Browsing by Subject "Remote Sensing"

Now showing 1 - 9 of 9
  • Thumbnail Image
    Item
    2015 Twin Cities Metropolitan Area Urban Tree Canopy Assessment
    (2017-01-03) Knight, Joe F; Rampi, Lian P; Host, Trevor K; jknight@umn.edu; Knight, Joseph, F
    A high-resolution (1-meter) tree canopy assessment was completed for the Twin Cities Metropolitan Area. Mapping of existing and potential tree canopy is critical for urban tree management at the landscape level. This classification was created from combined 2015 aerial imagery, LIDAR data, and ancillary thematic layers. These data sets were integrated using an Object-Based Image Analysis (OBIA) approach through multi-resolution image segmentation and an iterative set of classification commands in the form of customized rulesets. eCognition® Developer was used to develop the rulesets and produce raster classification products for TCMA. The results were evaluated using randomly placed and independent verified assessment points. The classification product was analyzed at regional scales to compare distributions of tree canopy spatially and at different resolutions. The combination of spectral data and LiDAR through an OBIA method helped to improve the overall accuracy results providing more aesthetically pleasing maps of tree canopy with highly accurate results.
  • Thumbnail Image
    Item
    Advancing Remote Sensing For Soybean Aphid (Hemiptera: Aphididae) Management In Soybean
    (2019-07) Marston, Zachary
    Soybean aphid, Aphis glycines Matsumura (Hemiptera: Aphididae), is the most economically important insect pest of soybean, Glycine max (L.) Merrill (Fabales: Fabaceae), in the north-central United States. Current management recommendations for soybean aphid include frequent scouting of soybean fields and application of foliar insecticides when soybean aphid populations exceed an economic threshold of 250 aphids per plant. The scouting process for soybean aphid is time consuming, expensive, and also fails to thoroughly assess populations across the entire field. Because of these drawbacks, 84% of soybean farmers want to reduce scouting efforts. In 2015, it was determined that soybean aphid-induced stress had a significant effect on red-edge and near-infrared (NIR) reflectance of soybean canopies, offering the potential to use remote sensing for soybean aphid scouting. Utilizing remote sensing for soybean aphid scouting may decrease human effort, increase spatial coverage, and ultimately increase the adoption of recommended management practices. However, it was unknown whether soybean aphid-induced stress could be detected from aerial platforms, whether these reflectance data of aphid-induced stress could be classified into treatment groups, and how confounding factors might affect classification results. My first chapter determined that soybean aphid-induced stress could be detected from an unmanned aerial vehicle (UAV) equipped with a multispectral sensor. Findings indicated that NIR reflectance decreased as aphid populations increased in both caged and open-field experiments. Chapter 2 evaluated ground-based hyperspectral samples and determined that soybean reflectance samples that were above the economic threshold of 250 aphids per plant could be classified with over 86% accuracy using linear support vector machine classification. Chapter 3 further evaluated ground-based hyperspectral samples in the presence of the confounding disease, soybean sudden death syndrome (SDS) caused by the fungal pathogen Fusarium virguliforme O’Donnell and T. Aoki (Hypocraeles: Nectriaceae). Findings indicated that when using linear support vector machines, it was difficult to differentiate between healthy and diseased samples; however, including the diseased group in the classification model decreased false positives for soybean aphid-induced stress. Overall, these findings advance the use of remote sensing for soybean aphid management and provide the first documentation of spectral classification of soybean aphid into threshold-based groups.
