Examining inter-regional and intra-seasonal differences in wintering waterfowl landscape associations among Pacific and Atlantic flyways
Data files
Dec 05, 2024 version files 4.80 GB
-
KBBX_screening_MJH.xlsx
49.82 KB
-
KBBX_summary_Winter.csv
1.05 GB
-
KDAX_screening_MJH.xlsx
48.25 KB
-
KDAX_summary_Winter.csv
1.64 GB
-
KDOX_screening_MJH.xlsx
48.14 KB
-
KDOX_summary_Winter.csv
296.74 MB
-
KHNX_screening_MJH.xlsx
44.23 KB
-
KHNX_summary_Winter.csv
305.16 MB
-
KLTX_screening_MJH.xlsx
46.37 KB
-
KLTX_summary_Winter.csv
355.25 MB
-
KMHX_screening_MJH.xlsx
47.05 KB
-
KMHX_summary_Winter.csv
456.78 MB
-
Modelling_Grids.zip
207.15 MB
-
Predictions_CVC.zip
284.61 MB
-
Predictions_MA.zip
198.47 MB
-
README.md
8.85 KB
Abstract
The Central Valley of California (CVC) and Mid-Atlantic (MA) in the U.S. are both critical sites for nationwide food security (California Poultry Federation 2016, Prosser et al. 2017), and many waterfowl species annually, especially during the winter, providing feeding and roosting locations for a variety of species. Mapping waterfowl distributions, using NEXRAD, may aid in the adaptive management of important waterfowl habitat and allow various government agencies to better understand the interface between wild and domestic birds and commercial agricultural practices. We used 9 years (2014–2023) of data from the US NEXRAD network to model winter waterfowl relative abundance in the CVC and MA as a function of weather, temporal period, environmental conditions, and landcover characteristics using Boosted Regression Tree modelling. We were able to quantify the variability in effect size of 28 different covariates across space and time within two geographic regions which are critical to nationwide waterfowl management and host a high density of nationally important commercial agriculture. In general, weather, geographic (distance to features), and landcover condition (wetness index) predictors had the strongest relative effect on predicting wintering waterfowl relative abundance in both regions, while effects of land cover composition were more regionally and temporally specific. Increased daily mean temperature was a major predictor of increasing relative waterfowl abundance in both regions throughout the winter. Increasing precipitation had differing effects within regions, increasing relative waterfowl abundance in the MA, while decreasing in general within the CVC. Increasing relative waterfowl abundance in the CVC is strongly tied to the flooding of the landscape and rice availability, whereas waterfowl in the MA, where water is less limiting, are generally governed by waste grain availability and emergent wetland on the landscape. Waterfowl relative abundance in the MA was generally higher nearer to the Atlantic coast and lakes, while in the CVC they were higher nearer to lakes. Our findings promote a better understanding of spatial associations of waterfowl to landscape features and may aid in conservation and biosecurity management protocols.
README: Examining inter-regional and intra-seasonal differences in wintering waterfowl landscape associations among Pacific and Atlantic flyways
List of Associated Documents and Descriptions of Files
(1) Screening Spreadsheets
.csv files of days of the winter November 1st to March 15th 2014-2022 for each of the 6 radars (KDAX, KBBX, KHNX, KDOX, KMHX, and KLTX).
File and variables included:
- RADAR = 4 Digit Radar Site Code
- DATE = Day of the Year in Month/Day/Year format
- SEASON = 'Winter' to distinguish from other season in projects dealing with songbirds.
- DOWNLOAD = Purposely left blank, it's filled in with a Yes or No by the data handler depending on the STATUS column
- STATUS = Classed as biological (B), contaminated (C), or No Data/missing (ND) when visually screened using NOAA's Weather Climate Toolkit. Biological days were used for modelling
- CONTAMINATION_TYPE = If STATUS = B, then 'blank', if STATUS = C OR ND, CONTAMINATION_TYPE is listed. P = Precipitation, CL = Clutter, and AP = Anomalous Propagation of the radar beam. ND STATUS is also left 'blank'.
