Search

Connecting forest management with the conservation of forest-associated animals requires an understanding of habitat quality, as well as identifying long-term silvicultural strategies that align with high quality habitat. It is, therefore, essential to characterize the spatio-temporal dimensions of habitat quality. Here, we leveraged multiple datasets to assess high quality habitat for female Canada lynx (Lynx canadensis), a federally threatened forest carnivore in the contiguous U.S. Our datasets included a spatially extensive sample of snowshoe hares (Lepus americanus) collected in 2013 (n = 1340 plots), an expansive time-series (i.e., 1972–2013) of forest structural classes derived from remote sensing, and a longitudinal dataset where we monitored habitat use and the reproductive success (i.e., litter of kittens present or absent) of female Canada lynx during 1999–2013 (n = 32 female lynx over 92 lynx years). Our results indicated that the probability of a female producing kittens was most associated with the connectivity of mature, multistoried forests (composed of mostly spruce-fir). However, the variation among female lynx accounted for ≈62% of the total variation explained in litter production, suggesting substantial individual-level variation. Thus, managers can contribute to increased reproductive success of female Canada lynx by facilitating the development of mature forests, but measuring that success will be difficult given the individual variation. In core areas of high quality females (i.e., produced kittens frequently), mature forest was 17% more abundant (i.e., ≈60% of the total core area), more connected, less clumpy, and exhibited 2.25-times larger patch sizes than the core areas of low quality females. At the home-range extent, patterns were less pronounced while the abundance of mature forests remained high (≈50%) for high quality females. Additionally, we demonstrated that the relative density of snowshoe hares was ≥2.8 times higher in advanced regenerating forests compared to all other structural classes, including mature forest. Advanced regenerating forests accounted for ≈18–19% of the core area and home range of high quality female lynx. Combined, our results suggest that a high quality mosaic for female Canada lynx contains ≈50–60% mature forest and ≈18–19% advanced regenerating forest. Furthermore, we used Forest Inventory and Analysis data to characterize the approximate age distribution of advanced regeneration and mature forest, which was relevant for rotation schedules of forest silviculture. Results indicated that advanced regeneration was ≈20 to 80 years old while mature forest was ≈50 to ≥200 years old. Our results provide novel insight into how forest management could increase habitat quality for female Canada lynx, and suggest that multiple silvicultural methods (e.g., intermediate treatments, regeneration harvests) could be employed to maintain a forest mosaic that enhances the ability of females to produce kittens. We concluded by providing a framework that integrates our new insights into a management context with the aim of conserving Canada lynx on multiple-use lands.

Forest managers are often tasked with balancing opposing objectives, such as altering forest structure and conserving forest-dwelling animals. Consequently, to develop holistic strategies managers require information on how forest manipulations influence species of conservation concern, particularly those that are federally threatened or endangered. Here, we characterized how differing silvicultural treatments (n = 1,293 – forest thinnings; removal of small trees, selection cuts; trees harvested in small patches, and regeneration cuts; clearcuts of nearly all trees) influenced the resource use of a threatened forest carnivore, Canada lynx (Lynx canadensis), over a temporal gradient of 1–67 years after treatment. To do this, we used an extensive GPS dataset on 66 Canada lynx (i.e., 164,593 locations) collected during 2004–2015 within the Northern Rocky Mountains, U.S. We used univariate analyses and hurdle regression models to evaluate the spatio-temporal factors influencing lynx use of treatments. Our analyses indicated that Canada lynx used treatments, but there was a consistent cost in that lynx use was low up to ∼10 years after all silvicultural actions. However, cumulative use (in both winter and summer) by lynx reached 50% at ∼20 years after a thinning treatment, whereas it took ∼34–40 years after a selection or regeneration cut. This indicated that Canada lynx used thinnings at a faster rate post-treatment than selection or regeneration cuts, and that lynx used selection and regeneration cuts in a similar fashion over time. Further, we discovered that lynx occupancy and intensity of treatment use was influenced by the composition of forest structure in the surrounding neighborhood. In some instances, the existing forest structure surrounding the treatment and the time since treatment interactively influenced lynx use; a pattern characterizing a spatio-temporal functional response in habitat use. This demonstrated that both the recovery time as well as the spatial context of a particular area are important considerations when implementing different silvicultural treatments for Canada lynx at the landscape scale. For example, if a selection cut was implemented with abundant mature, multi-storied forest (i.e., a preferred habitat by lynx) in the surrounding landscape, lynx would use these treatments less over time than if the neighborhood contained less mature forest. Forest managers can apply our spatio-temporal understandings of how lynx respond to forest silviculture to refine expectations and develop strategies aimed at both forest management and the conservation of Canada lynx.

Forests worldwide are impacted by a wide variety of disturbances that are happening more frequently with more intensity than in the past due to global climate change. Forest managers, therefore, need to identify new ways to quickly and accurately predict post-disturbance forest landscape composition. We suggest the use of Landsat satellite imagery and an image processing tool to map percent canopy cover (PCC) by species and sub-canopy species counts to be used in adaptive forest management strategies. We used zero-inflated models to successfully predict PCC and sub-canopy counts (number of regenerating trees per pixel, also called biotic legacies) for 4 tree species, along with overall PCC and percent mortality, for a large portion of the Rio Grande National Forest (RGNF) in 2013. The RGNF had recently been disturbed by spruce beetle (Dendroctonus rufipennis) infestation since the early 2000s and the West Fork Fire Complex in 2013. Our PCC models resulted in pseudo median differences between observed and predicted values of 0.2–6.5%, RMSE of 10.9–17.0%, and 95% confidence interval widths of 4.4–24.9%, depending on the species. The percent mortality model resulted in pseudo median differences between observed and predicted values of 1.1%, RMSE of 12.4%, and 95% confidence interval width of 4.6%. The sub-canopy PCC model resulted in a pseudo median differences between observed and predicted values of 1.3%, RMSE of 9.4%, and 95% confidence interval of 3.0%. The sub-canopy count models resulted in mean differences of 0.1–1.4 trees, RMSE of 3.0–13.4 trees, and 95% confidence interval widths of 1.1–5.0 trees, depending on species. By mapping PCC and sub-canopy counts, we have provided forest managers with knowledge of the current surviving forest (PCC) as well as the biotic legacies (sub-canopy counts) that can aid in forming hypotheses as to what the forest might become in the future, adding to the forest manager toolbox for forest management strategies. The methods described can be applied to a variety of issues within the field of disturbance ecology and, combined with change analyses, will provide forest managers with empirical evidence of current and future forest composition along with biological legacies that will impact forest regeneration.

