top of page

SCRAM 2: Frequently Asked Questions

If you have suggestions for additional topics to include on this page, please email kate.williams@briwildlife.org.  

Wow, SCRAM is working a lot faster now, what happened?

We integrated SCRAM with another offshore wind collision risk decision support tool from the United Kingdom (stochLAB; Caneco et al. 2022) that also uses the Band (2012) model as a basis for its estimates. The developers of stochLAB significantly improved the speed of the model runs and added QA/QC protocols as compared to the older model code on which SCRAM 1.0.3 was based. This change reduces the possibility of future errors and allows users to test multiple scenarios much faster.

In the web app, I’m seeing Option 2 and Option 3 listed for estimated collision risk, when before there was Option 1 and Option 3. Why the change?

The options are calculated almost exactly the same as before, but we have renamed Option 1 as Option 2 to make sure that SCRAM nomenclature is aligned with Band (2012) and stochLAB (Caneco et al. 2022). In Band and stochLAB, what is called Option 1 requires location-specific flight height data, while Option 2 uses general flight height data for the species. Since the latter case is what was originally called “Option 1” in SCRAM, we renamed this option to align with CRM nomenclature. Additionally, we use very slightly different avoidance rate estimates for Option 2 (formerly Option 1) than were used in SCRAM 1.0.3; before, a single set of avoidance values per species were used for both collision risk estimation options. Values used in both Option 2 and Option 3 are taken from Table A2 in Cook (2021). However, as stated in Adams et al. (2022), we generally recommend the use of Option 3 whenever possible, which explicitly pairs the altitudes of bird activity with collision risk in the rotor swept zone. If birds tend to fly in the lower portion rotor swept zone where the chances of collision are lower, then their chances of collision should be estimated as such. We account for this distribution profile in Option 3, but not Option 2. In Option 2, we only look at overall use of the rotor swept zone and the average collision risk of the whole area.

Why is the number of predicted collisions different between SCRAM 1.0.3 and 2?

Many changes were made to improve SCRAM between versions 1.0.3 and 2. First, the transition to the stochLAB code for the collision risk model helped us squash a few lingering bugs. Some of these changes seem to have generally decreased SCRAM 2 collision risk estimates relative to SCRAM 1.0.3. Second, we updated the movement models for all species; this included adding newer Motus data, implementing more stringent QA/QC procedures to filter out potentially erroneous detections in the tracking datasets for all three study species, after Loring et al. (2021), and eliminating the use of multiple movement states. We also incorporated data from multiple new transmitter types (GPS and PTT) for red knots. After these improvements, there were noticeable differences in the estimates of space use in the animals, though the general patterns were similar. In general, the highest and lowest occupancy values were smoothed out across the study area in SCRAM 2 as compared to SCRAM 1.0.3. Lastly, the regional population size estimates were updated on both an overall and monthly basis, which tended to decrease estimated collision risk. The net change in collision risk varied by species and location. For more information, please see the BOEM report on SCRAM 2 (https://espis.boem.gov/Final%20Reports/BOEM_2024-057.pdf).

What movement data are now included in SCRAM?

SCRAM 2 movement models are based on data from automated radiotelemetry, or “Motus” tags (Piping plover n=107, roseate tern n=134, red knot n=240), GPS transmitters (red knot n=81), and PTT transmitters (red knot n=25). Motus tracking technology involves the deployment of receiving stations that record detections of tags that come into range of the receiving radio antennas (Taylor et al. 2017), unlike GPS and PTT tags, which typically transmit data via satellite. The Motus data are all from tracking studies conducted in 2015-2017 that were funded by the Bureau of Ocean Energy Management (BOEM; Loring et al. 2018; Loring et al. 2019; Loring et al. 2021). GPS and PTT data are from six different studies conducted in 2020-2023, as discussed in the main text and in Appendix C of the SCRAM 2 report (https://espis.boem.gov/Final%20Reports/BOEM_2024-057.pdf). SCRAM 1.0.3 included only Motus data, including a much smaller Motus dataset for red knots than is included in SCRAM 2. Due to the coastal locations of Motus stations in 2015-2017, Motus detections in the dataset tend to be somewhat clustered towards the coastline, and there is substantial uncertainty in occupancy estimates for grid cells located well offshore (despite substantial improvements in movement models in SCRAM 2, which reduced this uncertainty from SCRAM 1.0.3). For red knots, GPS and PTT transmitters incorporated into SCRAM 2 provide location data for both coastal and offshore regions with substantially higher precision than Motus technology, improving resulting occupancy models. Flight height models in SCRAM 1.0.3 were based solely on Motus data (Loring et al. 2018; Loring et al. 2019). For SCRAM 2, red knot flight height models are now based on GPS tracking data (n=132 individuals). Future updates to SCRAM are expected to continue to incorporate additional tracking datasets (as they become available) into both movement models and flight height models.

