The sensitivity of topoclimatic models to fine-scale microclimatic variability and the relevance for ecological studies

Microclimatic loggers are increasingly used to collect data from various habitats and interpolate ecologically meaningful landscape-scale topoclimatic grids. However, it is unknown how sensitive these grids are to finer-scale variations in microclimate. We performed a sensitivity analysis using three microclimatic loggers at 27 sites for 5 months in a semi-arid region of Western Australia. We partitioned the within- and between-site variance in temperature and produced 100 different topoclimatic models using a random sensor from each site. For the coldest temperatures, we found within-site variance was negligible (3 %), and models were strong (r 2 = 0.74) and the coefficients consistent. However, for the hottest temperatures, there was substantial within-site variance (39 %), and models were weaker (r 2 = 0.27) and more sensitive. We concluded that careful site design is needed to maximise the reliability of topoclimatic grids, including using large sample sizes, ensuring there is low predictor collinearity and sampling full environmental gradients.