Flight-based hyperspectral imaging systems have the potential to provide valuable information for ecosystem and environmental studies, as well as aid in land management and land health monitoring. This paper examines a series of images taken over the course of three years that were radiometrically referenced allowing for quantitative comparisons of changes in vegetation health and land usage. The study area is part of a geologic carbon sequestration project located in north-central Montana, approximately 580 ha in extent, at a site requiring permission from multiple land owners to access, making ground based validation difficult. Classification based on histogram splitting of the biophysically based parameters utilizing the entire three years of data is done to determine the major classes present in the data set in order to show the constancy between data sets taken over multiple years. Additionally, a method of anomaly detection for both single and multiple data sets, using Median Absolute Deviations (MADs), is presented along with a method of determining the appropriate size of area for a particular ecological system. Detection of local anomalies within a single data set is examined to determine, on a local scale, areas that are different from the surrounding area and depending on the specific MAD cutoff between 50–70% of the anomalies were located. Additionally, the detection and identification of persistent (anomalies that occur in the same location over multiple data sets) and non-persistent anomalies was qualitatively investigated.

Low-cost flight-based hyperspectral imaging systems have the potential to provide important information for ecosystem and environmental studies as well as aide in land management. To realize this potential, methods must be developed to provide large-area surface reflectance data allowing for temporal data sets at the mesoscale. This paper describes a bootstrap method of producing a large-area, radiometrically referenced hyperspectral data set using the Landsat surface reflectance (LaSRC) data product as a reference target. The bootstrap method uses standard hyperspectral processing techniques that are extended to remove uneven illumination conditions between flight passes, allowing for radiometrically self-consistent data after mosaicking. Through selective spectral and spatial resampling, LaSRC data are used as a radiometric reference target. Advantages of the bootstrap method include the need for minimal site access, no ancillary instrumentation, and automated data processing. Data from two hyperspectral flights over the same managed agricultural and unmanaged range land covering approximately 5.8 km2 acquired on June 21, 2014 and June 24, 2015 are presented. Data from a flight over agricultural land collected on June 6, 2016 are compared with concurrently collected ground-based reflectance spectra as a means of validation.

Hyperspectral image analysis has benefited from an array of methods that take advantage of the increased spectral depth compared to multispectral sensors; however, the focus of these developments has been on supervised classification methods. Lack of a priori knowledge regarding land cover characteristics can make unsupervised classification methods preferable under certain circumstances. An unsupervised classification technique is presented that utilizes physically relevant basis functions to model the reflectance spectra. These fit parameters used to generate the basis functions allow clustering based on spectral characteristics rather than spectral channels and provide both noise and data reduction. Histogram splitting of the fit parameters is then used as a means of producing an unsupervised classification. Unlike current unsupervised classification techniques that rely primarily on Euclidian distance measures to determine similarity, the unsupervised classification technique uses the natural splitting of the fit parameters associated with the basis functions creating clusters that are similar in terms of physical parameters. The data set used in this work utilizes the publicly available data collected at Indian Pines, Indiana. This data set provides reference data allowing for comparisons of the efficacy of different unsupervised data analysis. The unsupervised histogram splitting technique presented in this paper is shown to be better than the standard unsupervised ISODATA clustering technique with an overall accuracy of 34.3/19.0% before merging and 40.9/39.2% after merging. This improvement is also seen as an improvement of kappa before/after merging of 24.8/30.5 for the histogram splitting technique compared to 15.8/28.5 for ISODATA.

In 2007, common vampire bats were the source of the first outbreak of paralytic bovine rabies in Uruguay. The outbreak coincided in space and time with the fragmentation of native grasslands for monospecific forestry for wood and cellulose production. Using spatial analyses, we show that the increase in grassland fragmentation, together with the minimum temperature in the winter, accounts for the spatial pattern of outbreaks in the country. We propose that fragmentation may increase the connectivity of vampire bat colonies by promoting the sharing of feeding areas, while temperature modulates their home range plasticity. While a recent introduction of the virus from neighboring Brazil could have had an effect on outbreak occurrence, we show here that the distribution of rabies cases is unlikely to be explained by only an invasion process from Brazil. In accordance with previous modeling efforts, an increase in connectivity may promote spatial persistence of rabies virus within vampire bat populations. Our results suggest that land use planning might help to reduce grassland fragmentation and thus reduce risk of rabies transmission to livestock. This will be especially important in the context of climatic changes and increasing minimum temperatures in the winter.

