- Author: Shane T Feirer
The first day of the ESRI User Conference was a good day of new technologies and good stories of Conservation and GIS. There were approximately 18,000 attendees in San Diego for the Conference with an additional 10,000 attending virtually.
Jack Dangermond and ESRI highlighted their new offerings like ArcGIS Reality that allows users to generate a digital twin of at scale from buildings, cities, to entire countries. ESRI has also released new story map templates (the brief, e-magazine) and theme and text enhancements. They have built in new enhancements to the ArcGIS Dashboards, these enhancements include better mobile support and additional support of Arcade.
There have been considerable investment in enhancing Spatial Analysis and Data Science, these included:
- 60+ New Tools
- GeoAI Models (Deep Learning)
- Analysis in the Map Viewer
- Time Series Forecasting
- Predictive Modeling
- Multivariate Indexing
- Pie and Donut Charts
- Big Data Tools
I am looking forward to learning more about the enhancements in the coming days.
Beyond ArcGIS Online and its apps, ESRI highlighted its work on integrating the ArcGIS Mapping system into Office 365 applications. Users of Office 365 can now easy map data in spreadsheets, in teams, sharepoint these new tools. This could bring more mapping technology to UCANR Academics and Staff.
- Interested in hearing about the latest projected climate data just released from California's Fifth Climate Change Assessment?
- Have a burning question about how climate change will impact California's agricultural systems?
- Want to hear how researchers, industry leaders, and state agencies are thinking about decision support for climate adaptation?
If you answered yes to any of these questions, join us at the California Adaptation Forum on July 31, 2023 in Pomona (one of the workshops will also be on Zoom) for a pair of pre-conference workshops that will address these and other topics.
The California Adaptation Forum, being held in person for the first time in 5 years, is the premier conference in the state that brings together community leaders, state and local agencies representatives, researchers, and others to share and discuss how Californians are adapting to climate change today and preparing for the future. Pre-conference workshops will be held on Monday July 3, including:
Using Climate Data & Tools for Decision-Making with Cal-Adapt
July 31, 1:00 – 2:45 pm. In-person
Now more than 10 years old, Cal-Adapt is California's primary source for downscaled climate data. This workshop will review the data and tools available through Cal-Adapt.org, and describe the latest generation of downscaled climate data just released from the Fifth Climate Change Assessment. You'll also get to see the upcoming “Cal-Adapt: Analytics Engine” in action, talk to the researchers who built it, and play with the tools in interactive working sessions. Bring your questions, curiosity, and enthusiasm to dive into the data and explore how
climate projections can help you and your community prepare for climate change.
- Nancy Thomas (Geospatial Innovation Facility, UC Berkeley)
- Grace Di Cecco (Eagle Rock Analytics)
- Nancy Freitas (Lawrence Berkeley National Lab and UC Berkeley)
- Justine Bui (Spatial Informatics Group)
Advancing decision support for climate adaptation in agriculture and natural resources
July 31, 2023. 3:00 – 5:00 pm. In-person and on Zoom
Information on climate change is one of our best resources for adaptation planning. We have a pretty good idea how weather and climate can affect agricultural and natural systems, and we have pretty good data on climate trends. Connecting the two is the realm of decision support tools. In this workshop, we'll hear from researchers asking questions about the impacts of climate changes on tree crops, agricultural pests, urban trees, and water resources. In the second half of the workshop, participants will be invited to discuss their burning questions about climate adaptation, and talk with researchers who are trying to harness climate data to develop decision support tools for adaptation.
- Tapan Pathak (Professor of Extension, UC Merced)
- Andy Lyons (Program Coordinator, UC ANR)
- Bob Klein (California Pistachio Research Board)
- Janet Hartin (CE Advisor, UC ANR)
- Michael Wolff (Senior Environmental Scientist, CDFA)
- Romain Maendly (Climate Change Technical and Policy Advisor, DWR)
Zoom info: This workshop will be held in person and on Zoom. Zoom participants will be included in the entire workshop - presentations, Q&A with the presenters, and breakout discussions. Register here.
- Author: Andy Lyons
- Author: Sean Hogan
- Author: Maggi Kelly
Find out what our GIS experts say about ChatGPT's answers to GIS questions
ChatGPT has taken the world by storm. Most of the media coverage has been around its ability to write essays, cover letters, and even poetry. But its language model is drawn from the web, which covers a lot more than prose. Its programming model in particular has impressed - and even scared - professional software engineers and data scientists.
