Can we reliably estimate species richness with large plots? an assessment through calibration training