ABCDEFGHIJKLMNOPQRSTUVWXYZAA
1
Working with eBird Status and Trends Data Products: Part I, Introduction to the eBird R Package
2
Workshop Live QuestionsLive Answers
3
1What to do about this?
> set_ebirdst_access_key(8dv3vnccfd8h)
Error: unexpected symbol in "set_ebirdst_access_key(8dv3vnccfd8h"		Try: set_ebirdst_access_key(“8dv3vnccfd8h”)
4
2> set_ebirdst_access_key(“8dv3vnccfd8h”)
Error: unexpected invalid token in "set_ebirdst_access_key(“"		live answered
5
3> set_ebirdst_access_key(“8dv3vnccfd8h”)
Error: unexpected invalid token in "set_ebirdst_access_key(“"		You may need to request a new access key -- perhaps the last one has expired? https://science.ebird.org/en/status-and-trends/download-data
6
4> set_ebirdst_access_key(“8dv3vnccfd8h”)
Error: unexpected invalid token in "set_ebirdst_access_key(“"		I had the same problem, but when I realized that i needed to load the ebirdst package (> library(ebirdst)).

Once that was done, I was able to load and store my access code.
7
5Same with me
8
6Do I need the ebird r package (I Forgot to download it)?A portion of the workshop requires the ebirdst R package. Installation instructions here: https://ebird.github.io/ebirdst/
9
7will you being sharing you sripts/ R files after?Yes, we will! All the code that Matt goes through today will be available after the webinar.
10
8Yay! You have access to download data products.
Your Access Key: 8dv3vnccfd8h
Expires: 2025-04-08
11
9I tried, but I got a message saying that the ' operation you requested cannot currently be performed'. Is this a problem with my pc?This sounds like it may be a problem with your PC. What happens when you re-install with the install.packages("ebirdst") function?
12
10Hi, is the delay for the 2023 data release for the status data?Yes, the data products that would have been released this fall, predicted to year 2023, will be release sometime in early 2025.
13
11what is denominator for "relative" abundanceRelative abundance is a prediction which is standardized for each species. It's 'relative' because it is standardized based on a big set of variables describing effort and detectibility for each species.
The definition of this unit is here: https://science.ebird.org/en/status-and-trends/faq#mean-relative-abundance
14
12is this interpolated into rarely scouted regions?These machine learning models do make predictions into areas with less data, based on what they have learned about relationships with the landscape they have learned from available data. This is discussed somewhat in this FAQ item: https://science.ebird.org/en/status-and-trends/faq#errors
15
13I've been tinkering/devloping a code based off the orginal R package from Dr. Dokter. I'm curious if you have found a workthrough to handle the pure volume of data that might come from reading NEXRAD data in a "real time" fasion [hosted through to my site]. I think I herd rest services here.. fingers crossed.Hi Isaac, this webinar is about the R package "ebirdst". For questions about R packages to deal with radar data, I would suggest reaching out directly to those package developers.
16
14What data is used to define breeding and non-breeding seasons? Is this based on eBird data exclusively or does it use a variety of data sources including birds of the world species accounts?Experts who review the Status data products set dates for two reasons: a) to make these combined seasonal maps, and b) to define dates for Trends analysis. These experts do this by looking for a majority of the population to be stationary, which is important for the Trends analysis.They draw on their own knowledge of the species and other sources like Birds of the World, but they are primarily looking at the animation of weekly data. Read more here: https://science.ebird.org/en/status-and-trends/faq#seasons
17
15How do I download this? This is probably a dumb question, but I have no knowledge on things like... github and this code:if (!requireNamespace("remotes", quietly = TRUE)) {
install.packages("remotes")
}
remotes::install_github("ebird/ebirdst")		We will discuss these parts in about 20-30 minutes. Are you familiar with the R programming language?
18
16Is there a way to export the shapes you drew to get the breeding population proportions Or to upload a kml or shp instead?When you draw a shape you can export the summaries and with that comes a CSV file with the Well Known Text (WKT) of the polygons you drew, which can be imported into R or GIS. You cannot upload KML or SHP at this point.
19
17Just to clarify, the weekly estimates are “on average” for how many years? could the model track changes in weekly abundance across different years?The weekly estimates use data from all years to train the machine learning model. When we predict to a certain year and gridcell, we use the variables (e.g., land cover conditions) specific to a single prediction year.