Recent developments for acquiring and distributing remotely-sensed data have greatly increased data availability to the user community. The past two decades have witnessed an explosion in data acquisition by a variety of ground, airborne and orbital sensors. The popularization of Unmanned Aerial Systems (UAS) and the development of reduced cost orbital platforms should guarantee that even higher data volumes will be available to future analysts. The past decades also saw the opening of image data archives (e.g., Landsat, CBERS, Sentinel), making access to a rich database of moderate resolution satellite images a reality across the globe. This increased volume and variety of remotely-sensed data increases the demand for methods and procedures for data handling and information extraction. This chapter describes recent efforts to expand the analyst’s data processing toolset and includes the theory and strategies used in manipulating remotely-sensed data by digital systems. The text focuses on presenting algorithms and techniques for image processing and analysis and emphasizes recent developments not covered by previous editions of the ASPRS Manual of Remote Sensing. Although the main topics covered by the chapter involve the direct processing of images, the text also covers concepts involved in processing remote sensing data that may not have been collected or stored as images, such as spectral curves acquired by spectroradiometers. Several sections of this chapter match this description, including Spectral Vegetation Indices and Spectral Mixture Analysis. Image processing includes not only the analysis of images, but also the necessary steps involved in preparing images for analysis, such as geometric correction, atmospheric correction and several techniques associated with image enhancement. Spectral indices resulting from the combination of multiple spectral bands are presented, with emphasis on the description of vegetated targets. A detailed treatment is given to the mixture problem resulting from the contribution of multiple materials within the instantaneous field of view (IFOV) of a given sensor. Because multiple applications can benefit from the increased explanation power provided by a large number of spectral bands, hyperspectral data processing is also presented and discussed. Further, the chapter addresses the benefits and challenges involved in combining datasets acquired by different systems (Data Fusion). Image classification addresses multiple strategies involved in assigning classes to images (e.g., Support Vector Machine, and Decision Trees); and includes advances in Object-Based Image Analysis (OBIA), particularly those related to image segmentation in preparation for classification. Given the increasing length of remotely-sensed data time series, particular attention is given to preparing sequences of images and data, including multiple techniques for smoothing, spike removal and the retrieval of metrics associated with temporal variations of targets. The chapter also brings multiple examples of use of products derived from processing remotely-sensed data as input to a variety of workflows, including modeling and analysis efforts. Finally, very current topics involving recent advances in image acquisition and availability, are presented for generating 3D surfaces from multiple images using Structure from Motion (SfM); processing of very large datasets (Big Data); and processing of images in the cloud are presented.

There is a pressing need to map changes in forest structure from the earliest time period possible given forest management policies and accelerated disturbances from climate change. The availability of Landsat data from over four decades helps researchers study an ecologically meaningful length of time. Forest structure is most often mapped utilizing lidar data, however these data are prohibitively expensive and cover a narrow temporal window relative to the Landsat archive. Here we describe a technique to use the entire length of the Landsat archive from Multispectral Scanner to Operational Land Imager (M2O) to produce three novel outcomes: (1) we used the M2O dataset and standard change vector analysis methods to classify annual forest structure in northwestern Montana from 1972 to 2015, (2) we improved the accuracy of each yearly forest structure classification by applying temporal continuity rules to the whole time series, with final accuracies ranging from 97% to 68% respectively for two and six-category classifications, and (3) we demonstrated the importance of pre-1984 Landsat data for long-term change studies. As the Landsat program continues to acquire Earth imagery into the foreseeable future, time series analyses that aid in classifying forest structure accurately will be key to the success of any land management changes in the future.

Evaluation of classification methods, whether in connection with the development of new methods or in an application setting, has been hampered by the lack of availability of adequate data and an approach for comparisons. We collected 30 mostly moderate-resolution, multispectral datasets to enable statistically rigorous comparisons of methods and have made those datasets available for other researchers. We developed a methodological approach to comparing classification methods and demonstrated the approach using six methods, C5.0, classification tree analysis, logistic model trees, multivariate adaptive regression splines, random forest, and support vector machines. We also demonstrated how these data and this approach can be used to address specific questions in addition to overall accuracy performance, including the relative effects of using derived components and ancillary data and the relative success in classifying rare classes. Most methods performed best by at least one metric with at least one dataset. Therefore, although random forest on average performed statistically significantly better than the other methods tested, we do not recommend this method as the sole option currently in remote sensing. Rather, our results suggest that remote sensing analysts should evaluate multiple methods with respect to any classification project, which can be accomplished through statistical software packages.

Ecological and land management applications would often benefit from maps of relative canopy cover of each species present within a pixel, instead of traditional remote-sensing based maps of either dominant species or percent canopy cover without regard to species composition. Widely used statistical models for remote sensing, such as randomForest (RF), support vector machines (SVM), and generalized linear regression (GLM), are problematic for this purpose as they often fail to properly predict the absence of a target species, especially in areas of high vegetation diversity, due to the relative abundance of absence observations (or zero values) in the reference data used to train predictive models. Experience has shown that RF, SVM, and GLM models trained on such reference data produce biased values of PCC, for example, in forested areas absent the target species, PCC is overestimated, while in forested areas where a target species PCC is abundant, PCC tends to be underestimated. We used zero-inflated regression modeling to reduce such bias and better predict PCC-by-species within each pixel in mixed conifer forests. Zero-inflated regression models use a two-step process to first predict the presence or absence of the target species, and then to predict continuous levels of PCC only where the target species is present. We compared the results of three widely used methods (RF, SVM, and GLM) to nine zero-inflated models for their ability to predict continuous PCC for each of five different conifer species in heterogeneous forests of northwestern Montana using Landsat TM and OLI imagery. Our best zero-inflated models resulted in a mean difference of − 3.84% to 2.26%, 95% confidence interval of 6.22% to 13.09%, and RMSE of 11.26% to 22.98%, depending on the species. The success of the zero-inflated model was robust across methods tested. Both the zero-inflated and traditional methods were successful in estimating continuous canopy cover, however, the traditional models showed a substantial bias by never correctly predicting the absence of the target species, while the zero-inflated models correctly predicted species absence 57% to 84% of the time, depending on the species. Visual inspection of the predicted maps compared to high-resolution imagery demonstrated that the zero-inflated models also more closely matched the landscape, as traditional models more often incorrectly predicted canopy cover in non-forested areas. Using the zero-inflated process dramatically reduced the bias of the results, allowing end users to make management decisions with increased confidence about where a target species is absent, something not possible with the traditional methods tested.

