AI Use Cases in Department of Interior

1. Land Use Plan Document and Data Mining and Analysis R&D

This project explores the potential of identifying patterns, rule alignments or conflicts, discovery, and mapping of geo-history and rules from unstructured planning documents. The outputs identify conflicts in resource management planning rules with proposed action locations, highlighting areas that require exclusion, restrictions, or stipulations as defined in the planning documents. This analysis helps in better understanding and managing land use plans, ensuring that proposed actions comply with existing rules and regulations, and supporting more informed decision-making in resource management.

2. Seasonal/Temporary Wetland/Floodplain Delineation using Remote Sensing and Deep Learning

This project investigates whether recent advancements in machine learning, particularly convolutional neural network (CNN) architecture in deep learning, can improve the delineation (mapping) of seasonal/temporary wetlands and floodplains using high temporal and spatial resolution remote sensing data. If successful, these new mapping techniques could inform the management of protected species and provide critical information to decision-makers during scenario analysis for operations and planning. The project aims to enhance the accuracy and reliability of wetland and floodplain mapping, supporting better environmental management and conservation efforts.

3. Data Driven Sub-Seasonal Forecasting of Temperature and Precipitation

This project involves running two year-long prize competitions where participants developed and deployed data-driven methods for sub-seasonal (2-6 weeks into the future) prediction of temperature and precipitation across the western US. The participants’ methods outperformed benchmark forecasts from NOAA. Reclamation is now collaborating with the Scripps Institute of Oceanography to further refine, evaluate, and pilot implement the most promising methods from these competitions. Improving sub-seasonal forecasts has significant potential to enhance water management outcomes, supporting better planning and resource allocation in the face of climate variability.

4. PyForecast

PyForecast is a water supply forecasting software developed by Reclamation that utilizes statistical and machine learning methods. This software employs various data-driven techniques to provide accurate water supply forecasts, aiding in better water resource management and planning.

5. Improved Processing and Analysis of Test and Operating Data from Rotating Machines

This project aims to improve the analysis of DC ramp test data from rotating machines. Traditionally, analyzing these tests requires engineering expertise to identify characteristic curves from voltage vs. current plots. By employing machine learning and AI tools like linear regression, the project seeks to automate this analysis using computer software. This approach is expected to provide faster and more reliable analysis of DC ramp tests conducted in the field, enhancing the efficiency and accuracy of machine diagnostics.

6. Sustained Casing Pressure Identification

This project addresses the identification of sustained casing pressure (SCP) problems in well platforms, which can lead to major safety issues. SCP is typically caused by gas migration from high-pressure subsurface formations through leaking cement sheaths, but it can also result from defects in tube connections, downhole accessories, or seals. To mitigate accidents, quickly identifying wells with SCP is crucial. BSEE has partnered with NASA’s Advanced Supercomputing Division to research the use of various AI techniques for this purpose, aiming to enhance safety and efficiency in well platform operations.

7. Well Activity Report Classification

This project involves researching the use of self-supervised and supervised deep neural networks to classify significant well events using data from Well Activity Reports. The goal is to develop a robust classification system that can accurately identify and categorize important events, improving the monitoring and management of well activities.

8. Well Risk

Building on the research into sustained casing pressure, NASA’s Advanced Supercomputer Division is developing machine learning models to identify precursors of risk factors for wells. By pinpointing these risk factors, the models aim to inform BSEE engineers of potential problems during different stages of well development, enhancing safety and operational efficiency.

9. Autonomous Drone Inspections

This project explores the development of autonomous drone systems to detect methane and inspect unsafe platforms on the outer continental shelf. Autonomous drones can perform inspections while ensuring the safety of inspectors and reducing the need for extensive training. This technology aims to enhance inspection capabilities and safety in challenging environments.

10. Level 1 Report Corrosion Level Classification

Level 1 surveys from BSEE assess the condition of well platforms, including images of components to estimate coating and structural conditions. These reports are crucial for identifying safety concerns and determining if additional audits are needed. Currently, the manual review process is time-consuming. To address this, BSEE has partnered with NASA’s Advanced Supercomputing Division to research AI techniques for automating the identification of excessive corrosion, aiming to significantly reduce processing time and enhance safety assessments.

