Geospatial as an Accelerator of Impact: Already Converging!
1. Smart Oceans 2020 | October 5th | Virtual Plenary
Geospatial as an Accelerator of Impact
Already Converging!
Dawn J. Wright, Ph.D.
Chief Scientist
Environmental Systems Research Institute (aka Esri)
@deepseadawn
2. Sea Surface Temperature
NOAA
Chlorophyll-a Concentration
NOAA
Global Species Range Rarity
E.O. Wilson Biodiversity Foundation
Ocean Warming Ocean Health Biodiversity
A Global Geospatial Framework is Emerging
3. Geospatial Provides a Proven Framework
Agricultural Science
Hydrology
Ecology
Geology/Geophysics
Conservation Biology
Forestry
Ocean ScienceGeographic Information Science
Spatial Data Science
Computer Science
Cartographic Science
Remote Sensing
Sustainability Science / Geodesign
/ Social Science
Climate Science
4. C-Accel Track A: KnowWhere Graph
Enriching and Linking Cross-Domain Knowledge Graphs using Spatially-Explicit AI
NSF Award 2033521, Lead PI K. Janowicz, UCSB
5. C-Accel Track A: KnowWhere Graph
Enriching and Linking Cross-Domain Knowledge Graphs using Spatially-Explicit AI
NSF Award 2033521, Lead PI K. Janowicz, UCSB
6. Esri is Working with a Loosely Connected Network of Organizations
Microsoft
National
Geographic
Society
Pristine Seas
Universities
Scripps, UCSB,
Lamont, more
UN
Esri
Users
NGOs
Conservation
Education
Esri
Science
GEO
Blue Planet
SDSN
SDGs
Esri
Partners
To Leverage and Extend This Global Framework
E.O. Wilson
Foundation
Ocean Data
Platform
C4IR-Ocean
Seabed
2030
NOAA
Big Data
Sustainability
POGO
NSF
OceanObs
RCN
9. “GeoAI”
Classification
Clustering
Prediction
• Maximum Likelihood
• Random Trees
• Support Vector Machine
• Empirical Bayesian Kriging (EBK)
• Areal Interpolation
• EBK Regression Prediction
• Ordinary Least Squares Regression and Exploratory
Regression
• Geographically Weighted Regression (GWR)
• K means | Spatially Constrained Multivariate Clustering
• Multivariate Clustering | Density-based Clustering
• Hot Spot Analysis | Image Segmentation
• Space Time Pattern Mining Outlier Analysis
10. “GeoAI” – Predicting Seagrass Distribution in a Warming Ocean
Australia could lose significant seagrass
habitat by 2030
AGU Eos feature, Jan 2018 +
Aydin et al., Estuaries &Coasts, in press
11. “GeoAI” – 3D Modeling of Dissolved Oxygen in Monterey Bay
Flip book esriurl.com/flipbook, p. 44
Learning module https://learn.arcgis.com/en/projects/interpolate-3d-oxygen-measurements-in-monterey-bay/
12. “GeoAI” – 3D Modeling of Dissolved Oxygen in Monterey Bay
Flip book esriurl.com/flipbook, p. 44
Learning module https://learn.arcgis.com/en/projects/interpolate-3d-oxygen-measurements-in-monterey-bay/
15. Tracking, Monitoring, Alerting via Sensor Networks and IoT
Analytics for Internet of Things (including IUU Vessels)
Situational
Awareness
Analytics
Alerting
Real-Time
Data
17. Smart Oceans 2020 | October 5th | Virtual Plenary
Geospatial as an Accelerator of Impact
Already Converging!
Dawn Wright
dwright@esri.com
www.esri.com/sciences --> Ocean Science
@deepseadawn
dusk.geo.orst.edu/Pickup/Esri/smartocean.pdf
18. Smart Oceans 2020 | October 5th | Virtual Plenary
Extra slides for Q&A
19. Geospatial Infrastructure for a Digital Ocean
DistributedServices
Real-Time Measurement
Extensive
Content
StoryMaps
Data Science
Geocoding
GeoEnrichment
Analytic
Services
Big Data
AI & ML
Designed to work together…
Open & Interoperable….