Ecologists and managers are motivated to predict the distribution of animals across landscapes as well as understand the mechanisms giving rise to that distribution. Satisfying this motivation requires an integrated framework that characterizes multi-scale habitat use and selection, as well as builds predictive models such as resource selection functions. However, the assumption of constant habitat use or selection is often made in such analyses, which ignores the possibility that individuals experiencing different conditions might respond differently. Assessing functional responses in habitat use evaluates how animal behavior changes with differing environmental conditions, which has basic and applied utility. Here, we combined these ideas into an integrated process that characterizes habitat relationships, predicts habitat, and assesses behavioral differences with changing environmental conditions. Our species of interest was Canada lynx (Lynx canadensis) in the Northern Rocky Mountains, which is a rare and federally threatened forest carnivore. Through our process, we developed multi-scale predictions of lynx distribution and learned that across scales and seasons, lynx use more mature, spruce-fir forests than any other structure stage or species. Intermediate snow depths and the distribution of snowshoe hares (Lepus americanus) were the strongest predictors of where lynx selected their home ranges. Within their home ranges, female and male lynx increasingly used advanced regeneration forest structures as they became more available (up to a maximum availability of 40%). These patterns supported the bottom-up mechanisms regulating Canada lynx in that advanced regeneration generally provides the most abundant snowshoe hares, while mature forest is where lynx appear to hunt efficiently. However, lynx exhibited decreasing use of stand initiation structures (up to a maximum availability of 25%). Land managers have an opportunity to promote lynx habitat in the form of advanced regeneration, but are required to go through the stand initiation phase. Thus, managers can apply the relative proportions of forest structure classes along with our response curves to inform landscape actions (e.g., timber harvest) targeted at facilitating the forest mosaic used and selected by Canada lynx. Collectively, the insights gleaned from our approach advance habitat conservation efforts and consequently are of broad utility to applied ecologists and managers.

Producers make the decision to adopt a particular agricultural practice within a range of social, economic, environmental, and agronomic constraints. The semiarid regions of the US northern Great Plains are dominated by dryland farming practices and the traditional practice has been to rotate small-grain cereals with summer fallow; however, producers are moving away from this practice. The area of fallow in northeastern Montana decreased by one-third and the area of pulse crops increased nearly six-fold during 2001–2012. We previously identified two key practices that are indicative of regionally changing agricultural practices: (1) the broad-scale adoption of cereal–pulse sequences, and (2) the conversion from continuous strip-cropping to block managed cereal-based sequences. Here, we examined the adoption of these two practices from a spatio-temporal perspective to determine if the observed patterns were consistent with those expected from a priori processes: random occurrence, spread and adoption of the practices due to social interaction as described in innovation diffusion theory, or adoption based on environmental factors. Our results suggest that the adoption and spread of both practices were likely constrained by the suitability of the physical environment. Available water, in particular, exerts a fundamental control on the decision whether or not to adopt either practice. We also found evidence for the expansion of these practices due, in part, to social factors, particularly during the early period of adoption. We conclude that producers made the decision whether or not to adopt these practices primarily as a function of environmental suitability and, to a lesser extent, within the context of social interactions.

Biomass from land enrolled in the Conservation Reserve Program (CRP) is being considered as a biofuel feedstock source. A quick, accurate and nondestructive method to estimate biomass yield would be valuable for land managers to ensure sustainable production. The purpose of this study was to compare the ability of regression models to estimate biomass yields using data from satellite and ground based remote sensing platforms. Biomass yields and plant spectral responses were obtained at different phenological stages over two growing seasons (2011e2012) on an 8.1 ha CRP pasture in central Montana. Regression models were constructed using the normalized difference vegetation index (NDVI) and various band combinations from a hand held Crop Circle sensor and from Landsat satellite images. All models showed reasonable accuracy in estimating biomass, with a difference of 0.05) between actual and estimated biomass. Results suggest that the usefulness of the spectral regions is a function of phenological growth stage. Red, red edge, and the near-infrared bands are more responsive at boot and peak growth stages while bands in the short-wave infrared increased the accuracy for the dormant stage biomass estimations. Land managers may construct spectral models to more effectively manage biomass resources.

Farmers implement an assortment of management practices to ensure the sustainability, economic viability, and resilience of their operation. Dryland farming practices dominate in the semiarid regions of the US northern Great Plains, where historical practice has been to rotate small-grain cereals with whole-year summer fallow; however, pulse crops (e.g., lentils) have become increasingly common in these regions as an alternative to fallow. The area of fallow in northeastern Montana, for example, has decreased by one-third, while the area of pulse crops has increased more than five-fold. Our objectives were: (1) to characterize the principal cropping sequences in northeast Montana during the period of regional pulse crop adoption (2001–2012); and (2) to identify changes in the relative proportions of these sequences during the same period. We identified crops at the field-level by class (cereal, pulse, fallow, or cereal–fallow strips) for 2001–2012 using multitemporal Landsat imagery in conjunction with the cropland data layer, cadastral data, ground reference data, and local producers’ records. The annual crop classifications were combined into a 12-character string for each field that represented the sequence of crop classes for 2001–2012. We then searched these strings for specific 2- and 3-year crop sequences with a string-matching algorithm. The most abundant sequences involved continuous cereal, block-managed cereal–fallow, and cereal–pulse. We also observed a steady decrease in the abundance of cereal–fallow sequences managed by strip-cropping that were coincident with increases in block-managed cereal–fallow sequences and with increases in pulse production. We conclude that, over the study's time frame, regional producers grew more cereal crops and fallowed fields less frequently, but did not appear to strongly adhere to specific sequences. Furthermore, strip-cropping as a management practice has declined substantially.