- TARGET_ID = Intentionally left blank
- SCREENER = Initials of researcher who visually screened raw radar data.
- SURFACE_WIND, WIND_DIRECTION, APPROXIMATE_SAMPLING_TIME, TARGET_SPEED, GROUND_HEADING, and COMMENTS intentionally left blank and can be used for data annotation by the person screening the weather radar data. We use wind direction and target speeds to distinguish between songbirds and insects, but given this was a waterfowl study, and in winter, we do not have to distinguish between insects. This is our standard template and is why numerous columns are blank.
(2) Modelling Grids and Summary-Winter
I provide modelling grids and all their properties used for the 6 radar domains. These can be read in GIS or other software such as R. The modelling grids are a 250m resolution raster of the unique identifier values for each grid cell that match up with the “id” variable in summary_winter files of data tables. Use this raster for visualizing data as a spatial map.
File includes:
- id = Cell ID in radar modelling grids
- range = The range the radar beam is scanning from the centroid of the radar (meters)
- azimuth = The angle the radar beam is scanning at, ranging from 0.25 to
- groundht = Refers to the ground height (meters) or the elevation of the ground surface at a specific location relative to sea level. This value is used to calculate the actual height of radar targets
- pblock = Quantifies how much of the radar beam's energy is reduced or blocked due to obstructions. Expressed as a percentage.
- vprfilter = Refers to adjustments or filters applied to account for the Vertical Profile of Reflectivity (VPR). The VPR describes how radar reflectivity changes with altitude and is used to better interpret and process radar data.
- clutter = Was a manual clutter filter applied (binary, 1 = yes, 0 = no)
- pwater = Quantifies how much of the reflectivity is over water. Expressed as a percentage.
- waterfilt = Mask of reflectivity values over water. Was a manual water filter applied (binary, 1 = yes, 0 = no)
- utm_e = Easting
- utm_n = Northing
- elev = elevation above sea level (meters) of the radar itself
- hectares = Measure of area (hectares) of land that can be used for analyzing or interpreting radar data.
- awater = Measure of area (hectares) of water that can't be used for analyzing or interpreting radar data.
- lcv11 = How many 30 x 30m grid cells of Open Water
- lcv12 = How many 30 x 30m grid cells of Ice/Snow
- lcv20 = How many 30 x 30m grid cells of Grain
- lcv21 = How many 30 x 30m grid cells of Developed-Open Space
- lcv22 = How many 30 x 30m grid cells of Developed- Medium
- lcv23 = How many 30 x 30m grid cells of Developed- High
- lcv24 = How many 30 x 30m grid cells of Developed- Very High
- lcv25 = How many 30 x 30m grid cells of Developed-Other
- lcv30 = How many 30 x 30m grid cells of Barren
- lcv31 = How many 30 x 30m grid cells of Barren
- lcv41 = How many 30 x 30m grid cells of Forest-Deciduous
- lcv42 = How many 30 x 30m grid cells of Forest-Coniferous
- lcv43 = How many 30 x 30m grid cells of Forest-Mixed
- lcv44 = How many 30 x 30m grid cells of Forest-Other
- lcv52 = How many 30 x 30m grid cells of Scrub/Shrub
- lcv71 = How many 30 x 30m grid cells of Grassland
- lcv81 = How many 30 x 30m grid cells of Pasture/Hay
- lcv82 = How many 30 x 30m grid cells of Cultivated Crops
- lcv84 = How many 30 x 30m grid cells of Sod
- lcv90 = How many 30 x 30m grid cells of Woody/Forested Wetland
- lcv91 = How many 30 x 30m grid cells of Unidentified
- lcv92 = How many 30 x 30m grid cells of Unidentified
- lcv93 = How many 30 x 30m grid cells of Unidentified
- lcv94 = How many 30 x 30m grid cells of Unidentified
- lcv95 = How many 30 x 30m grid cells of Emergent Herbaceous Wetland
- lcvsum = sum of 30x30 m landcover gridcells for that specific id
NOTES:
Values of lcv01v95 and similar denote proportions of that landcover type that makes up that 250m x 250m grid, values of z### are the rraw reflectivity values of aerial targets measured in dbz.