  • Thumbnail Image
    Item
    Computational Techniques to Identify Rare Events in Spatio-temporal Data
    (2018-05) Mithal, Varun
    Recent attention on the potential impacts of land cover changes to the environment as well as long-term climate change has increased the focus on automated tools for global-scale land surface monitoring. Advancements in remote sensing and data collection technologies have produced large earth science data sets that can now be used to build such tools. However, new data mining methods are needed to address the unique characteristics of earth science data and problems. In this dissertation, we explore two of these interesting problems, which are (1) build predictive models to identify rare classes when high quality annotated training samples are not available, and (2) classification enhancement of existing imperfect classification maps using physics-guided constraints. We study the problem of identifying land cover changes such as forest fires as a supervised binary classification task with the following characteristics: (i) instead of true labels only imperfect labels are available for training samples. These imperfect labels can be quite poor approximation of the true labels and thus may have little utility in practice. (ii) the imperfect labels are available for all instances (not just the training samples). (iii) the target class is a very small fraction of the total number of samples (traditionally referred to as the rare class problem). In our approach, we focus on leveraging imperfect labels and show how they, in conjunction with attributes associated with instances, open up exciting opportunities for performing rare class prediction. We applied this approach to identify burned areas using data from earth observing satellites, and have produced a database, which is more reliable and comprehensive (three times more burned area in tropical forests) compared to the state-of-art NASA product. We explore approaches to reduce errors in remote sensing based classification products, which are common due to poor data quality (eg., instrument failure, atmospheric interference) as well as limitations of the classification models. We present classification enhancement approaches, which aim to improve the input (imperfect) classification by using some implicit physics-based constraints related to the phenomena under consideration. Specifically, our approach can be applied in domains where (i) physical properties can be used to correct the imperfections in the initial classification products, and (ii) if clean labels are available, they can be used to construct the physical properties.
  • Thumbnail Image
    Item
    Machine learning algorithms for spatio-temporal data mining
    (2008-12) Vatsavai, Ranga Raju
    Remote sensing, which provides inexpensive, synoptic-scale data with multi-temporal coverage, has proven to be very useful in land cover mapping, environmental monitoring, forest and crop inventory, urban studies, natural and man made object recognition, etc. Thematic information extracted from remote sensing imagery is also useful in variety of spatiotemporal applications. However, increasing spatial, spectral, and temporal resolutions invalidate several assumptions made by the traditional classification methods. In this thesis we addressed four specific problems, namely, small training samples, multisource data, aggregate classes, and spatial autocorrelation. We developed a novel semi-supervised learning algorithm to address the small training sample problem. A common assumption made in previous works is that the labeled and unlabeled training samples are drawn from the same mixture model. However, in practice we observed that the number of mixture components for labeled and unlabeled training samples differ significantly. Our adaptive semi-supervised algorithm over comes this important limitation by eliminating unlabeled samples from additional components through a matching process. Multisource data classification is addressed through a combination of knowledge-based and semi-supervised approaches. We solved the aggregate class classification problem by relaxing the unimodal assumption. We developed a novel semi-supervised algorithm to address the spatial autocorrelation problem. Experimental evaluation on remote sensing imagery showed the efficacy of our novel methods over conventional approaches. Together, our research delivered significant improvements in thematic information extraction from remote sensing imagery.
  • Thumbnail Image
    Item
    Machine Learning for Advancing Spaceborne Passive Microwave Remote Sensing of Snowfall
    (2022-09) Vahedizade, Sajad
    Falling snow is one of the key elements of the water and energy cycle that occurs in response to a complex cascade of macro and microphysical processes. While an accurate observation of spatiotemporal variability of snowfall is lacking due to the sparse network of ground-based gauges and their intrinsic challenges, remote sensing from spaceborne satellites has provided a global picture of snowfall through near-global observations. Bayesian passive microwave (PMW) retrievals of snowfall have been developed to detect precipitation phase and retrieve its rate using coincident data from the active and passive sensors onboard the CloudSat and the Global Precipitation Measurement (GPM) satellites. These Bayesian techniques often rely on mathematical matching of the observed vectors of brightness temperature (TB) with an a priori database of precipitation profiles and their corresponding TBs. In the present dissertation, we analyzed the effects of surface and atmospheric state variables on the PMW retrievals through Silhouette Coefficient (SC) analysis. The Neyman-Pearson (NP) hypothesis testing is employed to improve these retrievals by conditioning them to the associated physical variables that affect the PMW signatures including the cloud total liquid (LWP) and ice water path (IWP). The presented approach determines thresholds for IWP and LWP that enable identification of non-snowing and snowing clouds, which can mislead the retrieval algorithms to falsely detect or miss the snowfall events. Inspired by advances in deep learning approaches, a dense and deep neural network architecture is proposed. The presented framework first detects the precipitation occurrence and its phase, and then estimates its intensity using key physical variables including those capturing cloud microphysical properties. The results suggest the proposed framework could effectively reduce the uncertainties in the retrievals and improve their accuracy compared to the existing reanalysis data and official GPM products.