Recent research demonstrated that ${\mmb CO}_{\mmb 2}$ gas leaking from underground can be identified by observing increased stress in overlying vegetation using spectral imaging. This has been accomplished with both visible/near-infrared (Vis/NIR) sunlight reflection and long-wave infrared (LWIR) thermal emission. During a 4-week period in summer 2011, a controlled ${\mmb CO}_{\mmb 2}$ release experiment was conducted in Bozeman, Montana, as part of a study of methods for monitoring carbon sequestration facilities. As part of this experiment, reflective and emissive imagers were deployed together to enable a comparison of these two types of imaging systems for vegetation-based ${\mmb CO}_{\mmb 2}$ leak detection. A linear regression was performed using time as the response variable with red and NIR reflectances, Normalized Difference Vegetation Index (NDVI), and LWIR brightness temperature as predictors. The regression study showed that the reflectance and LWIR brightness temperature data together explained the most variability in the data (96%), equal to the performance of the Vis/NIR reflectance data alone, followed by NDVI alone (90%), and LWIR data alone (44%). Therefore, the two types of imagers contributed in a synergistic fashion, while either method alone was capable of gas detection with increased statistical variability.

The utility of Enhanced Thematic Mapper Plus (ETM+) has been diminished since the 2003 scan-line corrector (SLC) failure. Uncorrected images have data gaps of approximately 22% and gap-filling schemes have been developed to improve their usability. We present a method to classify a northeast Montana agricultural landscape using ETM+ SLC-off imagery without gap-filling. We use multitemporal data analysis and employ an object-oriented approach to define objects, agricultural fields, with cadastral data. This approach was assessed by comparison to a pixel-based approach. Results indicate that an ETM+ SLC-off image can be classified with better than 85% overall accuracy without gap-filling.

Airborne remote sensing has been proposed as a potential large-area monitoring tool for geologic carbon sequestration (GCS) sites. Elevated soil CO 2 levels from an underground CO 2 leak could induce a plant stress response that is spectrally discernible from the air. A controlled subsurface CO 2 release experiment was conducted during the growing season at the outdoor Zero Emissions Research and Technology (ZERT) center to simulate a CO 2 leak scenario. Simultaneously, aerial imagery was collected to obtain a time series used to identify and characterize the simulated CO 2 leak prior to, during, and after the three week CO 2 release. A theoretical framework was developed for analytical strategies that could be implemented to detect a CO 2 leak using aerial hyperspectral imagery with minimal a priori knowledge of when and where a surface leak is going to occur. Areas of inferred CO 2 stressed vegetation were identified using an unsupervised clustering algorithm. The spectral signatures of this vegetation informed the development of a red edge index (REI) used to quantify the CO 2 stress signal and chart vegetation health trajectories over the course of the CO 2 release experiment. REI was found to be significantly lower (Welch's p -value 2 stressed vegetation as compared to healthy vegetation. Furthermore, maximum differences in REI ( REI healthy vegetation − REI CO 2 stressed vegetation ) were observed at the height of the CO 2 release followed by a subsequent decrease in differences once CO 2 injection ceased. These results suggest that there was a cumulative vegetation stress response followed by a possible vegetation recovery and that aerial hyperspectral imaging may be a plausible method for detecting CO 2 leaks from GCS sites.

Landsat data are increasingly used for ecological monitoring and research. These data often require preprocessing prior to analysis to account for sensor, solar, atmospheric, and topographic effects. However, ecologists using these data are faced with a literature containing inconsistent terminology, outdated methods, and a vast number of approaches with contradictory recommendations. These issues can, at best, make determining the correct preprocessing workflow a difficult and time-consuming task and, at worst, lead to erroneous results. We address these problems by providing a concise overview of the Landsat missions and sensors and by clarifying frequently conflated terms and methods. Preprocessing steps commonly applied to Landsat data are differentiated and explained, including georeferencing and co-registration, conversion to radiance, solar correction, atmospheric correction, topographic correction, and relative correction. We then synthesize this information by presenting workflows and a decision tree for determining the appropriate level of imagery preprocessing given an ecological research question, while emphasizing the need to tailor each workflow to the study site and question at hand. We recommend a parsimonious approach to Landsat preprocessing that avoids unnecessary steps and recommend approaches and data products that are well tested, easily available, and sufficiently documented. Our focus is specific to ecological applications of Landsat data, yet many of the concepts and recommendations discussed are also appropriate for other disciplines and remote sensing platforms.

Multispectral vegetation reflectance measurements were used as an indirect method of sensing CO2 gas leaking from underground in a controlled release experiment in Bozeman, Montana, USA. The leak location is identified through time-series analysis of the reflectances and the normalized difference vegetation index (NDVI), evaluated at a test location and a control location. Vegetation reflectance changes that correlated with root-level CO2 exposure were distinguishable from changes attributed to seasonal factors including precipitation, wind, air temperature variation, etc. The NDVI of the vegetation became steadily smaller until saturating approximately twenty days after the beginning of the release. However, before reaching the threshold values, both reflectance and NDVI values changed more rapidly when exposed to elevated CO2 fluxes.