Standards Based….
Secure….
SharedApps
Tracking
Field
Operations
Search &
Discover Publishing
Services
Remote Sensing
Content
Management
Open Data
Containers
Editor's Notes
...involving and supporting many organizations and communities, all seeking to share data sets and services, dramatically extending the impact of geographic information systems or GIS, which provides a SPATIAL FRAMEWORK toward helping ocean data achieve its full benefit
Because Geospatial is special – yes, points, line, areal observations, but photography/videography and other imagery, seismics, surfaces and volumes of ocean parameters. Geospatial lies at the heart of just about everything that matters to us in the ocean such as
WHERE to best establish and enforce additional marine protected areas, especially in the high seas
WHERE and HOW to sustainably feed a rapidly growing population with ocean-based protein?
WHERE to address hot spots of rapidly declining ocean oxygen and increasing ocean acidification?
WHERE to mitigate and adapt to a changing climate? All of these are inherently spatial issues.
Are we not all beginning to create and speak the common SPATIAL language of maps and even 3D scenes that we can mash up and integrate dynamically? Indeed Geospatial is providing a proven framework in a virtuous circle of measurement, analysis & modeling, mapping & visualization, planning & evaluation leading to decision-making and most importantly ACTION in many fields, especially the ones show, which are quickly converging under this geospatial umbrella. They are all building integrated platforms that are bringing together and applying a collective geographic knowledge, sharing it back and forth among these fields and their stakeholders (academia, industry, policy, philanthropy).
To wit, my colleagues and I at Esri are no-cost senior personnel on a new NSF Convergence Accelerator Pilot Phase II Active project LED BY GEOGRAPHERS that greatly desires to integrate with Smart Oceans because of the value add that it is building for a host of domains
Knowledge graphs power search engines, catalogs, apps – building a new much more open web of structured scientific data an open Open Knowledge Network (semantics, ontologies, vocabularies, standards)
Where the X-Ray is an emerging technology called “geoenrichment” which is the enhancing of EXISTING data with additional location-based context (it could be demographics, infographics, blue economy financials, fisheries stock assessments, movement data, migrations, even parameters from climate models). It fills in needed attributes that are currently missing
AI-powered geoenrichment services with an open cross-domain knowledge graph
https://www.nsf.gov/awards/award_visualization.jsp?org=NSF&pims_id=505777&ProgEleCode=095Y,096Y&from=fund#region=US-CA&instId=0013201000
AI-powered geoenrichment services with an open cross-domain knowledge graph
Seeking to add value to a host of domains, including in the future, Smart Oceans!
Knowing where and when things happen is key to understanding why they happen. Our graph, services, and ontologies will be of value across domains and to other OKNs.
This team will also be the first to incorporate satellite data, drone imagery, and maps into a knowledge graph
Our millions of users and thousands of business partners
Collaborations both formal and informal with the ocean organizations and initiatives in yellow
Centre for the 4th Industrial Revolution-Ocean in Norway is an affiliate of the World Economic Forum’s C4IR Network, and is currently building the Ocean Data Platform
POGO = Partnership for Global Ocean Observation
RCN = Research Coordination Network
E.O. Wilson Foundation in building the Half Earth Map
Microsoft in building their Planetary Computer and partnering in AI for Earth
A big part of this effort is our new Living Atlas Indicators of the Planet, which is like a “report card” for the entire Earth. If you go to this URL the indicators fill up your entire browser window. We’ve partnered with @microsoft, @natgeo, and the @UnitedNations Sustainable Development Solutions Network to compile 18 pressing topics affecting our planet with real-time or near real-time data to match. We keep these indicators updated up to the minute with hosted Python notebooks touching the data in the cloud with AI that provides dynamic integration of data from multiple sources, and then running consistent analytics on the fly. Click on each indicator to go deeper with higher resolution maps and resources
The result is a single point to understand the day to day and year to year changes that are occurring on Earth. As you can see these are marked by the relevant SDG and there are several for the ocean. The Living Atlas, by the way is one of the world’s largest catalogs of geospatial data, information, apps (including ~200 million map requests by ~2 million users per DAY) and is being used by Microsoft as a content provider for its AI for Earth initiative and as a backbone for its emerging Planetary Computer initiative.