.csv file continued......
- long = Longitude of grid cell
- lat = Latitude of grid cell
- dcoast = Distance to coast (kilometers) from radar centroid
- mbeam = Height of the middle of the radar beam (meters)
- bbeam = Height of the bottom of the radar beam (meters)
- tbbeam = Height of the top of the radar beam (meters)
- corrz = How well radar signals correlate across the radar volume scan. This variable is important because it helps identify areas where radar returns may be unreliable due to issues such as beam blocking, ground clutter, or other distortions. The higher the correlation, the more reliable the data is likely to be.
- propdist = refers to the propagation distance or distance to the radar beam at a particular point in time. It is essentially the distance from the radar site to the location where the radar signal is being reflected or detected. Expressed as a percentage. The decreasing amount of area covered as a function of increasing beam height
Missing Data notes
Unless otherwise stated above, 0 values are true zeroes. Missing values indicate data were not available/corrupt/missing at that grid or interval.
(3) Predictions
Description: The Mean Vertically Integrated Reflectivity (VIR) of waterfowl refers to a radar-derived measurement of the total reflectivity of waterfowl targets (like flocks of birds) along the radar's vertical scan. It is an estimate of the radar reflectivity accumulated over the entire column of air from the ground up to the maximum height scanned by the radar.
Items:
.rds files of predicted mean vertically integrated reflectivity (VIR) for each radar domain.
Content Description: There are two prediction folders
Predictions_CVC.zip = Mean VIR predictions for the Central Valley of California region (Radars KDAX, KBBX, and KHNX)
Predictions_MA.zip = Mean VIR predictions for the Mid-Atlantic region (Radars KDOX, KLTX, and KMHX)
Zip File Contents
We created mean VIR for every half-month period through 2020-2022 beginning with Early and late November (eNov, and lNov) and ending in March (eMar) for each year and each domain (MA or CVC).
eNov, lNov, eDec, lDec, eJan, lJan, eFeb, lFeb, and eMar
(4) Accessing the Archive Data ︎or Getting Data from Other Radars
https://aws.amazon.com/public-datasets/nexrad/
The full historical archive from NOAA from June 1991 to present is available.
The NEXRAD Level II archive data is hosted in the noaa-nexrad-level2 Amazon S3 bucket in the us-east-1 AWS region. The address for the public bucket is: https://noaa-nexrad-level2.s3.amazonaws.com.
Each volume scan file of archival data is available as an object in Amazon S3. The basic data format is:
/
Where:
All files in the archive use the same compressed format (.gz). The data file names are, for example, KDAX20010101_080138.gz. The file naming convention is:
GGGGYYYYMMDD_TTTTTT
Where:
GGGG = Ground station ID (map of ground stations)
YYYY = year
MM = month
DD = day
TTTTTT = time when data started to be collected (GMT)
Note that the 2015 files have an additional field on the file name. It adds “_V06” to the end of the file name. An example is KDAX20150303_001050_V06.gz.
Methods
We analyzed NEXRAD data from November 1 to March 15 (2014–2023) to estimate waterfowl densities in the Central Valley of California and the Mid-Atlantic. Radar reflectivity, measured in units of Z (mm⁶ m⁻³), was used to quantify waterfowl biomass by assessing the density and size of targets.
To capture waterfowl departing feeding grounds, we analyzed radar scans during evening civil twilight when detection was highest. Data were collected from the lowest elevation sweeps (0.5°) with volume dimensions of 250 m x 0.5°. Scans were manually screened to exclude precipitation and other non-bird sources, using NOAA’s Weather and Climate Toolkit and Amazon Web Services for radar data.
Reflectivity in dBZ was converted to total bird reflectivity (η) [cm² km⁻³] and grouped into 10 m altitude bins to create a vertical profile of reflectivity (VPR). The VPR, derived from the five lowest sweeps (0.5°–5.5° tilt) within 5–20 km of the radar, was adjusted for range bias to calculate vertically integrated reflectivity (VIR) [cm² ha⁻¹], representing the average waterfowl biomass across the air column.