But domain experts who have tested ChatGPT also caution that using the world wide web as the corpus of knowledge for training ChatGPT is both a blessing and vulnerability. Because as we all know, the web is a mixed bag of content. A lot of it is accurate and of good quality, but not everything. It takes humans years of experience to develop the wisdom to sift the wheat from the chaff - how is ChatGPT doing after just a few months out of the oven?
We decided to test it out with our domain knowledge - geospatial data, science, and technology. Like everyone, we're curious about the accuracy of ChatGPTs answers to common GIS questions that have a known answer, and how it deals with questions that don't have a single answer (like which GIS software is “best”). But we were also interested in what types of questions it can handle, and when it essentially gives a non-answer.
To kick the tires on ChatGPT's geospatial domain knowledge, we asked it a series of questions in a few categories that we have a lot of experience with: general GIS, GIS software (i.e., ArcGIS), and spatial programming with Python and R. Our experts then gave each response a score of 1-5 stars, plus a little commentary.
See below the questions we asked, or jump to the end for our conclusions. The results might surprise you!
General GIS Questions
1. What's the best spatial projection to use for a GIS analysis of a forest?
IGIS experts score: 4.5 stars
Slightly more emphasis could have gone to scale, and the trade-offs of preserving area, direction, positional accuracy, and cartographic shape; however, this is a very good answer! SH
2. What is the most common format for spatial data?
IGIS experts score: 4.5 stars
ChatGPT answers the question accurately, however I was struck by the oddity of including “GeoTIFF” as a competing or comparable file format to a Shapefile, given that a GeoTIFF is a raster format and a Shapefile is a vector format. This comparison is not technically wrong, but requires further explanation. SH
What is the best format for vector spatial data, and why?
The best format for vector spatial data depends on the specific needs of the analysis and the software being used. However, some of the most common and widely used formats for vector spatial data include:
- Shapefile (.shp): Shapefile is a widely used format for vector data that is supported by many GIS software packages. It stores geometric and attribute data in separate files and can represent point, line, and polygon features.
- KML (.kml): KML (Keyhole Markup Language) is an XML-based format for storing geographic data and is often used for displaying geographic data in Google Earth.
- GeoPackage (.gpkg): GeoPackage is an open, standards-based format for storing geospatial data that can store both vector and raster data. It is designed for use on mobile devices and is gaining popularity as a format for sharing geospatial data.
The choice of format will depend on factors such as the purpose of the analysis, the size and complexity of the data, and the software being used. It is always a good idea to consult with experts in the field, such as GIS professionals, to determine the best format for a given analysis.
IGIS experts score: 3.5 stars
ChatGPT does a good job at picking some good (and common) vector data formats; however, in my opinion it doesn't do a great job at explaining why its selected formats are the best, but rather just gave some basic examples of use-cases. SH
What is the best source of data to make a map of wildfire risk in California?
IGIS experts score: 4.5 stars
This is a very good answer for primary sources of fire risk data; however I would have also liked to have seen mention of secondary data sources, such as ESRI or Living Atlas. SH
Is it possible to classify historic black and white aerial photography to distinguish between grasses, shrubs, and trees?
IGIS experts score: 2.5 stars
This answer would have been good, if it were not for the third paragraph, which doesn't really apply to black and white imagery in any practical sense. Near-infrared reflectance and black and white imagery are very much two different things. SH
GIS Software Questions
Where do I go in ArcGIS Pro to split a big polygon into multiple polygons using a road network to define the regions?
IGIS experts score: 2 stars
I attempted to use the above work-flow to split a polygon, and the directions were both inaccurate and not very clear. In my opinion, a better way of tackling this geoprocessing step would have been to use the search tool in the geoprocessing tool box, as this approach would be much less likely to change with updated versions of ArcGIS Pro. SH
Write Python code to convert a multi-layer raster stack to Zarr.
IGIS experts score: 5 stars
The provided code works and is easy to follow. I was impressed that it offered an alternative library and anticipated the next question - how to upload to cloud storage (because Zarz is a cloud optimized format). AL
Write a R script to import a polygon layer from a file geodatabase, and plot it using leaflet.
IGIS experts score: 4 stars
The code is correct, and the explanations are helpful. Impressive that it knows how to combine functions from two R-packages that are often presented separately. It didn't however check the projection of the polygon data to verify the data are in WGS 84, which is required by leaflet. AL
Overall we found ChatGPT very impressive. We gave it a series of questions on geospatial concepts, data, desktop software, and programming, and in all cases it was able to interpret our questions sensibly. Although it missed some of the details, the responses were always relevant, and in many cases it was able to bridge different domains of knowledge (such the scale of projections and the scale of forests).