  • Thumbnail Image
    Item
    Nursery Production Method Performance Evaluation Assessed With The Normalized Difference Vegetation Index Derived From An Unmanned Aircraft System Mounted Single-Imager Sensor
    (2020-03) Bahe, Michael
    Trees provide many benefits to urban areas including enhanced human health, pollution mitigation, and reductions in residential energy consumption. The goal of urban forest managers is to develop mature trees with large crowns to maximize these benefits. Urban trees have the highest mortality rate during the initial years post planting, known as the establishment period. In an era of planting trees to reach quotas, the looming fact is many perish during establishment limiting goal achievement. Nursery production methods (NPM) are a controllable factor in practice that may have an impact on establishment success. In this study, urban trees planted in situ from four common NPM’s (balled and burlapped, smooth plastic containers, spring planted bareroot, and gravelbed bareroot) were monitored for three years post planting using the normalized difference vegetation index (NDVI). This data was derived from high-resolution imagery collected with an unmanned aircraft system (UAS). First, the single-imager multispectral sensor selected for this project was evaluated for effectiveness in determining tree health. This was done in a controlled growth chamber environment. Results showed the single-imager sensor derived NDVI values were effective indicators of tree stress within species groups. Second, a novel technique to isolate tree crowns for spectral data analysis with UAS derived imagery was utilized to compare the health of newly planted trees in situ from the four NPM’s. Analysis of the effect NPM’s had on tree health during the establishment period showed minimal differences between the study groups thus providing evidence that each is a viable option for practitioners in urban areas.
  • Thumbnail Image
    Item
    Remote sensing for regional assessment and analysis of Minnesota lake and river water quality
    (2012-05) Olmanson, Leif Gordon
    Beginning soon after the launch of the first Landsat satellite, researchers began investigating the use of Landsat imagery to monitor the water quality of our lakes and coastlines. The earliest use of Landsat imagery was for simple qualitative observations which included locating and mapping pollution and pollution plumes. Shortly thereafter, field measurements of water quality were correlated with Landsat data and later these correlations were used for quantitative assessment of water quality (e.g., turbidity, chlorophyll and water clarity). This dissertation expands on this earlier work and describes results of research to develop and use remote sensing tools for regional water quality assessment to improve the understanding and management of Minnesota's lakes and rivers. It includes four major components. First, a 20-year, 1985-2005, comprehensive water clarity database for more than 10,500 lakes at approximately five-year intervals for the time period 1985-2005, which includes almost 100,000 individual estimates of lake water clarity, was compiled and evaluated. Second, the results of a statistical analysis of the Landsat database for geospatial and temporal trends of water clarity over the 20-year period, as well as trends related to land cover/use and lake morphometry, are reported. Third, the advantages of improved spectral and temporal resolution and disadvantages of the lower spatial resolution of the global MODIS and MERIS systems are evaluated for regional-scale measurements of lake water clarity and chlorophyll of large lakes in Minnesota and compared with Landsat. Finally, aerial hyperspectral spectrometers were used to collect imagery with high spatial and spectral resolution for use in identifying, measuring and mapping optically related water quality characteristics of major rivers in Minnesota for three time periods that represent different water quality and flow regimes.