Speaking of AI, GEOAI, Geospatial AI has made great strides with both vector and raster data in the areas of classification, clustering, and prediction with some examples of the specific types of functions in each category.
Modeling Dissolved Oxygen in Monterey Bay -
https://dusk.geo.orst.edu/Pickup/Esri/Science_flipbook/#p=44
Scientists and researchers often encounter the problem of unknown values within a range of known values. To solve this problem, they use mathematical and statistical methods to create or interpolate new data points. They use these data points to fill in the gaps and model phenomena occurring across a landscape or in 3D space. Researchers use interpolation to accurately predict values for new data points using the existing values of a limited number of sample data points where the measurements were gathered. Researchers have begun using Empirical Bayesian Kriging 3D, a geostatistical interpolation method developed by Esri, to get new estimated values between the known values, and then model the measurements in 3D.
Kriging
Kriging is a method of interpolation used in spatial analysis in which data points exist at specific locations in space and time. Each data point is defined by geographic coordinates (i.e., latitude, longitude, and often, elevation) and other measurements, such as the amount of particulate matter in the air. Kriging applies the basic principle that distributed data points are spatially correlated. This principle assumes that while everything is related to everything else, near things are more related than distant things.
Researchers can use kriging to predict unknown values for any data point, such as elevation, rainfall, chemical concentrations, and noise levels. Empirical Bayesian Kriging 3D is used when the data points are distributed within a geographic volume, such as a square-mile study area of an ocean, ranging from the ocean surface to the ocean floor.
Cross validation
Empirical Bayesian Kriging 3D provides cross-validation tools to assess how well the model predicts values at unknown locations. Cross validation removes a measured point and then uses all remaining points to go back to the location of the removed point. This process is called the “leave-one-out” validation method. The measured value at the hidden point is then compared to the prediction value from cross validation. The difference between these two values is called the cross-validation error and is performed on every input point.
Modeling dissolved oxygen levels in Monterey Bay
Recent research has established that global oxygen levels in the ocean have declined for decades. Dissolved oxygen is the essential ingredient for life beneath the surface of lakes, rivers, and oceans. Using data from the World Ocean Database (WOD) provided by NOAA’s National Centers for Environmental Information, researchers measured the levels of dissolved oxygen by sampling water at different locations in Monterey Bay on the California coast. From above, the sample locations would appear as dots on a map riding on a flat surface. However, researchers sampled at multiple depths at each location, leaving lots of distance between measurements, where dissolved oxygen levels are unknown. Researchers filled in the blanks using Empirical Bayesian Kriging 3D to interpolate the values between these known measurements and create estimated but reliable measurements at each sampling depth.
The result is a 3D map layer stack of surface models, each depicting ranges of dissolved oxygen levels. It is easy to see that dissolved oxygen levels vary across each surface and vertically through the layers. Cross-validation tools and charting allow researchers to explore any location across each surface and slide up and down through the surface layer stack at a specific location.
Modeling Dissolved Oxygen in Monterey Bay -
https://dusk.geo.orst.edu/Pickup/Esri/Science_flipbook/#p=44
Scientists and researchers often encounter the problem of unknown values within a range of known values. To solve this problem, they use mathematical and statistical methods to create or interpolate new data points. They use these data points to fill in the gaps and model phenomena occurring across a landscape or in 3D space. Researchers use interpolation to accurately predict values for new data points using the existing values of a limited number of sample data points where the measurements were gathered. Researchers have begun using Empirical Bayesian Kriging 3D, a geostatistical interpolation method developed by Esri, to get new estimated values between the known values, and then model the measurements in 3D.