Where it works well: ChatGPT is most impressive in domains like coding that are fairly well-defined and have highly structured ontologies and syntax. It's probably not too surprising that a computer program is most proficient in understanding computer languages. Programming is also probably one of the easier domains for it to train on: answers to common questions on the internet probably converge pretty quickly. We predict it will have the greatest utility for questions on getting started, for example which function to use, and well-defined use cases. Most advanced topics, like debugging or designing the architecture of a program to address a specific use-case, are much more complex and will continue to require human expertise and experience synthesizing information.
Other fields it might work well in. Extrapolating this conclusion further, we could also predict AI could become very useful in fields like law, medicine and other highly technical areas where domain knowledge is well-developed, highly structured, and requires storage capacity well beyond what the human brain can handle. It will be interesting - and scary - to see how AI intrudes upon fields that have historically been seen as some of the most intellectually demanding and prestigious occupations out there.
AI in education: There are many active discussions about the use of AI in education, as there is no doubt that these tools are going to change education and training. There is considerable worry about how these tools might facilitate exam cheating in higher education, and some instructors have responded by trying to keep AI tools out of the classroom. However, that seems like trying to hold back an ocean back with a sand castle. Many teachers are figuring out ways to embrace and incorporate tools like ChatGPT in their teaching. It is clear that these tools can also help to accelerate the learning curve, but also require strengthening skills in discernment, critical thinking, the value of diverse perspectives and experiences, and the ways in which knowledge serves as a form of power.
Learning how to talk to ChatGPT: We found that to get good answers you have to ask good questions. This is perhaps an extension of the well-known adage, ‘garbage in, garbage out' (which also applies to ChatGPT's training data). As we learned from trial and error, if you don't ask a clearly expressed question, using standard terminology and with all the needed parameters, the results are more unpredictable, general, and probably not helpful. Again this is not surprising given that you're talking to a computer who doesn't know your background and can't read your body language. Indeed one of the skills we think will need to be taught to use ChatGPT effectively is articulating good questions. Perhaps not coincidentally, learning how to ask better questions is also one of the primary learning outcomes of many K-16 curricula.
Recognizing and reconciling multiple realities and knowledges. Our domain of interest - geospatial science, data, and tools - is fairly cut-and-dry, and the ‘one universal truth' paradigm isn't so far fetched. That being said, there are still shades of gray, old and new standards that often coexist together, scale dependencies, data modeling choices, philosophical differences, power dynamics, etc. To its credit, when there isn't a single answer or the decision has consequences, ChatGPT likes to invoke ‘It is always a good idea to consult with experts in the field…' That often feels like a cop-out, but it's probably the right thing to do.
In other domains, such as many social science fields and the humanities, multiple situated knowledges and interpretations are not an outlier but an intrinsic characteristic of the field. Whether this new generation of AI can recognize and articulate the value in different types of knowledge, time will tell.
What we'd like to see more of: transparency. Transparency is a tall ask for AI. The algorithms are extremely complex and fuse together training data in ways that are hard if not impossible to unpack (even if you know the algorithm!). Not knowing where information comes from is a huge paradigm shift from traditional scholarship, and not just a bit unsettling. In conventional research, transparency is baked in through practices like literature reviews and citations. This is how scholars build upon the work of each other, allowing the reader to trace the lineage of knowledge and evaluate for herself whether a new discovery or idea holds water. ChatGPT on the other hand draws from a massive treasure trove of online content, however the specifics of how sources are chosen and evaluated are complex and known only to an elite few software engineers. To say its sources are anonymous is incorrect - they're just hidden. You don't have to be a George Orwell to think of ways this could produce unintended and undesirable consequences in a big way.
Does ChatGPT level the playing field? Technologies like ChatGPT can level the playing field to some degree, allowing poorly resourced communities to catch up in terms of education and technology. Think of the positive impacts on universities in poor countries. However once transformative technologies get monetized, they tend to exacerbate existing inequalities. Hopefully the powerful companies that are developing these tools will listen to the smart people inside and outside the company and find ways to keep this technology in the public domain.
In conclusion, the journey to Cyborg is a slippery slope. Long before ChatGPT, people freaked out about the transformative impacts of search engines, cell phone apps, and even Wikipedia. Is ChatGPT going to be a game changer in the world of GIS? Only time will tell. We will definitely be using it for what it's already good at, and are not too worried about our jobs being taken over by robots - yet!
