SMAST Seminar - DFO - September 25, 2024 - "Spatially Balanced Sampling and Other Heresies" By: Paul Rago
When: Wednesday,
September 25, 2024
3:00 PM
-
4:00 PM
Where: > See description for location
Description: Department of Fisheries Oceanography
"Spatially Balanced Sampling and Other Heresies "
Paul Rago, Retired, formerly Northeast Fisheries Science Center (NEFSC)
Wednesday, September 25, 2024
3:00 - 4:00 pm
SMAST E 101-102 and via Zoom
Abstract:
Stratified random sampling is widely used in natural resource surveys around the world. It offers numerous theoretical and practical benefits, particularly when strata are well designed and sampling allocation is proportional to some measure of abundance. While simple random sampling within strata is unbiased in expectation, it can be inefficient if a realized sample allocation yields stations too close together, and perceived as problematic if known “hot spots†don’t show up in the sample. Systematic sampling is often used to address these problems but estimation of variances for such designs is still being debated in the literature. Spatially-balanced sampling is an alternative that retains the desirable attributes of random sampling while achieving an appropriate degree of spatial coverage and distance between samples. A particularly useful version of spatially-balanced sampling is known as Generalized Random Tessellation Stratified (GRTS) sampling. GRTS is a numerically intensive method that generates random samples from a spatially-distributed population based on a recursive hierarchical method for ordering the potential sampling sites. Estimates of means and totals are based on inclusion probabilities for each station using the classic Horvitz-Thompson methods. Importantly, inclusion probabilities can vary among stations in response to prior information about relative abundance (e.g., hot spots). Thus, GRTS can “flex†station locations within strata and guide station allocation among strata.
In this seminar I address the heretical nature of unequal inclusion probabilities and how such designs can be used to improve surveys for Atlantic sea scallops. Existing survey strata are based on depth and geographic criteria that are independent of the spatial fishing areas now used in management. This mismatch complicates the estimation of scallop biomass. Simulation studies are used to demonstrate the advantages of using prior information for reducing variance of estimates and responding to trends in abundance. Reliance on prior information also creates risks of biased estimation if the prior information is wrong. I illustrate several ways to reduce the risks of biased estimates. Recent declines in the Mid-Atlantic suggests that environmentally-driven mortality may play an increasingly important role in the future. Future surveys need to be both responsive to such changes and robust to misspecification of predicted abundance.
************************************************************
Join Zoom Meeting
https://umassd.zoom.us/j/93758230260?pwd=OHJ5UDloQkZZaCtXcTlBNlR6Qm0rQT09
Meeting ID: 937 5823 0260
Passcode: 426839
************************************************************
For additional information, please contact Callie Rumbut at c.rumbut@umassd.edu
"Spatially Balanced Sampling and Other Heresies "
Paul Rago, Retired, formerly Northeast Fisheries Science Center (NEFSC)
Wednesday, September 25, 2024
3:00 - 4:00 pm
SMAST E 101-102 and via Zoom
Abstract:
Stratified random sampling is widely used in natural resource surveys around the world. It offers numerous theoretical and practical benefits, particularly when strata are well designed and sampling allocation is proportional to some measure of abundance. While simple random sampling within strata is unbiased in expectation, it can be inefficient if a realized sample allocation yields stations too close together, and perceived as problematic if known “hot spots†don’t show up in the sample. Systematic sampling is often used to address these problems but estimation of variances for such designs is still being debated in the literature. Spatially-balanced sampling is an alternative that retains the desirable attributes of random sampling while achieving an appropriate degree of spatial coverage and distance between samples. A particularly useful version of spatially-balanced sampling is known as Generalized Random Tessellation Stratified (GRTS) sampling. GRTS is a numerically intensive method that generates random samples from a spatially-distributed population based on a recursive hierarchical method for ordering the potential sampling sites. Estimates of means and totals are based on inclusion probabilities for each station using the classic Horvitz-Thompson methods. Importantly, inclusion probabilities can vary among stations in response to prior information about relative abundance (e.g., hot spots). Thus, GRTS can “flex†station locations within strata and guide station allocation among strata.
In this seminar I address the heretical nature of unequal inclusion probabilities and how such designs can be used to improve surveys for Atlantic sea scallops. Existing survey strata are based on depth and geographic criteria that are independent of the spatial fishing areas now used in management. This mismatch complicates the estimation of scallop biomass. Simulation studies are used to demonstrate the advantages of using prior information for reducing variance of estimates and responding to trends in abundance. Reliance on prior information also creates risks of biased estimation if the prior information is wrong. I illustrate several ways to reduce the risks of biased estimates. Recent declines in the Mid-Atlantic suggests that environmentally-driven mortality may play an increasingly important role in the future. Future surveys need to be both responsive to such changes and robust to misspecification of predicted abundance.
************************************************************
Join Zoom Meeting
https://umassd.zoom.us/j/93758230260?pwd=OHJ5UDloQkZZaCtXcTlBNlR6Qm0rQT09
Meeting ID: 937 5823 0260
Passcode: 426839
************************************************************
For additional information, please contact Callie Rumbut at c.rumbut@umassd.edu
Contact: > See Description for contact information
Topical Areas: Faculty, SMAST, Students, Graduate, Civil and Environmental Engineering, College of Engineering, Lectures and Seminars