Kriging
Kriging is a method of interpolation used in spatial analysis in which data points exist at specific locations in space and time. Each data point is defined by geographic coordinates (i.e., latitude, longitude, and often, elevation) and other measurements, such as the amount of particulate matter in the air. Kriging applies the basic principle that distributed data points are spatially correlated. This principle assumes that while everything is related to everything else, near things are more related than distant things.
Researchers can use kriging to predict unknown values for any data point, such as elevation, rainfall, chemical concentrations, and noise levels. Empirical Bayesian Kriging 3D is used when the data points are distributed within a geographic volume, such as a square-mile study area of an ocean, ranging from the ocean surface to the ocean floor.
Cross validation
Empirical Bayesian Kriging 3D provides cross-validation tools to assess how well the model predicts values at unknown locations. Cross validation removes a measured point and then uses all remaining points to go back to the location of the removed point. This process is called the “leave-one-out” validation method. The measured value at the hidden point is then compared to the prediction value from cross validation. The difference between these two values is called the cross-validation error and is performed on every input point.
Modeling dissolved oxygen levels in Monterey Bay
Recent research has established that global oxygen levels in the ocean have declined for decades. Dissolved oxygen is the essential ingredient for life beneath the surface of lakes, rivers, and oceans. Using data from the World Ocean Database (WOD) provided by NOAA’s National Centers for Environmental Information, researchers measured the levels of dissolved oxygen by sampling water at different locations in Monterey Bay on the California coast. From above, the sample locations would appear as dots on a map riding on a flat surface. However, researchers sampled at multiple depths at each location, leaving lots of distance between measurements, where dissolved oxygen levels are unknown. Researchers filled in the blanks using Empirical Bayesian Kriging 3D to interpolate the values between these known measurements and create estimated but reliable measurements at each sampling depth.
The result is a 3D map layer stack of surface models, each depicting ranges of dissolved oxygen levels. It is easy to see that dissolved oxygen levels vary across each surface and vertically through the layers. Cross-validation tools and charting allow researchers to explore any location across each surface and slide up and down through the surface layer stack at a specific location.
There is also the emerging concept of the “digital twin,” which will be powerful for the ocean – this is a virtual representation of an object, process, or system that bridges the gap between the physical and digital worlds. However, it is MORE than just a visualization, as when implemented with IoT and AI it can accelerate innovation, build consensus, and save time and money by iteratively modeling changes, testing how components or systems function, and troubleshooting malfunctions inexpensively in a virtual world.
Esri has been writing about digital twins in the geospatial realm since 2017, especially with regard to ports (e.g., https://www.esri.com/about/newsroom/publications/wherenext/digital-twin-for-supply-chain-management/)
A recent news article in Science reports that the European Union is finalizing plans for an ambitious “digital twin” of planet Earth that would simulate the atmosphere, ocean, ice, and land with unrivaled precision, providing forecasts of floods, droughts, and
fires from days to years in advance.
To reach the ultimate goal of being the first port in the world to accept autonomous (self-sailing) ships, the port of Rotterdam is already working with Esri to create a digital twin of the port. The port calls this a “moonshot”
Rotterdam’s GIS-powered digital twin would allow port managers to view the operations of all the primary players, provide an accurate, current picture of what is going on in the port, everything from the weather to how many ships are sailing about, their speed, and where they are headed. Simulations would be run digitally to improve efficiency and save money in the real port. Rotterdam officials anticipate being able to pinpoint the best times for ships to berth and offload or take on cargo, because the digital twin simulations will give them the optimal water depths and berth vacancies, among other variables.
https://www.esri.com/about/newsroom/podcast/port-of-rotterdam-the-digital-transformation-of-europes-largest-port/
https://www.esri.com/about/newsroom/publications/wherenext/rotterdam-autonomous-ships-and-digital-twin/
Digital twins rely on sensor networks and IoT especially to monitor and track IUU fishing via AIS and precise radio frequency signals – showing also an R Studio session bridged with ArcGIS where appropriate feature services, image services, REST APIs, JSONs, and geodatabases can be more fully analyzed and interactively mapped
(e.g., Hawkeye 360 is a Silver-level Esri business partner, https://www.he360.com/about/)
HawkEye 360 is a Radio Frequency (RF) data analytics company. We operate a first-of-its-kind commercial satellite constellation to identify, process, and geolocate a broad set of RF signals. They extract value from this unique data through proprietary algorithms, fusing it with other sources to create powerful analytical products that solve hard challenges for our global customers. Their products include maritime domain awareness and spectrum mapping and monitoring; their customers include a wide range of commercial, government and international entities.
