“Overland Assessment of Physical Environments for Mobility in Complex Ecological Landscapes”

closed

A.    Short Description of Funding Opportunity ERDC seeks applications for: Overland landscape data collection and accuracy for assessments of synthetically created datasets. Artificial intelligence (AI), machine learning (ML), and Deep Learning (DL) can help rapidly create geospatial products and aid in understanding and predict the physical and biotic connectivity and formation of landscapes. These techniques are being applied to various components of data production that are influencing the planning and execution of Government programs in addition to publicly created and provided data sets. Techniques and understanding the uses and limitations of their use is paramount in planning multi-domain operations and systems. We lack the ability to traditionally scientifically ground truth these algorithms as this may not be physically or technically feasible, especially in contested environments, but intense data collection and selection to train these models will help increase the accuracy of the resulting products. The work we are seeking shall involve creating detailed vegetation, landform, and structure scientific models. Awardee will conduct and perform research assessments of physical and biotic environments support of mobility in assured position, navigation, timing (APNT) activities and tests in forested and mixed topographic relief areas. Awardee will compare and contrast different datasets provided and produce accuracy assessments of these datasets to inform widely used land cover products and derivatives is essential. Results should also inform the wider scientific literature in this area. Resulting products will help inform and shape the USGS National Land Cover Database and the Virginia Land Cover Database. B. Program Description/Objective: Awardee will compare and contrast provided synthetic datasets with field data to produce a comprehensive accuracy assessment for landcover derived data. Awardee will complete a detailed field collection data in several land cover areas of the upper coastal plain/lower piedmont area of Fort AP Hill, Virginia. Some examples of data needed are surface soils, vegetation, landforms, cover percentages, stem estimates and counts, canopy size, barriers that could affect mobility, edge habitats, and anthropogenic features. The work would involve a minimum mapping unit of approximately 0.5m accuracy or better to help drive accuracy assessments during AI/ML/DL comparisons. Traditional scientific accuracy assessments between synthetic data (AI/ML products), traditional foundation datasets (land cover) and ground truth field verified data will be completed. Successful applicant will provide foundational geospatial final products and accuracy assessment results of synthetic data assessments. Successful applicant will provide access to ongoing spatial data during collection including current and detailed aerial imagery, ground data collection using field survey grade GPS systems (web based) to follow progress. USACE-ERDC-GRL will assist recipient in formulating a proposed data dictionary for collection and successful awardee will also create documentation of processes and workflows that could include videos, manuals, and other media documenting collection processes. Data should be accessible and in a standard geospatial format for ingestion into mapping and modeling programs. We would like to capture data over approximately 15 representative areas of landforms and locations within the area of interest of Ft. AP Hill, Virginia. The timing of data collection will involve leaf-on and off scenarios to better understand the confusion matrix in accuracy. Traditional accuracy assessments of these products will be performed to provide the community with an understanding of the use and limitations.

High