20
18Is the estimation of abundance relying on some kind of urbanisation index or is it completely independant and just using eBird observations ?A previous Birds of the World Webinar described the methdology for creating these abundance estimates (https://www.youtube.com/watch?v=VKh9L6hjdlM&t=194s).

These are machine learning models that are trained on over 150 features describing the observation process and the landscape, which includes a nighttime lights index to describe urbanization.
21
19Nope. I'm pretty new to ebird, and I've only used it to learn, identiFy, upload and study on my local birds.This webinar is about using the eBird Status and Trends data products, which are modeled from eBird data, to do futher analyses or applications. You don’t need the eBird app for today, but what software you need will depend on what you want to do with the data products.
22
20why estimate for a recent year (2021) over estimateing for average overall?The machine learning models we’re using are designed for prediction and they perform much better when we include year as a feature or predictor. However, this means that, for computational reasons, we must choose to make a prediction for a single year. That said, this single year is fairly representative of recent years, because the year feature is treated as a continuous, numerical feature in the model, so borrows more readily from recent years. You can think of it as a smooth, in a sense.
23
21Can one draw a polygon in a region to get all species documented in that polygon? as well as the related information?Because the eBird modeled layers (like relative abundance) are species-specific, you would need to do this separately for each species.
There is a speedier way to do this using the R package, which includes a download giving the % of each species population (among other summaries) in all countries, states, and provinces in the world. When we send out a feedback form later in the session, please submit this request and we'll show you how to do it next session!
24
22Are these products available as map services to bring into web maps?At the moment, no. The Terms of Use of the products prohibit this at the current moment.
25
23Are there any plans to implement trend summaries outside the breeding season (specifically migration?)
26
24Is it possible to select trends fro specific years? For instance, a positive trend between 2018-2021, but then a negative trend between 2021-2024 will give a “uncertain” change. In addition, if the whole range of the species is included (from 2012-2022), how is handled range expansions in locations not occupied in the previous years; for example, a species colonized a region in 2018 does not have estimates in 2012, or the change from 2012 (o individuals) to 2022 will be positive, no mather a transient dynamic is ocurring (positive and then stationary abundance). How to track this?This is a great question! Right now, the data are only estimated as the change for a specific baseline (e.g., 2012), not the year-to-year trajectories. The model is able to detect a range expansion, and it will show up as a very high % per year increase. For example, going from very rare to relatively common might indicate a % change of greater than 100% over the time period.
27
25Is it possible to select trends fro specific years? For instance, a positive trend between 2018-2021, but then a negative trend between 2021-2024 will give a “uncertain” change. In addition, if the whole range of the species is included (from 2012-2022), how is handled range expansions in locations not occupied in the previous years; for example, a species colonized a region in 2018 does not have estimates in 2012, or the change from 2012 (o individuals) to 2022 will be positive, no mather a transient dynamic is ocurring (positive and then stationary abundance). How to track this?
28
26It seemed very few people reported Long-tailed Shrike from souther China and SE Asia. It is a common species there. So there may be a bias if the data is not representing the true status.This is a great question. The modeling workflow is designed to deal with some of these areas of data sparsity, but, like any model, it's not perfect. In general, we find that if there are areas with good data representation, the model can still do an excellent job predicting relative abundance to other areas with less data. Each species goes through a series of quality controls and predictive performance checks before going on the website. Note that some local areas may still have descrepancies -- that's a good place to fill in the missing data and submit eBird checklists!
29
27It seemed very few people reported Long-tailed Shrike from souther China and SE Asia. It is a common species there. So there may be a bias if the data is not representing the true status.thanks, we are trying to get more people use eBird in these countries.
30
28why are the trends estimated for different seasons in different regions?Trends are estimated for the same season across the entire range of each species. For some species, the trends are breeding season. For others, the trends are nonbreeding season. This is just based on the quantity of the available data, but it's a species-level (not region-level) decision.
31
29I have been using python and R for all of my analysis so far, do you think there are benifts to using Arc/QGIS or is it just preference ?Good question -- it's just a preference. If R and python are working well for you, keep it up! QGIS is a useful tool for people who are not as familiar with R but want to visualize the data quickly. It is also handy for making pretty maps!
32
30Are the confidence intervals actually Bayesian credible intervals from a posterior distribution?For eBird Status, the CIs come from an ensemble modeling approach, so they are not Bayesian posteriors.
For eBird Trends, the CIs come from a series of replicated models producing a distribution of trend estimates at each point on the map. More similar to bootstrapping than Bayesian.