It sounds like there were a lot of changes to the Motus movement models, what happened there?

It started with an update to the red knot tracking dataset, to incorporate newer Motus data into SCRAM 2. The movement model didn’t fit the data properly, so we simplified the model to improve convergence. For consistency, we applied a similar approach and improvements to the piping plover and roseate tern datasets. The greatest change simplified the models from two-state behavioral models to one-state models. Previously we adjusted the collision estimates to only include occupancy from birds in a migratory movement state. However, following changes to the underlying red knot dataset that added data and improved QA/QC (e.g., data filtering procedures), the model could no longer reliably differentiate two movement states. This means that collision risk estimates for SCRAM 2 may be higher in some nearshore grid cells and lower in offshore grid cells as compared to SCRAM 1.0.3. Future model development is expected to revisit a two-state model, as new tracking data become available.

Why did you add the Band Annex 6 models to the SCRAM web application?

We added functionality to run a stochastic version of the Band (2012) model (specifically Band Annex 6, the collision risk model for migrants) in the SCRAM web application. Band (2012) Annex 6 uses nearly the same underlying collision risk model as SCRAM, but estimates “flux,” or the number of animals that are present to potentially collide with a turbine, based on range-wide estimates of migratory corridor width. Essentially, the Band migratory model (i.e., Annex 6) assumes animals are evenly distributed across the width of the user-designated migratory corridor. In contrast, SCRAM’s movement modeling approach estimates the occupancy to be higher in places that more animals moved through, and lower where fewer animals occurred (based on individual tracking data). A random distribution of animals, as in the Band migrant model, isn’t often a reasonable assumption, but it can be helpful for making approximate collision risk estimates where other data are lacking. By allowing users to contrast the two models, it can more easily be seen how certain assumptions play out in the collision risk estimates, and will better allow those differences to inform conservation decision making. Including the Band (2012) Annex 6 collision risk model in the tool will also simplify the process of running both models for the same wind energy facilities, ensure consistent reporting of data inputs and outputs between the two types of models, and facilitate the use of a stochastic, rather than deterministic, version of the Band model.

Have the models been peer-reviewed?

Yes! In addition to reviews by experts at BOEM and the U.S. Fish and Wildlife Service (USFWS), both SCRAM 1 and SCRAM 2 reports, web applications, user manual, and code have gone through external peer-review processes by the U.S Geological Survey (USGS). All the underlying code is also fully available on GitHub for anyone to review and use (https://github.com/Biodiversity-ResearchInstitute/SCRAM2).

Anything else we should know about?

