How many bootstrap replications are necessary for estimating remote sensing-assisted, model-based standard errors?