  • Thumbnail Image
    Item
    Spatial Assessment of Boreal Forest Carbon
    (2015-06) Kristensen, Terje
    The ability to accurately map and monitor forest carbon (C) has gained global attention as countries seek to comply with international agreements to mitigate climate change. However, attaining precise estimates of forest C storage is challenging due to the inherent heterogeneity occurring across different scales. To develop cost-effective sampling protocols, there is a need for more unbiased estimates of the current C stock, its distribution among forest compartments and its variability across different scales. As a contribution to this work, this dissertation used high-resolution field measurements of C collected from different forest compartments across a boreal forest stand in South East Norway. In the first paper, we combined the use of airborne scanning light detection and ranging (lidar) systems with fine-scale spatial C data relating to vegetation and the soil surface to describe and contrast the size and spatial distribution of C pools across the forest. We found that predictor variables from lidar derived metrics delivered precise models of above and belowground tree C, which comprised the largest of the measured C pool in our study. We also found evidence that lidar canopy data correlated well with the variation in field layer C stock. By using topographical models from lidar ground returns we were able to establish a strong correlation between lidar data and the organic layer C stock at a stand level. In the search for an effective tool to measure and monitor forest C pools, we found the capabilities of lidar to map forest C encouraging. In the second paper, we used a geostatistical approach to analyze the fine-scale heterogeneity of the soil organic layer (forest floor) C storage. Our results showed that the C stocks were highly variable within each plot, with spatial autocorrelation distances < 3 m. Further, we established that a minimum of 20 to 25 inventory samples is needed to determine the organic layer C stock with a precision of �0.5 kg C m-2 in inventory plots of ~2000 m2. In the third paper, we investigated how the short-range spatial variability of organic layer C affects sampling strategies aiming to monitor and detect changes in the C stock. We found that sample repeatability rapidly declines with sample separation distance, and the a priori sample sizes needed to detect a change a fixed change in the organic layer C stock vary by a factor of ~4 over 15 to 125 cm separation distance. Unless care is taken by the surveyor to ensure spatial sampling precision, substantially larger samples sizes, or longer time intervals between baseline sampling and revisit are required to detect a change. In the final paper, we utilized the nested sampling protocol to investigate the spatial variability of organic layer C across different scales and incorporated inventory expenses in the development of a cost-optimal sampling approach. Because precise estimates are costly to obtain, it is of great interest for surveyors to develop cost-efficient sampling protocols aimed at maximizing the spatial coverage, while minimizing the estimate variance. We found that the majority of the estimate variance is confined within small subplots (100 m2) of the forest (25 km2), emphasizing the importance of considering the short-range variability when conducting a large-scale inventory. Further, this chapter demonstrated how optimal allocation of sampling units (plot, subplot and sample) is not only a function of the variance component within that dimension, but also changes with the sampling unit costs and the acceptable margin of error. We found that the costs of conducting an organic layer C inventory could be reduced by more than 60% by increasing the inventory uncertainty from �0.25 Mg C ha-1 to �0.5 Mg C ha-1. Finally, we established that sampling costs can be reduced with as much 80% by conducting a double sampling procedure that utilizes the correlation between organic layer C stock (r = 0.79 to 0.85) and measurements of layer thickness.
  • Thumbnail Image
    Item
    The Vulnerability of Northern High-Latitude Ecosystems to Climate and Disturbance-Induced Change
    (2018-06) Pastick, Neal
    Arctic and boreal regions have experienced unprecedented changes in recent decades as the result of climate change. Increasing air temperatures have led to widespread warming and degradation of permafrost, significant shifts in vegetation composition and productivity, and increases in disturbance frequency and extent that can have profound impacts on ecosystems and human populations across the globe. Despite a legacy of studies describing the heightened sensitivity of arctic and boreal ecosystems to change, there has not been a comprehensive assessment of historical and projected trends in landscape properties, disturbances, and drivers of change throughout all of Alaska. Such an assessment is immensely challenging because of spatially-heterogeneous dynamics and interactions among numerous factors that influence ecosystems throughout the State. Consequently, additional research is needed to better characterize permafrost-affected landscapes and their potential response to further perturbations. This dissertation presents important improvements in the mechanistic understanding and characterization capabilities of changing permafrost landscapes by combining field measurements, time series analyses, climate reanalysis data, and remote sensing into an integrated modeling framework. The primary goal is to improve understanding of how and why globally significant permafrost landscapes are changing by means of: (1) Characterizing climate, permafrost, disturbance, and vegetation dynamics that exert strong controls on energy, water, and biogeochemical cycling; (2) Quantifying underlying drivers of change related to contemporary trends in land and water surfaces observed by remote sensing; and; (3) Providing novel approaches and baseline information to fill critical observational gaps identified by the remote sensing community and permafrost and ecosystem scientists. This research supports the science priorities of federal agencies (e.g. United States Geological Survey, National Aeronautics and Space Administration) and techniques and results are highly relevant to climatic, hydrologic, ecologic, topographic, and cryospheric studies. This research provides critically needed information on the temporal and regional distribution of landscape properties and conditions, which is instrumental in determining the vulnerability and resilience of northern high-latitudes regions to climate and disturbance-induced change, and benefits both the research community and the policy community in the management of Arctic and boreal landscapes.