Yellowstone National Park (YNP) contains the world's largest concentration of geothermal features and is legally mandated to protect and monitor these natural features. Remote sensing is a component of the current geothermal monitoring plan. Landsat satellite data have a substantial historical archive and will continue to be collected into the future, making it the only available thermal imagery for historical analysis and long-term monitoring of geothermal areas in the entirety of YNP. Landsat imagery from Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) sensors was used to examine change trajectories for terrestrial emittance among spatial groupings from 1986 to 2007. Trajectories of locations with known change events were also examined. Relationships between the spatial groupings and several change vectors (distance to geologic faults, distance to large water bodies, and distance to earthquake swarms) were explored. The analysis showed the strongest relationship between absolute difference...

After nearly a century of height suppression, willows (Salix spp.) in the northern range of Yellowstone National Park, USA, are increasing in height growth as a possible consequence of wolf (Canis lupus) restoration, climate change, or other factors. Regardless of the drivers, the recent release of this rare but important habitat type could have significant implications for associated songbirds that are exhibiting declines in the region. Our objective was to evaluate bird response to releasing willows by comparing willow structure and bird community composition across three willow growth conditions: height suppressed, recently released, and previously tall (i.e., tall prior to the height increase of released willows). Released and previously tall willows exhibited high and similar vertical structure, but released willows were significantly lower in horizontal structure. Suppressed willows were significantly shorter and lower in horizontal cover than released or previously tall willows. Bird richness increased along a gradient from lowest in suppressed to highest in previously tall willows, but abundance and diversity were similar between released and previously tall willows, despite lower horizontal cover in the released condition. Common Yellowthroat (Geothlypis trichas) and Lincoln's Sparrow (Melospiza lincolnii ) were found in all three growth conditions; however, Yellow Warbler (Dendroica petechia), Warbling Vireo (Vireo gilvus), Willow Flycatcher (Empidonax traillii ), and Song Sparrow (Melospiza melodii ) were present in released and previously tall willows only. Wilson's Warbler (Wilsonia pusilla) was found in previously tall willows only, appearing to specialize on tall, dense willows. The results of our a priori habitat models indicated that foliage height diversity was the primary driver of bird richness, abundance, and diversity. These results indicate that vertical structure was a more important driver of bird community variables than horizontal structure and that riparian and willow- dependent bird species have responded positively to increased willow growth in the region.

A pilot cropland carbon sequestration program within north central Montana has allowed farmers to receive carbon credit for management adjustments associated with changing from tillage-based agricultural systems to no-till. Carbon credit can also be obtained by adopting conservation reserve, where cropland is planted into perennial vegetation. Summer fallowing is also considered within the crediting process as credit is not given in years that a field is left un-vegetated. The carbon sequestration program has been advocated as a means to mitigate climate change while providing an added source of income for Montana farmers. There is lack of data, however, pertaining to the percentage of lands within this region that have not converted to no-till management, lands under certain crop intensities (e.g. those that are cropped every growing season vs. those that use a fallow-crop-fallow system), or cropland that have converted to perennial vegetation outside of the popular Conservation Reserve Program. Data is also sparse concerning the amount of soil organic carbon that might be sequestered given a conversion to no-till or conservation reserve. This study established regional percentage estimates of cropland under no-till, various degrees of crop intensity, and conservation reserve within north central Montana. Literature-based carbon sequestration estimates were used to generate carbon gain data associated with the conversation to no-till and to conservation reserve. These estimates were then applied to the area-based cropland statistics to estimate potential regional carbon sequestration associated with these management changes.

Conservation tillage management has been advocated for carbon sequestration and soil quality preservation purposes. Past satellite image analyses have had difficulty in differentiating between no-till (NT) and minimal tillage (MT) conservation classes due to similarities in surface residues, and may have been restricted by the availability of cloud-free satellite imagery. This study hypothesized that the inclusion of high temporal data into the classification process would increase conservation tillage accuracy due to the added likelihood of capturing spectral changes in MT fields following a tillage disturbance. Classification accuracies were evaluated for Random Forest models based on 250-m and 500-m MODIS, 30-m Landsat, and 30-m synthetic reflectance values. Synthetic (30-m) data derived from the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) were evaluated because high frequency Landsat image sets are often unavailable within a cropping season due to cloud issues. Classification results from a five-date Landsat model were substantially better than those reported by previous classification tillage studies, with 94% total and ≥ 88% class producer's accuracies. Landsat-derived models based on individual image scenes (May through August) yielded poor MT classifications, but a monthly increase in accuracy illustrated the importance of temporal sampling for capturing regional tillage disturbance signatures. MODIS-based model accuracies (90% total; ≥ 82% class) were lower than in the five-date Landsat model, but were higher than previous image-based and survey-based tillage classification results. Almost all the STARFM prediction-based models had classification accuracies higher than, or comparable to, the MODIS-based results (> 90% total; ≥ 84% class) but the resulting model accuracies were dependent on the MODIS/Landsat base pairs used to generate the STARFM predictions. Also evident within the STARFM prediction-based models was the ability for high frequency data series to compensate for degraded synthetic spectral values when classifying field-based tillage. The decision to use MODIS or STARFM-based data within conservation tillage analysis is likely situation dependent. A MODIS-based approach requires little data processing and could be more efficient for large-area mapping; however a STARFM-based analysis might be more appropriate in mixed-pixel situations that could potentially compromise classification accuracy.