- Author: Andy Lyons
- Author: Sean Hogan
- Author: Maggi Kelly
- Contributor: Shane Feirer
The IGIS Team is pleased to share our workshop schedule for Spring 2023. Workshops are the best part of our 'trilogy' of strategic goals - research, technical projects, and training - because this is where we get to share the tips and tricks we pick up every day. See descriptions below, and if you're not available to attend one that really interests you keep an eye open for the recordings on our YouTube Channel.
Introduction to ArcGIS Online
Friday January 27, 2023 • 1:00 - 4:00pm • online • free
ArcGIS Online is the online component of the ESRI geospatial ecosystem, and the foundation for web mapping, story maps, and mobile data collection. This workshop will provide an overview of ArcGIS Online and teach participants how to create a web map. This workshop is a prerequisite for the Story Maps and Field Maps workshops, and recommended for ArcGIS Pro.
Requirements: No experience required. Participants must have an ArcGIS Online account set up prior to the workshop (free for all UCANR employees, temporary accounts available for others). Details and registration.
Introduction to ArcGIS Pro
Friday February 17, 2023 • 1:00 - 4:00pm • online • free
ArcGIS Pro is ESRI's powerhouse desktop application for all things GIS. It can do anything from basic cartography to advanced geospatial modeling. This introduction will get you started creating maps with local and online GIS data.
Requirements: Participants must have ArcGIS Pro installed on their personal computer prior to the workshop. ArcGIS Pro is for Windows only. Licenses are free for all UCANR employees, and temporary accounts available for others. No experience needed, but the ArcGIS Online workshop (January 27) or equivalent experience is strongly encouraged. Details and registration.
Introduction to Jupyter Notebooks in ArcGIS
April 7, 2023 • 1:00 - 4:00pm • online • free
Jupyter Notebooks are a user-friendly and interactive way to write Python code. ArcGIS Pro supports Juptyer notebooks natively, opening the door to a wide range of options for automation and extensibility. This workshop will get you started using Jupyter Notebooks in ArcGIS Pro to automate workflows, perform geoprocessing tasks, create data summaries, and import downscaled climate data from Cal-Adapt.
Requirements: Basic familiarity with ArcGIS Pro is expected. Licenses for ArcGIS Pro are free for all UCANR employees, and temporary accounts available for others. Experience with Python is helpful but not required. Details and registration.
June 26 - 30, 2023 • CSU Monterey Bay
DroneCamp is a unique, one-week intensive short course that covers everything you need to know to get started using drones for data collection and research: equipment, safety and regulations, flight planning, flight instruction, data processing, and data analysis. Now in our 7th year, DroneCamp is a collaborative effort between UCANR/IGIS, CSU Monterey Bay, UC Santa Cruz, UC Merced, and others.
Requirements: No experience necessary! Registration will open in January 2023. This program has always sold out so save your spot early. Scholarships available. For details, see https://dronecampca.org/.
Access to fast internet isn't just a luxury anymore. In our modern age, broadband is a necessity to take advantage of all kinds of basic services, from education, to banking, civic engagement, navigation, and even health care. Most broadband in the USA is provided by Internet Service Providers companies (ISPs), and while we're getting closer to universal coverage there is still a ways to go!
California was an early pioneer in broadband mapping, to show exactly where broadband is available and where it isn't. For the past 10 years, the Federal Communication Commission (FCC) has been collecting broadband data from ISPs for the entire country, and putting it on a national map. Actually there are two maps - one for fixed internet access points (i.e., homes and businesses), and one for mobile broadband. The latest version of the national broadband map was released in late November, and the FCC is now inviting the public to submit feedback through January 13, 2023.
Although ISPs are the primary sources of data, the FCC encourages everyone to check out the map. If the reported availability and speeds don't seem to tally with your experiences, they've also made it easy to submit feedback, known as ‘challenges'. And indeed thousands of citizens have been doing exactly this.
What does this have to do with me?
Exploring the National Broadband Map can be fun and is definitely educational - especially this year where the level of granularity is higher than ever. You can see which ISPs are serving your area, and what kind of connection speeds they claim.
But the map isn't simply for public information. The map with all the feedback submitted is also being used to identify where internet infrastructure investments are needed most to achieve the ultimate goal of universal access. And if you've been following the news recently, there's a lot of funding coming out of Washington and Sacramento right now to improve digital infrastructure.