11. Data Mining, Machine Learning and the IHS Markit Databases

This project supports the DOI Secretarial Priority Project on Smart Energy Development by using machine learning to identify potential conflicts between energy development and other priorities. It aims to extract spatial patterns for future development and lay the groundwork for new skills, analyses, and products for the ERP and Mission Area. The project also focuses on building internal knowledge about the capabilities of machine learning for the ERP.

12. Aluminum Criteria Development in Massachusetts

The USGS, in collaboration with MassDEP, is collecting water-quality data at freshwater sites in Massachusetts. This data will be used to demonstrate a process for calculating aluminum criteria based on water chemistry parameters like pH, DOC, and hardness, using a multiple linear regression model developed by the EPA in 2017. This project aims to provide a scientific basis for setting aluminum criteria in water quality management.

13. Multi-scale modeling for ecosystem service economics

This project aims to expand the ARIES modeling framework using AI and decision rules to create a system that selects models and data based on contextual factors like climate and socioeconomics. By using national and global datasets, the system will map ecosystem services (ES) with greater accuracy. The goal is to implement this intelligent modeling system across the U.S., providing a consistent, AI-supported framework for ES assessment and valuation. This includes integrating national economic accounts data with ES data for more timely and comprehensive insights at both national and subnational levels.

14. WOS.OS.NHM National Temperature Observations

This project aims to reduce the burden on Science Centers by improving the collection, storage, analysis, and processing of quality assurance data for water temperature measurements. The objectives include modifying software for processing and storing discrete water temperature data, implementing workflows and QA checks to support new temperature policies, and creating a pilot program to help Science Centers perform 5-point temperature checks. These improvements are expected to increase the deployment of sensors in the water temperature network, enhancing data quality and coverage.

15. WRA.HIHR.WAIEE Building capacity for assessment and prediction of post-wildfire water availability

This project focuses on building capacity for assessing and predicting post-wildfire water availability in the western U.S. Objectives include collecting harmonized datasets from fire-affected basins to develop, calibrate, and validate water-quality models; analyzing these datasets to assess regional differences in water quality impairment drivers; and developing a decision tree and standardized plan for post-fire water-quality monitoring. The project also aims to create a rapid response plan for immediate post-fire data collection, establish the state of science for post-fire water quality impairment, and characterize critical drivers. Additional goals include building a catalog of methods for remotely sensed water quality measurement, developing a catalog of critical data needs for geospatial prediction of wildfire impacts, and constructing a blueprint for incorporating missing water-quality processes into models. The project will also prepare a plan for incorporating wildfire effects on water availability into rapid prediction and participate in developing a framework for cross-Mission Area integration of predictive approaches for post-fire hazards.

16. WRA.NWC.WU Gap analysis for water use

The Water Use Program is conducting a detailed gap analysis of national water-use data to better understand uncertainties in water-use estimates and inform future data collection and modeling efforts. The primary objectives are to identify dominant water-use categories in different U.S. regions, pinpoint data gaps that, if filled, would improve model performance, and identify methods for data estimation to fill these gaps. Additional objectives include enhancing understanding of data quality to inform model prediction uncertainties, collaborating with model developers to understand model sensitivity to input data, and improving data quality related to water extraction, delivery, and consumption for key water-use categories. National models for public supply, irrigation, and thermoelectric use are currently under development.

17. WRA.NWC.IWAA National Extent Hydrogeologic Framework for NWC

This project aims to provide nationally consistent predictions of groundwater quality, focusing on salinity and nutrient levels relevant for human and ecological uses, and their influence on surface water. The objectives are organized into three tasks: 1) Groundwater-Quality Prediction for salinity, providing accurate predictions of groundwater salinity to document availability for human and ecological uses; 2) Groundwater-Quality Prediction for nutrients, offering reliable predictions of nutrient concentrations in groundwater; and 3) Incorporating Groundwater-Quality Predictions into comprehensive water availability assessments. This involves developing strategies to couple groundwater quality predictions with flow and flux simulations from process-based models to quantify available groundwater of specified quality and determine its impact on surface water quantity and quality.

18. WRA.NWC.IWAA National-Extent Groundwater Quality Prediction for the National Water Census and Regional Integrated Water Availability Assessments

This project aims to provide consistent national predictions of groundwater quality, focusing on salinity and nutrients, which are important for both human and ecological uses. It also seeks to understand how groundwater quality affects surface water. The project will develop strategies to integrate these predictions into broader water availability assessments, including the National Water Census and regional Integrated Water Availability Assessments. This integration will help in better managing water resources by providing a comprehensive understanding of water quality and availability.

