Buffer Size and Region.N
Posted: Mon Sep 14, 2015 3:50 pmHello,
I am looking to compare SECR-based estimates of population size to those estimated in program MARK. I am largely interested in qualitatively comparing and understanding the different processes that create the differences, as well as using the comparison as a double check of my work.
I undertand that the estimation of density can be influenced when the buffer is too small, so buffer only has to be large enough to encompass all individuals that use the area of interest. However, I am less clear on the effects of buffer size for region.N, as the buffer appears to have large effects on this parameter. I conducted the following comparison to highlight the modest differences in density, but major differences in N with changing buffer sizes.
My gut assumption would be that the most accurate buffer width for calculating N would be the buffer at which density estimates asymptote?, thus movements of all individuals are accounted for while not adding extra area? It appears to me density asymptotes at some buffer width, while N increases continually. Which may make sense if N is estimated as the derived density*area.
I am looking to compare SECR-based estimates of population size to those estimated in program MARK. I am largely interested in qualitatively comparing and understanding the different processes that create the differences, as well as using the comparison as a double check of my work.
I undertand that the estimation of density can be influenced when the buffer is too small, so buffer only has to be large enough to encompass all individuals that use the area of interest. However, I am less clear on the effects of buffer size for region.N, as the buffer appears to have large effects on this parameter. I conducted the following comparison to highlight the modest differences in density, but major differences in N with changing buffer sizes.
My gut assumption would be that the most accurate buffer width for calculating N would be the buffer at which density estimates asymptote?, thus movements of all individuals are accounted for while not adding extra area? It appears to me density asymptotes at some buffer width, while N increases continually. Which may make sense if N is estimated as the derived density*area.
- Code: Select all
#####With large Buffer
##Model 1
SN.bear.fit1 <- secr.fit(SN.ALL,model=list(D~1, g0~1), buffer = 4 * RPSV(SN.ALL)$SN2010)
SN.bear.fit1
link estimate SE.estimate lcl ucl
D log 4.731945e-05 1.082982e-05 3.038853e-05 7.368342e-05
g0 logit 2.313856e-01 3.736411e-02 1.662727e-01 3.124413e-01
sigma log 8.994555e+03 1.128712e+03 7.040131e+03 1.149155e+04
##Check mask width
mask.check(SN.bear.fit1)
spacing
buffer 2026.03395180444 1519.52546385333 1013.01697590222
1e+05 -30.42934 -30.42934 -30.42934
150000 -30.42934 -30.42934 -30.42934
2e+05 -30.42934 -30.42934 -30.42934
##project abundance
region.N(SN.bear.fit1)
estimate SE.estimate lcl ucl n
E.N 62.27256 14.25208 39.99140 96.96762 9
R.N 65.74089 11.86799 46.82338 94.12008 9
#####With SMALL Buffer
##Model 2
SN.bear.fit2 <- secr.fit(SN.ALL,model=list(D~1, g0~1), buffer = 1 * RPSV(SN.ALL)$SN2010)
SN.bear.fit2
link estimate SE.estimate lcl ucl
D log 6.753998e-05 1.052341e-05 4.985740e-05 9.149392e-05
g0 logit 2.126351e-01 3.624230e-02 1.501539e-01 2.921771e-01
sigma log 8.289553e+03 8.172233e+02 6.836258e+03 1.005180e+04
##Check mask width
mask.check(SN.bear.fit2)
spacing
buffer 762.584034826112 571.938026119584 381.292017413056
13339.587 -30.66813 -30.66110 -30.66131
20009.3805 -30.29588 -30.29648 -30.29586
26679.174 -30.26892 -30.26890 -30.26890
##project abundance
region.N(SN.bear.fit2)
estimate SE.estimate lcl ucl n
E.N 14.39947 2.243583 10.62956 19.50643 9
R.N 12.04958 NaN NaN NaN 9