33
31Are the confidence intervals actually Bayesian credible intervals from a posterior distribution?Ok, thank you for clarifying!
34
32Anyone know what the translate tool used in QGIS to get rid of the zeros is for ArcGIS Pro?I am unaware, but FYI the translate tool in QGIS is the gdal_translate command line tool from the GDAL libraries.
35
33sorry I may have missed it - where can I access QGIS?https://qgis.org/ - go to “downloads” and select your platform!
36
34Tom is using the software QGIS as tool for visualizing the products. You can download that software here https://qgis.org/download/
37
35
38
36
39
37@Andrew Stillman Is QGIS amenable to a reproducible workflow or are interactions always through the app’s UI?I believe from the background QGIS uses python coding. You can save the code of your workflow or use your own coding in R or Python if you are familiarized with spatial analyisis with coding
40
38@Andrew Stillman Is QGIS amenable to a reproducible workflow or are interactions always through the app’s UI?I don’t use QGIS in that way, but Orlando is right, you can write Python scripts to do analysis and interface with QGIS. There may also be other methods for creating reproducible workflows in QGIS.
41
39Is this accesible later as a video? I'm not ready For this level oF programing yet.Yes, this video is recorded and it will be posted publically for people to re-watch later. We will also share the R code used in this webinar.
42
40Is this accesible later as a video? I'm not ready For this level oF programing yet.
43
41Can we get a download of the QandA?Thanks for bringing this up. I was just checking this with the meeting hosts. We’ll put the meeting chat in the blog post that comes out after this webinar (along with the video recording!).
44
42
45
43Will we learn how to estimate trends data for species that are not presented in the Status and Trends website?We don’t have plans to give webinars on this topic. Please refer to the methdology paper: https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.14186
46
44Will we learn how to estimate trends data for species that are not presented in the Status and Trends website?Are some presented on the website but not in the package? I'm trying with sarus crane, which is on the website at https://science.ebird.org/en/status-and-trends/species/sarcra1/abundance-map, but am getting an error:
> ebirdst_download_status(species = "sarcra1")
Error in ebirdst_download_status(species = "sarcra1") : The requested species was not modeled by Status and Trends.
47
45Will we learn how to estimate trends data for species that are not presented in the Status and Trends website?Unfortunately, some species are spread across two version years, which will be doing away with during the upcoming release. Apologies.

You can access it programmatically by installing previous version of package: remotes::install_version("ebirdst", version = "2.2021.3")
48
46Is there any reason that QGIS may be preferred over ArcGIS for this use? (Not even sure of the difference myself, might be more for me to research, but interested in hearing any tips/thoughts relating to this work.)live answered
49
47Is there any reason that QGIS may be preferred over ArcGIS for this use? (Not even sure of the difference myself, might be more for me to research, but interested in hearing any tips/thoughts relating to this work.)My personal thought: licences!!! QGIS can be used without licenses. Many users without ArcGIS licenses can’t work with that software
50
48Is there any reason that QGIS may be preferred over ArcGIS for this use? (Not even sure of the difference myself, might be more for me to research, but interested in hearing any tips/thoughts relating to this work.)QGIS also works on Macs, whereas ArcGIS doesn’t (without something complicated like running windows on your Mac)
51
49Can you download trends for several species at the same time?You can do this with the R package functionality for downloading Trends data products, which we will cover particularly in the third webinar, but Matt will preview today.
52
50Nice comment about ArcGIS!
53
51is there a way to visualize the CIs for the trend info?Yes, the downloaded data format has the confidence intervals, so you could calculate CI width if you wanted or just visualize them directly. Matt will also likely cover this topic directly in the third webinar.
54
52is there a way to visualize the CIs for the trend info?ty
55
53How did you develop status and trends for so many species? It seems like a huge undertaking! Did you have a repeatable workflow you used for each species?It absolutely IS a huge undertaking! It requires the use of super-computing clusters to scale it up to many species (e.g., through NSF). It's a standardized production workflow for all species, and it's run by a team of computer scientists, data scientistics, and statisticians.
56
54How did you develop status and trends for so many species? It seems like a huge undertaking! Did you have a repeatable workflow you used for each species?
57
55Any idea when the recording of this webinar will be made available?It will take a few days to process, but then it will be available here: https://youtube.com/playlist?list=PLgSpqOFj1Ta7bnCNBAlWcN76UbnLthyO1&si=T05d-vYQdIKpfN7r
58
56How we can access to the recording of this class and others?There will be a blog post on the Birds of the World webinar website with a link to the video recording.