Severe mountain pine beetle (MPB) epidemics can degrade ecosystem services and socioeconomic assets. Mapping outbreak progression provides tools to mitigate damages and analyze MPB attack processes. Current time-series methods for mapping disturbance focus on extent rather than severity. Infestation severity, defined by within-pixel percentage, is more robust for answering a variety of ecologic questions. We develop a time-series regression approach to map infestation severity from 2005 to 2015 in the U.S. Central Rocky Mountains. Covariates include spectral data from all available dates of Landsat imagery, topographic data, and US Forest Service aerial detection survey (ADS) polygons. We collect model reference data by interpreting National Agricultural Imagery Program images. Validation against a randomly selected subset of the data results in no statistical difference between predicted and observed severity. The mean absolute deviation is 7.7% with a root-mean-square error of 9.9%. Average (maximum) severity increased from 9.4% (49.7%) in 2005 to 17.6% (58.8%) in 2015. Our raster maps identify widespread, lower severity infestation absent from the ADS. Our maps can improve mitigation efforts by allowing managers to: address low-severity infestations before they intensify, monitor intensifying infestations within previously identified outbreak extents, and combine infestation severity with other forest metrics.

Yellowstone National Park (YNP) is legally mandated to monitor geothermal features for their future preservation, and remote sensing is a component of the current monitoring plan. Landsat imagery was explored as a tool for mapping terrestrial emittance and geothermal heat flux for this purpose. Several methods were compared to estimate terrestrial emittance and geothermal heat flux (GHF) using images from 2007 (Landsat Thematic Mapper) and 2002 (Landsat Thematic Mapper Plus). Accurate estimations were reasonable when compared to previously established values and known patterns but were likely limited due to inherent properties of Landsat data, the effects of solar radiation, and variation among geothermal areas. Landsat data can be valuable for calculation of GHF in YNP. The method suggested in this paper is not highly parameterized. Landsat data provide the means to calculate GHF for all of YNP and have the potential to enable scientists to identify locations for in-depth study.

An inexpensive and reproducible method for monitoring rangelands over the northern range (NR) of Yellowstone National Park was developed utilizing Landsat imagery. A 1999 map of rangeland vegetation communities was created using boosted classification tree analysis. The 1999 map and a 1985 image were utilized in a change vector analysis resulting in a classified map for 1985. Classification accuracies ranged from 72.30 percent to 83.65 percent for the 1999 map and 72.60 percent to 88.73 percent for the 1985 map, depending upon level of class detail, demonstrating that Landsat-class data can be effectively used for efficient change analysis that maintains accuracy while reducing compound error. Spatial patterns of change were compared to theories from other studies related to change in the NR and were found to be consistent with effects from fire suppression, precipitation, and urban growth but not with trophic cascade from wolves or beaver effects.

A hyperspectral imaging system was used to monitor vegetation during a subsurface controlled release of carbon dioxide (CO2). From August 3 to 10, 2007, 0.3 tons CO2/day were released through a 70 m horizontal pipe located at a nominal depth of 1.8 m below the surface. Hyperspectral images of alfalfa plants were collected during the controlled release and used along with classification tree analysis to study changes in the reflectance spectra as a function of perpendicular distance from the horizontal pipe. Changes in the reflectance spectra near the red edge (650–750 nm) were observed over the course of the controlled release experiment for plants within a perpendicular distance of 1 m of the release pipe. These results indicate monitoring vegetation over a carbon sequestration site has the potential to allow monitoring of the integrity of the CO2 storage.

article i nfo We used an object-oriented approach in conjunction with the Random Forest algorithm to classify agricultural practices, including tillage (till or no-till (NT)), crop intensity, and grassland-based conservation reserve (CR). The object-oriented approach allowed for per-field classifications and the incorporation of contextual elements in addition to spectral features. Random Forest is a classification tree-based advanced classifier that avoids data over-fitting associated with many tree-based models and incorporates an unbiased internal classification accuracy assessment. Landsat satellite imagery was chosen for its continuous coverage, cost effectiveness, and image accessibility. Classification results for 2007 included producer's accuracies of 91% for NT and 31% for tillage when applying Random Forest to image objects generated from a May Landsat image. Low classification accuracies likely were attributed to the misclassification of conservation-based tillage practices as NT. Results showed that the binary separation of tillage from NT management is likely not appropriate due to surface spectral and textural similarities between NT and conservation-type tillage practices. Crop and CR lands resulted in producer's accuracies of 100% and 90%, respectively. Crop and fallow producer's accuracies were 95% and 82% in the 2007 classification, despite post-senesced vegetation; misclassification within the fallow class was attributed to pixel-mixing problems in areas of narrow (b100 m) strip management. A between-date normalized difference vegetation index approach was successfully used to detect areas having "changed" in vegetation status between the 2007 and prior image dates; classified "changed" objects were then merged with "unchanged" objects to produce crop status maps. Field crop intensity was then determined from the multi-year analysis of generated crop status maps.

This paper evaluates the potential for remote mapping of river bathymetry by (1) examining the theoretical basis of a simple, ratio-based technique for retrieving depth information from passive optical image data; (2) performing radiative transfer simulations to quantify the effects of suspended sediment concentration, bottom reflectance, and water surface state; (3) assessing the accuracy of spectrally based depth retrieval under field conditions via ground-based reflectance measurements; and (4) producing bathymetric maps for a pair of gravel-bed rivers from hyperspectral image data. Consideration of the relative magnitudes of various radiance components allowed us to define the range of conditions under which spectrally based depth retrieval is appropriate: the remotely sensed signal must be dominated by bottom-reflected radiance. We developed a simple algorithm, called optimal band ratio analysis (OBRA), for identifying pairs of wavelengths for which this critical assumption is valid and which yield strong, linear relationships between an image-derived quantity X and flow depth d. OBRA of simulated spectra indicated that water column optical properties were accounted for by a shorter-wavelength numerator band sensitive to scattering by suspended sediment while depth information was provided by a longer-wavelength denominator band subject to strong absorption by pure water. Field spectra suggested that bottom reflectance was fairly homogeneous, isolating the effect of depth, and that radiance measured above the water surface was primarily reflected from the bottom, not the water column. OBRA of these data, 28% of which were collected during a period of high turbidity, yielded strong X versus d relations (R2 from 0·792 to 0·976), demonstrating that accurate depth retrieval is feasible under field conditions. Moreover, application of OBRA to hyperspectral image data resulted in spatially coherent, hydraulically reasonable bathymetric maps, though negative depth estimates occurred along channel margins where pixels were mixed. This study indicates that passive optical remote sensing could become a viable tool for measuring river bathymetry. Copyright © 2009 John Wiley & Sons, Ltd.