Aside from helping steer resources where they're needed most, if you decide to contribute data to the national broadband map, you'll be joining one of the largest citizen science projects in the country. Thousands of residents just like yourself have used the mobile app and other tools provided by FCC to check their actual internet speed. And if you discover the claims by the ISPs are perhaps a bit rosy, you have the option to submit feedback in the form of 'challenge' data.
How do I participate?
The map is designed for the public. You can use it to see what internet service is available in your area, and the expected speeds. If you still want to keep going, you can double-check your cellular internet speed outside or while driving using a phone app. If you find any discrepancies on either the mobile or fixed broadband maps, you can submit feedback through the website or the mobile app. If enough people submit similar feedback from your area, the FCC will ask the ISPs to explain themselves, and either update the map or improve service.
First things first is to simply visit the map and find your area. You can type in an address or the name of a city (then wait a second for the drop down box to update), or use the pan and zoom controls to find your area. Once you've found your area, remember there are two maps to look at. The fixed internet map (with the green dots) will probably include your house if any ISPs offer service in your area (whether or not you use their service). The mobile broadband map shows the expected cellular internet speed using hexagons that get smaller as you zoom in.
Click on the map to see what it says about your upload and download speed. Internet speeds are measured in megabytes per second (Mbps), and to qualify as 'broadband' the ISP must offer at least 25 Mbps for download and 3 Mbps for upload (that definition may soon be changing). Remember that the reported speeds are provided by the ISPs, and may or may not be what you actually experience.
Verify your mobile data speed
Want to keep going? Let's start by looking at how you can test your mobile (cellular) internet speed, and if you desire how to report what you actually experience.
To measure the actual connection speed you're getting on your phone or mobile device, you can download the 'FCC Speed Test' app (available for Android and iOS). A few caveats about using the mobile app:
- The whole point of the mobile app is to test the speed of mobile broadband. This means:
1) you need to disconnect from WiFi when you use the app to test connection speeds, and
2) testing is going to eat into your cell phone data usage. Many cell phone plans come with unlimited mobile data, in which case this may not be a big deal. But others give you a monthly quota, after which they slow down, cut off, and/or charge you more. When you install the app, you'll be asked how much data to allow it to use within your monthly billing cycle, from 500 MB up to 10 GB.
- Sharing your data with the FCC is optional. If you just want to use the mobile app to see your connection speed, you can opt-out of sharing data and skip entering your name and email.
- The app does however require permission to access your location to run a test, even if you're not sharing the data. If this makes you nervous, there are other apps you can use to check your internet speed that don't ask for your location.
- The national mobile broadband map reports connection speeds outside. The FCC isn't particularly interested in how fast your cell phone data is inside your house, which of course is affected by all kinds of other things starting with how thick your walls are. You can use the app anywhere, but to submit data to the FCC make sure you step outside or go for a drive.
- Depending where you live, you may have noticed that internet speeds vary throughout the day. For example, when a lot of people stream video during prime time, individual speeds may go down. The FCC wants to know about that also. The app has an option to run tests periodically in the background. If you choose to enable this, it will still stay within the overall monthly cap you specified when installing the app.
Provide feedback on the fixed broadband map
The feedback the FCC wants on the fixed broadband map is a little different. Like the mobile app, they're interested in hearing if your internet speed is significantly different from what the ISPs report. But more fundamentally they're interested in whether the ISP service is actually available in practice. If you've tried to order service but the ISP never replied, or asked for an excessive installation fee, the FCC wants to know about that. They also want to know if your address is mislocated on the map, or in some cases missing completely!
To submit a challenge about the availability of service or the accuracy of a location, you essentially complete a web form. Click on the location on the map, and then either the ‘availability challenge' or ‘location challenge' links. For a preview of what that looks like, see this YouTube tutorial.
What about affordability?
We all know that ‘availability' only translates to ‘accessibility' if the service is affordable. For better or worse, the FCC national broadband map does not attempt to capture affordability. You can (and should) use the 'Availability Challenge' link to report if the installation fee for fixed broadband is excessive. However beyond that, the national map may not be the best place to register a complaint about broadband affordability.
Work with your community
While the FCC has made it easy to submit challenges to the map, that doesn't mean it will ask the ISPs to respond to every single submission. They need to see several challenges for the same area before they will aggregate them and ask the ISP to respond. So if you really feel the map needs to be reality-checked, encourage your neighbors and community to get involved. You will get feedback if your challenge is aggregated and sent to an ISP. Plus you never know, your efforts may just lead to an update on the map, with upgrades to infrastructure and service to follow!