59
57are there refs to the changes in your modeling from package year to package year? I;m thinking the last ref is Fink 2019No, we currently do not have anything newer than this paper: https://esajournals.onlinelibrary.wiley.com/doi/full/10.1002/eap.2056 , which the supplementary material describes the methods thoroughly.

Part of our change in release cadence is to publish a paper on the latest methods.
60
58are there refs to the changes in your modeling from package year to package year? I;m thinking the last ref is Fink 2019
61
59How do you use your access key online for downloading GIS data like Tom did?Go here and fill out a data access request form: https://science.ebird.org/en/status-and-trends/download-data
62
60How do you use your access key online for downloading GIS data like Tom did?I have done that already, but the downloads page still just says "key required". where does one enter the key you get so that you can actually download the gpkg files...
63
61The abundance estimates are modeled in weekly base, but the trends is annual change from `trends_start_year` to `trends_end_year`; how these two temporal resolutions are connected? As much as I understand the double machine learning algorithm uses a discrete exponential growth model as underlying population of the state process, but I don’t understand if the annual abundance is an average across weeks? or how is it defined?The weeks are used to define the season of data that is used in the Trends analysis, so for a breeding season we use all checklists from the start of the breeding season and the end of the breeding season. Then, in the Trends analysis, one of the models, the marginal model, estimates average relative abundance for the start and end years of the trends time period.
64
62The abundance estimates are modeled in weekly base, but the trends is annual change from `trends_start_year` to `trends_end_year`; how these two temporal resolutions are connected? As much as I understand the double machine learning algorithm uses a discrete exponential growth model as underlying population of the state process, but I don’t understand if the annual abundance is an average across weeks? or how is it defined?
65
63Do i need github? I have the acess key, but no github (i have no idea what it is).You do not need github, it’s just the source of the repository.
66
64How do you set your access key in the R session that Matt talked about (it was not in the zoom screen)Please see the access key section in the intro here: https://ebird.github.io/ebirdst/
67
65How do you set your access key in the R session that Matt talked about (it was not in the zoom screen)using the funcion set_ebirdst_access_key(), right?
68
66How do you set your access key in the R session that Matt talked about (it was not in the zoom screen)Or if that fails as it does for me, set the env variable EBIRDST_KEY=“XXXXXX” before starting r
69
67How do you set your access key in the R session that Matt talked about (it was not in the zoom screen)Correct.
70
68Sorry iF i'm asking too many questions, but where do I type this acess key?Use the set_ebirdst_access_key("XXXXX”), see here: https://ebird.github.io/ebirdst/
71
69Sorry iF i'm asking too many questions, but where do I type this acess key?Sorry, I meant the Function.
72
70What determines which birds are available through the r package?All the species which are available on the website are also available through the R package. If you run into issues trying to access a certain species, note that a small subset of species might not be part of the most recent package update because they we're part of the most recent production run. In these cases, the data can be accessed either directly from the website or from an older version of the R package (the package version will match the "prediction year" of the dataset).
73
71are we downloading rasters like when you downlaod from the website?Yes, that’s the primary format that people download through the R package. But the range boundaries as vector data are available as well.
74
72Yea Matt for using vierer friendly settings in Rstudio so we can read your code! Well done!
75
73is there a way to get a key now, so as to get a dataset to follow along?Absolutely! Get your key here: https://science.ebird.org/en/status-and-trends/download-data
76
74los subtitulo pueden ser en españolPuedes hacer eso en tu Zoom?
77
75los subtitulo pueden ser en españolen la flechita hacia arriba en el icono de CC esta la opcion del idioma
78
76Is it possible to obtain previous years of Status & Trends data?Last year's data estimates are available from the previous version of the R package (for example, if a species isn't available from the most recent R package version because it wasn't part of the most recent run -- this is a small subset of species).

However, note that previous years of relative abudnance estimates are not directly comparable to the most recent year if you're working with the same species. For example, the way that these data products are estimated means that you can't look at year 1 and year 2 and infer population change. Becuase the models are constantly updated and improved, you can't compare between years. (note that, down the road, we are working to accomplish this using eBird trends).
