Moving from Empirical Estimation of Humidity to Observation: A Spatial and Temporal Evaluation of MTCLIM Assumptions Using Regional Networks

Humidity is an essential component to the study and practice of hydrology, agriculture, ecology, and human health. Evapotranspiration estimation, human and animal comfort, hydrologic modeling, and many other applications require spatially and temporally continuous humidity data as direct input. The majority of this input comes in the form of high resolution, gridded data products. However, because of the relative difficulty in obtaining direct measurements of humidity in comparison to other variables, a lack of these observations has historically existed across North America, and the globe. This led to the development of empirical algorithms in order to estimate humidity, most notably, MTCLIM (Running, et al., 1987), which has served as the basis for several similar, but modified, techniques. In order to estimate humidity from available data, such as temperature and precipitation, these techniques make major assumptions about the spatial and temporal behavior of humidity. These major assumptions include: 1) dew point remains constant throughout the day, 2) minimum temperature can be used as a proxy to the average daily dew point, and 3) humidity is a very temporally and spatially homogenous variable, meaning that both changes on short time scales and over short distances are negligible. These assumptions are evaluated across the coterminous United States and southern Canada, through a geographic, seasonal, and diurnal climatology of humidity (dew point) using observed humidity measurements from 12000 stations for the period 1948 – 2014. Our results show that the validity of the above assumptions varies significantly by season and ecoregion, including coastal proximity and topographic complexity. Through a brief overview of the station networks now available, we propose that there is now enough station data available to directly interpolate humidity across the United States on a daily basis, greatly reducing the need to estimate this variable.