Most ecological analyses and forecasts use weather station data or coarse interpolated, gridded temperature data. Yet, these products often poorly capture the microclimates experienced by organisms that live near the surface and respond to fine scale spatial and temporal environmental variation. Historic and projected future environmental data derived from spatial interpoation or dynamic modelling better capture the microclimate relevant to organisms and ecosystems. This site is designed to help you select microclimate datasets for projects characterizing responses to spatial and temporal environmental variability.
We temporally and spatially compare several related types of environmental data to weather station observations, which we assume reflect actual environmental conditions. We focus on datasets that offer (near) hourly data. Although several datasets are global, regional datasets are focused on the United States. We start by examining environmental forcing data, generated by spatial interpolation or dynamic modelling, which afford high spatial and temporal resolution but tend to be at a reference height (1-2m) above where most organisms reside. These environmental forcing data can then be processed using microclimate models, which characterise the processes of heat transport through air and soil to estimate vertical temperature and wind profiles. We use both detailed NicheMapR microclimate models and simple TrenchR vertical profile models to scale conditions from sensor to organism height. We include data both that we have processes using microclimate algorithms and that is pre-computed and stored as a microclimate dataset.
For weather station observations, we focus on three weather stations in Colorado, Oregon, and Hawaii that were chosen to represented different climate regimes and ecosystems, as well as examining weather stations across the United States in the US Climate Reference Network (USCRN). We examine environmental data from 2017, a recent year for which data is available from all datasets. We evaluate the data for both summer (month of July) and winter (month of January) conditions. We focus on environmental variables that are needed to estimate body temperatures of organisms using microcliamte and biophysical modelling.
Tabs in this application allow you to interactively filter datasets based on your data needs, to temporally compare monthly timeseries, to spatially compare microclimate datasets to weather station observations, and to compare the body temperature estimates produced by the microclimate datasets. Once you select a dataset, the TrEnCh Project Microclimate Data Users' Guide provides intructions for accessing and processing the data.
What are these stats?
We calculate three statistics based on hourly data.
Pearson correlation coefficient is the measurement of the strength of the relationship between two variables and their association with each other. It takes a value between -1 and 1, and the value further away from 0, the stronger the two variables are related.
Bias represents the average difference between the two datasets for each data point. The smaller the bias, the closer the data are to each other.
Root Mean Squared Error (RMSE) is defined as the square root of the average squared error. RMSE is heavy on the largest errors. The smaller the RMSE, the closer the data are to each other.