79
77
80
78Can we set the ebirdst data directory with the ebirdst_data_dir() function or a different function?same here, the route in the function is not on my computer
81
79Can we set the ebirdst data directory with the ebirdst_data_dir() function or a different function?ebirdst_data_dir() lets you find out what the default directory is. If you want to set it differently you need to provide the path you want to the path parameter when you call ebirdst_download_status()

If you’ve already downloaded to another path, you can use that path parameter in load_raster() for example.
82
80Where do you find the polygons limits like Wyoming used in the script? Would it work if we were to filter with a province/state?download the rnaturalearth package
83
81Where do you find the polygons limits like Wyoming used in the script? Would it work if we were to filter with a province/state?Yes, the polygon boundaries we're using in the script come from the rnaturalearth R package. You can use any shapefile to crop and mask the dataset to your shapefile boundary.
84
82How would alititudinal migrants or species exhibiting chain migration (where different populations are found in the same location throughout the year) handled when the coding defines them as resident vs migrant, but the process is more complex than True/False?Classification as migrant doesn’t mean that weekly patterns aren’t captured by the model, often they are, but they’re subtle and we’re choosing to summarize as a year-round representation. However, if experts reviewed the weekly information and passed the animation, then that weekly information is available for download in R as well. However, subtle and/or irruptive movements are less likely to be captured.
85
83Wyoming boundary script has issues for me:
Error in h(simpleError(msg, call)) :
error in evaluating the argument 'x' in selecting a method for function 'vect': no applicable method for 'filter' applied to an object of class "c('SpatialPolygonsDataFrame', 'SpatialPolygons', 'Spatial')"
86
84what if you want multiple stateslive answered
87
85For the next session, would it be possible to send the code beforehand so we can update/install packages beforehand.There won’t be additional packages before the next session, but we will send out the code ahead of time.
88
86Is this going to end at 1:30 or 2 pm ET?The prepared content will end at 1:30, but we will have time for questions until 2.
89
87will you provide the code from this session?Yes, we've posted the code in the chat: https://raw.githubusercontent.com/ebird/ebirdst/refs/heads/main/examples/ebirdst-bow-webinars.R
90
88Could you also get % of the relative abundance for a protected area vs let’s say the state level?Yes, if you have spatial data about the boundaries of the protected area, it works the same.
91
89When applying the func= sum at the local scale you are just adding the abundance values of the selected region, right? could you walk us on teh interpretation of the values ?live answered
92
90If we want to bulk downloard brids from the website (old version is that possible)You can install a previous version of the R package and download in bulk that way. For example, “remotes::install_version("ebirdst", version = "2.2021.3”)”
93
91just a small question how long is our sessionWe will go until 1:30 ET (6 more minutes), but we will stay live for Q&A until 2 PM (ET).
94
92can we calculated for a protected area where the ebird records are less in some cases. what can be done to rectify it.The eBird relative abundance are predictions from an ensemble of models (not raw data), so the workflow is able to generate predictions even in regions where the raw data are sparse.
95
93can we calculated for a protected area where the ebird records are less in some cases. what can be done to rectify it.Thank you. so we can compare protected areas and unprotected areas?
96
94Based on the fact that all datapoints are human species observation, thus where there is better human coverage the data is more representative would you say seabirds (especially those that are high-seas specialist) are significantly more data poor?Seabirds over the high seas are data poor, definitely. Data near the coast I think is reliable, if we’re showing it. In the upcoming version, we’ve added 5 covariates over the ocean (bathymetry, bathymetric slope, monthly chlorophyll, monthly sea surface temperature, and monthly sea surface temperature anomaly) and we’re hoping to release more data over open water, although it will be far from complete for most species.
97
95Based on the fact that all datapoints are human species observation, thus where there is better human coverage the data is more representative would you say seabirds (especially those that are high-seas specialist) are significantly more data poor?Thanks for the information. Are you interested in data sources that are not observations? I work for the Department of Conservation in NZ and we sit on an abundance of satellite tracking data for many species of Albatross and Petrels :)
98
96Based on the fact that all datapoints are human species observation, thus where there is better human coverage the data is more representative would you say seabirds (especially those that are high-seas specialist) are significantly more data poor?Data integration is an exciting opportunity and challenge but is very much at a research phase at this point, I don’t have a direct lead right now.
99
97Where do I input the Function?Would you clarify this question? Which function?
If you would like to follow along with Matt's code, please find it here: https://raw.githubusercontent.com/ebird/ebirdst/refs/heads/main/examples/ebirdst-bow-webinars.R
100
98Where do I input the Function?I meant the set_ebirdst_access_key("XXXXX") Function.