Yep, there are a few other noteworthy changes in SCRAM 2, including: 1. We are now limiting SCRAM’s Motus model output to a smaller geographic area where there was consistent Motus station coverage during the tracking studies that make up our movement dataset. This area ranges approximately from northern Cape Cod to the Virginia-North Carolina border. If you ask SCRAM for a movement model-based collision risk estimate in the NES outside of that geographic area, you will only get one for red knots, and it will be 100% based on satellite tracking data (not Motus tags). The Band Annex 6 model will also provide collision risk predictions throughout the NES study area. We hope to expand the geographic area to which we can make confident predictions using Motus data in future iterations of SCRAM, but that will require new data and better coverage. 2. SCRAM produces monthly collision risk estimates for a slightly different subset of month/species combinations than before. This is because we 1) added new Motus data and GPS/PTT data to our red knot models, 2) incorporated new Motus data for red knots and implemented additional quality control filtering to remove potentially erroneous data points for all three species, and 3) removed occupancy and collision estimates for piping plovers in September, when only 4 individuals were detected (below the sample size threshold of 5). 3. We updated the monthly regional population size estimates of birds thought to be present in the NES study area that could be available to collide with turbines. These updates were based on the best available science from the USFWS and aligned the numbers in SCRAM 2 with those being used in the most recent Band (2012) model runs. In general, these updates tended to decrease monthly regional population size estimates, as USFWS refined their estimates of when animals were moving through the NES during fall migration. 4. We updated the avoidance rates for all three species to average the values recommended by Cook (2021) and Ozsanlav-Harris et al. (2023). These two reports used similar (but not identical) datasets from Europe to estimate avoidance rates of offshore wind turbines by gulls and terns, and developed differing estimates of avoidance rates. Upon review, neither analysis seems like it was a clear improvement over the other, and thus we chose to average the avoidance rates from the two reports (SCRAM 1 avoidance rates were based solely on values from the Cook report, and were lower than the averaged values from the two reports included in SCRAM 2).

Where do I find more information about SCRAM?

All documentation on SCRAM is available at https://www.collisionrisk.org/scram. For SCRAM 2, this includes the BOEM report, technical summary, web application, user manual, and model code.

FAQ References:

Adams E, A Gilbert, P Loring, KA Williams. 2022. Transparent Modeling of Collision Risk for Three Federally Listed Bird Species in Relation to Offshore Wind Energy Development: Final Report. Washington, DC: U.S. Department of the Interior, Bureau of Ocean Energy Management Report No.: OCS Study BOEM 2022-071. 

 

Band B. 2012. Using a collision risk model to assess bird collision risks for offshore windfarms. The Crown Estate as part of the Strategic Ornithological Support Services Programme, SOSS-02. 62 pp. 

 

Caneco B, Humphries G, Cook AS, Masden E. 2022. Estimating bird collisions at offshore windfarms with stochLAB. https://hidef-aerial-surveying.github.io/stochLAB/. 

 

Cook ASCP. 2021. Additional analysis to inform SNCB recommendations regarding collision risk modelling. BTO Research Report. 739:48. 

 

Loring P, Lenske A, McLaren J, Aikens M, Anderson A, Aubrey Y, Dalton E, Dey A, Friis C, Hamilton D, et al. 2021. Tracking Movements of Migratory Shorebirds in the US Atlantic Outer Continental Shelf Region. Sterling (VA): US Department of the Interior, Bureau of Ocean Energy Management. OCS Study BOEM 2021-008. 104 p. 

 

Loring P, Paton P, McLaren J, Bai H, Janaswamy R, Goyert H, Griffin C, Sievert P. 2019. Tracking Offshore Occurrence of Common Terns, Endangered Roseate Terns, and Threatened Piping Plovers with VHF Arrays. Sterling, Virginia: US Department of the Interior, Bureau of Ocean Energy Management. 

 

Loring PH, McLaren JD, Smith PA, Niles LJ, Koch SL, Goyert HF, Bai H. 2018. Tracking Movements of Threatened Migratory rufa Red Knots in U.S. Atlantic Outer Continental Shelf Waters. OCS Study BOEM 2018-046. U.S. Department of the Interior, Bureau of Ocean Energy Management, Sterling, VA. 145 pp. 

 

Ozsanlav-Harris L, Inger R, Sherley R. 2023. Review of data used to calculate avoidance rates for collision risk modelling of seabirds. Joint Nature Conservation Committee. JNCC Report 732:60. 

 

Taylor PD, Crewe TL, Mackenzie SA, Lepage D, Aubry Y, Crysler Z, Francis CM, Guglielmo CG, Hamilton DJ, Holberton RL, et al. 2017. The Motus Wildlife Tracking System: A collaborative research network to enhance the understanding of wildlife movement. Avian Conservation and Ecology. 12(1). doi:10.5751/ACE-00953-120108.

Collision Risk Models for Birds and Wind Energy Development

This website was funded with support from the U.S. Fish and Wildlife Service under Cooperative Agreement F24AC01956.

JOIN OUR EMAIL LIST

bottom of page