And one of our longest standing ocean public-private partnerships to date is with these agencies and organizations to create a 3D basemap of the entire ocean (1/4 deg or 27 km horizontally and at 102 depth zones down to 5000 m) of the top physical parameters that control ocean’s ecology – a massive statistical clustering of global T, S, O2, nitrate, phosphate, silicate NOAA’s authoritative World Ocean Atlas, which in turn is based on the World Ocean Database of the UNESCO IOC IODE and NOAA. Published in these outlets, also available in apps for the web, phone, or tablet.
And a number of use cases are underway to deploy this 3d basemap for ecological studies or better yet to build biological/biodiversity data INTO the EMUs themselves
the capabilities foundational for digital oceanography, to strengthen the geospatial infrastructure that you’re already creating, a digital nervous system of sorts that’s intelligent, open, interoperable, based on standards, and secure. This is Esri’s ecosystem of tools that extend well beyond mapping, from data collection in the field to artificial intelligence/machine learning functions. This is all extremely powerful. Yes, I know. You had no idea!
And a number of use case underway to deploy the 3d basemap for ecological studies or to build biological/biodiversity data INTO the EMUs themselves
Wright, D.J., Kavanaugh, M.T., Henry, L.-A., Brandt, A., Saeedi, H., Bednarsek, N., Van Graafeiland, K., Butler, K.A., Breyer, S., and Sayre, R.G., Use cases of Ecological Marine Units for improved regional ocean observation data integration within the Marine Biodiversity Observation Network (MBON) (Highlighted), Eos, Trans. AGU, 99, Fall Meet. Suppl., Paper B41L-2897, 2018.
We report on a series of use cases underway to augment and test the viability of the global Ecological Marine Units (EMUs). EMUs were commissioned in 2015 by the Group on Earth Observations (GEO) as a means of developing a standardized and practical global ecosystems classification and map for the oceans. They are a key outcome for the GEO Biodiversity Observation Network (GEO BON), and a recent contribution to the Marine Biodiversity Observation Network (MBON).EMUs are comprised of a global 3D point mesh framework of 52 million ocean observations of salinity, temperature, dissolved oxygen, nitrate, silicate, and phosphate from the NOAA World Ocean Atlas. Many cite the need to scale this global framework down regionally and up temporally. Hence, over 15 teams of researchers are implementing EMUs in regional use cases, based on their own higher-resolution data for a richer geospatial accounting framework and visualization of species distributions.
Among these are use cases in temperate upwelling, shallow subtropical and polar regions, where boundaries of surface seascapes are compared to surface EMUs, and at seasonal scales. The EU-funded ATLAS project is comparing EMUs to species-based biogeographic clusters of Vulnerable Marine Ecosystems in the North Atlantic to further refine UNESCO's Global Open Ocean and Deep Seafloor effort for this region. German researchers compiling 5000-6000 deep-sea distribution records from expeditions to the Sea of Okhotsk, the Aleutian Trench, and the Kuril-Kamchatka Trench are comparing their EMU use case with the ATLAS use case. Another use case seeks to add data on NE Pacific carbonate chemistry and pteropod shell dissolution to the EMU 3-D point mesh network to provide information on the responses of ecosystems to influences such as ocean acidification.
In sum, we are building a strong user community based on these use cases to improve understanding of global and regional drivers of biogeography, refine tools to classify and prioritize areas for improved marine management including area-based management tools, and to enhanced visualizations of ocean trends and/or forecasts.