19. WRA.HIHR.WQP Process-guided Deep Learning for Predicting Dissolved Oxygen on Stream Networks

This project aims to develop a model that predicts daily minimum, mean, and maximum levels of dissolved oxygen (DO) in stream segments of the Lower Delaware River Basin. The model will use nationally available datasets to make these predictions. Accurate DO predictions are crucial for maintaining healthy aquatic ecosystems and ensuring water quality for human use.

20. WRA.NWC.EF Economic Valuation of Ecosystem Services in the Delaware River Basin

This project focuses on the economic valuation of ecosystem services in the Delaware River Basin. It aims to create a comprehensive data and model inventory, develop a database for fish data, and use AI/ML models to predict fish abundances and sizes under different future climates and reservoir operations. The project will also develop economic valuation models for the fishery resource and evaluate their validity. By linking these models, the project will enable the assessment of tradeoffs between water use and fisheries resources. A prototype web application will be developed for internal USGS use to facilitate understanding and assessment of these models.

21. WRA.NWC.IWAA Model Application for the National IWAAs and NWC

This project supports the National Water Availability Assessment reports and the National Water Census by developing model applications. The objectives include providing initial model applications, long-term projections, routine updates of current conditions, and short-term forecasts. These objectives cover various hydrologic sub-disciplines such as water budgets, water use, water quality, aquatic ecosystems, and drought. The project will involve strategic planning and implementation of available model applications, with an integrated approach to accommodate multiple sub-disciplines and expertise.

22. WRA.WPID.IWP.PUMP Turbidity Forecasting

This project focuses on improving national hydrological forecast models to provide water quality forecasts, specifically turbidity, which is crucial for water resource managers. Accurate turbidity forecasts help in managing water quality and ensuring safe water for various uses.

23. WRA.WPID.IWP.PUMP ExaSheds stream temperature projections with process-guided deep learning

This three-year project aims to enhance process-guided deep learning (PGDL) stream temperature models by incorporating new forms of process guidance and merging techniques from previous USGS and DOE projects. The focus will be on ensuring the models are robust enough to make reliable projections under future climate conditions. This will improve the accuracy of stream temperature forecasts in the Delaware River Basin, aiding in long-term water resource management.

24. Vegetation and Water Dynamics

This project involves tracking vegetation phenology to monitor drought and capture unique signatures of irrigated agriculture and invasive species. It uses tools for drought mapping and monitoring, and a livestock forage assessment tool to quantify drought effects on forage availability. The project also focuses on understanding spatiotemporal surface water dynamics to inform permafrost degradation and methane emission hotspots. By improving the accuracy of phenology tracking through remote sensing, the project aims to provide valuable geospatial data for water management and ecological comparisons between irrigated and non-irrigated systems.

25. DOMESTIC WELL VULNERABILITY SES INDICATORS NEW HAMPSHIRE

This project aims to investigate the vulnerability of private wells in New Hampshire by examining statistical associations between well data (geology, land use, construction, hydraulics, and chemistry) and socioeconomic status (SES) data at both homeowner and statewide levels. It seeks to identify indicators of vulnerability to well water availability and quality, particularly concerning arsenic and uranium contamination. The findings will be disseminated to scientific and general audiences, as well as targeted community groups, to inform and improve water management practices.

26. Two-Dimensional Detailed Hydraulic Analysis

This project involves conducting detailed hydrologic analysis and developing a two-dimensional hydraulic model to aid in decision-making for floodplain management and the protection of life and property. Objectives include topographic surveys to verify or augment existing data, hydrologic analysis of Joachim Creek to produce discharge-frequency values for various flood flows, and developing a calibrated two-dimensional hydraulic model for these flows. The project will also produce flood profiles and model-derived flood maps for Joachim Creek, illustrating inundation extents, water-surface elevation, depth, and velocity. These maps will help in understanding flood risks and planning non-structural flood mitigation measures.

27. AI/ML for aquatic science

This project focuses on developing AI algorithms for individual fish recognition using computer vision and deep learning. Objectives include developing baseline AI models, improving recognition performance, evaluating AI detection of diseased fish (blotchy bass syndrome), and assessing deep learning models for individual recognition and respiration rate using video data from lab and natural settings.

28. TMDL and Data Mining Investigations