected because of lower costs of acquiring and processing the data. Nomenclature: Yellow starthistle, Centaurea solstitialis L. #3 CENSO. Additional index words: Airborne charge-coupled devices, conditional kappa, image analysis, remote sensing, videography, weed detection. Abbreviations:K̂i, conditional kappa, where i denotes the image classification category (high, moderate, and none).

Predicting the likelihood of yellow starthistle (Centaurea solstitialis) occurrence using landscape characteristics

Abstract Yellow starthistle is an invasive plant species common in the semiarid climate of central Idaho and other western states. Early detection of yellow starthistle and estimation of its infestation potential in semiarid grasslands have important scientific and managerial implications. Weed detection and delineation of infestations are often carried out by using ground survey techniques. However, such methods can be inefficient and expensive in detecting sparse infestations. The distribution of yellow starthistle over a large region may be affected by various landscape variables such as elevation, slope, and aspect. These exogenous variables may be used to develop prediction models to estimate the potential for yellow starthistle invasion into new areas. A nonlinear prediction model has been developed using a polar coordinate transformation of landscape characteristics to predict the likelihood of yellow starthistle occurrence in north-central Idaho. The study region included the lower Snake River and parts of the Salmon and Clearwater basins encompassing various land-use (range, pasture, and forest) categories. The model provided accurate estimates of yellow starthistle incidence within each specified land-use category and performed well in subsequent statistical validations. This prediction model can assist land managers in focusing their efforts by identifying specific areas for survey. Nomenclature: Yellow starthistle, Centaurea solstitialis L. CENSO.

Using landscape characteristics as prior information for Bayesian classification of yellow starthistle

Abstract Yellow starthistle is an invasive plant of canyon grasslands in north-central Idaho. The distribution of yellow starthistle is associated with general landscape characteristics that include land use and specific terrain-related features such as elevation, slope, and aspect. Slope and aspect can be considered as indicators of plant community composition and distribution. Hence, these variables may be incorporated into prediction models to estimate the likelihood of yellow starthistle occurrence because plant communities differ in susceptibility to invasion. An empirically derived nonlinear model based on landscape characteristics has previously been developed to predict the likelihood of yellow starthistle occurrence in north-central Idaho. Although the model was used to predict the invasion potential of yellow starthistle into new areas, it could also be used as auxiliary data for classifying this weed species in remotely sensed imagery. To accomplish this, the predicted values from the model are regarded as prior information on the presence of yellow starthistle. A Bayesian image classification algorithm using this prior information is then applied to a corresponding set of remotely sensed data. This results in a map indicating the posterior probabilities of yellow starthistle occurrence given the landscape characteristics. This technique is demonstrated and is shown to reduce omissional error rates by 50% when the landscape characteristics are incorporated into the classification process. Nomenclature: Yellow starthistle, Centaurea solstitialis L. CENSO.

ASSESSING VARIABILITY OF AGREEMENT MEASURES IN REMOTE SENSING USING A BAYESIAN APPROACH

Remote sensing imagery is a popular accessment tool in agriculture, forestry, and rangeland management. Spectral classification of imagery provides a means of estimating production and identifYing potential problems, such as weed, insect, and disease infestations. Accuracy of classification is traditionally based on ground truthing and summary statistics such as Cohens Kappa. Variability assessment and comparison of these quantities have been limited to asymptotic procedures relying on large sample sizes and gaussian distributions. However, asymptotic methods fail to take into account the underlying distribution of the classified data and may produce invalid inferential results. Bayesian methodology is introduced to develop probability distributions for Cohens Conditional Kappa that can subsequently be used for image assessment and comparison. Techniques are demonstrated on a set of images used in identifYing a species of weed, yellow starthistle, at various spatial resolutions and flying times.

Development of a Dispersal Model for Balsam Woolly Adelgid, Adelges piceae Ratzeburg (Hemiptera: Adelgidae), to Facilitate Landscape-Level Management Planning

The balsam woolly adelgid (Adelges piceae Ratzeburg) attacks subalpine fir (Abies lasiocarpa (Hook.) Nutt.) in eastern Washington, Oregon, and northern Idaho. Historical balsam woolly adelgid distributions present an opportunity to understand climatic factors that influence the species’ distribution at a landscape scale. The distribution data allows for creation of predictive models that detail the likelihood of occurrence and associated geographic data allow modeling of species dispersal. Predictive variables linked to the distribution of the hosts and to abiotic environmental conditions were utilized to create a spatial probability model of occurrence. Balsam woolly adelgid predominantly disperses by wind, and hence, both wind speed and wind direction were used to create a dispersal probability model. Results from wind dispersal modeling suggested that two-thirds of the new infestations were due to July and August wind direction and speed. Average July winds ranged from 0.5 to 3.27 m/s, flowing south westerly, and August winds ranged from 0.43 to 1.55 m/s, flowing north easterly. Land managers can use the results of the predictive model to better understand where current infestations are likely to expand. Prediction of where the balsam woolly adelgid might move allows managers to adjust actions to respond to future insect movement and establishment.

PREDICTION OF YELLOW STARTHISTLE SURVIVAL AND MOVEMENT OVER TIME AND SPACE

Yellow starthistle is a noxious weed that has become a serious plant pest with devastating impact on ranching operation and natural resources in western states. Early detection of yellow starthistle and predicting its spread has important managerial implications and greatly reduce the economic losses due to this weed. The dispersal of yellow starthistle consists of two main components, plant survival and seed movement. Resources and direct factors relating to these components are not typically available or are difficult to obtain. Alternatively, topographic factors, such as slope, aspect and elevation, are readily available and can be related to plant survival and seed movement. In this study, several GIS network models incorporating these topographic factors are considered for the prediction of yellow starthistle spread. The models differed in their assessment of the costs of movement derived from these factors. Models were evaluated based on their predictive ability and residual analysis. The optimal model gave an accurate estimate of the dispersal boundary for the study area. Further validation of the estimated model using an independent data set from a larger area also verified its predictive capability.

MODELING DISPERSAL OF YELLOW STARTHISTLE IN THE CANYON GRASSLANDS OF NORTH CENTRAL IDAHO

Yellow starthistle is an invasive plant species that reduces productivity and plant diversity within the canyon grasslands of Idaho. Early detection of yellow starthistle and predicting its spread have important managerial implications that could greatly reduce the economic/environmental losses due to this weed. The spread of an invasive plant species depends on its ability to reproduce and disperse seed into new areas. Typically, information on the factors that directly affect a plant’s ability to reproduce and subsequently disperse seed is not available or difficult to obtain. Alternatively, topographic factors, such as slope and aspect as well as competitive correlates such as vegetation indices related to plant community biomass could be used to model plant survival and seed movement. In this research, several spatial network models incorporating these variables were considered for the prediction of yellow starthistle dispersal. Models will differed in their assessment of plant movement costs, which can be separated into two processes, survival to reproduction and seed dispersal. The candidate models were evaluated based on their predictive ability and biological relevance. Topographical variables, slope and aspect, were found to be significant contributors to yellow starthistle dispersal models, whereas vegetation indices did not improve the prediction process. The optimal model was applied to an area in central Idaho for predicting the dispersal of yellow starthistle in 1987 given a known 1981 infestation.