article i nfo Local, field-scale, VisNIR-DRS soil calibrations generally yield the most accurate predictions but require a substantial number of local calibration samples at every application site. Global to regional calibrations are more economically efficient, but don't provide sufficient accuracy for many applications. In this study, we quantified the value of augmenting a large global spectral library with relatively few local calibration samples for VisNIR-DRS predictions of soil clay content (clay), organic carbon content (SOC), and inorganic carbon content (IC). VisNIR models were constructed with boosted regression trees employing global, local+global, and local spectral data, using local samples from two low-relief, sedimentary bedrock controlled, semiarid grassland sites, and one granitic, montane, subalpine forest site, in Montana, USA. The local+global calibration yielded the most accurate SOC predictions for all three sites (Standard Error of Prediction (SEP)= 3.8, 6.7, and 26.2 g kg −1 ). This was similarly true for clay (SEP=95.3 and 102.5 g kg −1 ) and IC (SEP=5.5 and 6.0 g kg −1 ) predictions at the two semiarid grassland sites. A purely local calibration produced the best validation results for soil clay content at the subalpine forest site (SEP=49.2 g kg −1 ), which also had the largest number of local calibration samples (N=210). Using only samples from calcareous soils in the global spectral library combined with local samples produced the best SOC and IC results at the more arid of the two semiarid sites. Global samples alone never achieved more accurate predictions than the best local+global calibrations. For the temperate soils used in this study, the augmentation of a large global spectral library with relatively few local samples generally improved the prediction of soil clay, SOC, and IC relative to global or local samples alone.

Moderate resolution satellite imagery traditionally has been thought to be inadequate for mapping vegetation at the species level. This has made comprehensive mapping of regional distributions of sensitive species, such as whitebark pine, either impractical or extremely time consuming. We sought to determine whether using a combination of moderate resolution satellite imagery (Landsat Enhanced Thematic Mapper Plus), extensive stand data collected by land management agencies for other purposes, and modern statistical classification techniques (boosted classification trees) could result in successful mapping of whitebark pine. Overall classification accuracies exceeded 90%, with similar individual class accuracies. Accuracies on a localized basis varied based on elevation. Accuracies also varied among administrative units, although we were not able to determine whether these differences related to inherent spatial variations or differences in the quality of available reference data.

Accurate, efficient, and repeatable mapping of changes in wetlands and riparian areas (referred to collectively as wetlands) is critical for monitoring human, climatic, and other effects on these important systems. We used Landsat-based satellite imagery from 1988 and 2001 to map changes in wetland ecosystems in the Gallatin Valley of southwest Montana. Stochastic gradient boosting (SGB) was used to classify the 2001 image, and change vector analysis (CVA) was used to identify locations where wetland areas might have changed between 1988 and 2001. These potentially changed locations again were classified for the 1988 Landsat image using SGB. Areas of change constituted 3.4% of the study area, thus only this small percentage of the image was reclassified for the 1988 image. Overall change detection accuracy was 76%, although changes along the periphery of wetland boundaries and in areas of smaller upland inclusions were not distinguished as well as other changes. Overall accuracies of the SGB wetland classification maps were 81% for 1988 and 86% for 2001. CVA significantly reduced the number of pixels involved in the historical image classification compared to conducting independent classifications, thus reducing the potential for compounding classification errors in unchanged areas.

Ecological site descriptions (ESDs) based on soil maps, Landsat 7 ETM+ band values, and vegetation index data from 12 scenes were used as predictive variables in linear regression estimates of total biomass using field data from five Montana ranches. Bandwise regression explained the most variability (53%) when ESDs were not included, followed by tasseled cap components (51%), the soil adjusted vegetation index (44%), and the normalized difference vegetation index (41%). ESDs improved the amount of variability explained to 66% for bandwise regression and 65% using tasseled cap components.

The Kyoto Protocol entering into force on 16 February 2005 continues to spur interest in development of carbon trading mechanisms internationally and domestically. Critical to the development of a carbon trading effort is verification that carbon has been sequestered, and field level measurement of C change is likely cost prohibitive. Estimating C change based on agricultural management practices related to carbon sequestration seems more realistic, and analysis of satellite imagery could be used to monitor and verify these practices over large areas. We examined using Landsat imagery to verify crop rotations and quantify crop residue biomass in north central Montana. Field data were collected using a survey of farms. Standard classification tree analysis (CTA) and boosted classification and regression tree analysis (BCTA) were used to classify crop types. Linear regression (LM), regression tree analysis (RTA), and stochastic gradient boosting (SGB) were used to estimate crop residue. Six crop types were classified with 97% accuracy (BCTA) with class accuracies of 88–99%. Paired t -tests were used to compare the difference between known and predicted mean crop residue biomass. The difference between known and predicted mean residues using SGB was not different than 0 ( p -value = 0.99); however root mean square error (RMSE) was large (1981 kg ha −1 ), implying that SGB accurately predicted regional crop residue biomass but not local predictions (i.e., field or farm level). The results of this study, and previous research classifying tillage practices and estimating soil disturbance, supports using satellite imagery as an effective tool for monitoring and verifying agricultural management practices related to carbon sequestration over large areas.