Dewfall and rain wetting dynamics in an old growth Pseudotsuga menziesii canopy: modelling canopy wetness with random forest classification for use with long term climate data
Abstract
How long a wet canopy stays wet directly impacts forest dwelling organisms. Many tree species rely on leaf water uptake to alleviate water stress, poikilohydric bryophytes and lichens require surface wetness to maintain a positive carbon balance, and prolonged latent heat flux to the atmosphere buffers canopy temperatures, protecting canopy dwelling organisms from extreme heat. The microclimatic factors which determine residence time of moisture from rain and dew change substantially with height in large tree canopies, and future changes in precipitation and nighttime temperature will change wetness regimes throughout the canopy. In our study, we deployed sensors to measure surface wetness, relative humidity, and air temperature across a vertical gradient in the Discovery Tree, a ~450-year-old Pseudotsuga menziesii (Douglas-fir) at the H.J. Andrews Experimental Forest located in the Western Cascades of Oregon. We used a simple set of rules to create three classes of canopy wetness (dry, wet by dew, wet by rain) for the three extant years of in-canopy data. We employed random forest machine learning to build two models which predict wetness classes: one using within-tree observations of vapor pressure deficit, antecedent dewpoint depression, and time since last rainfall, and a second using the same observations from a nearby 4m tall weather station. These models reproduce observed wetness classes over a three year period with 91% and 86% overall classification accuracy. Then, we used our model built on nearby station data to predict wetness classes for the previous twenty years, using both the original station data and the station data with air temperature elevated by a constant 2o C. By artificially raising air temperature, we predicted an 18% (4% s.d.) mean decrease in time wet by dew annually. Our results show that commonly observed weather variables can be used to predict canopy wetness dynamics with a high degree of accuracy, including changes in dewfall caused by increasing nighttime temperatures. The demonstrated techniques will be useful in predicting changes in canopy moisture regimes and the subsequent impact on biota in forest systems where dewfall input is significant.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B53P2626S
- Keywords:
-
- 0414 Biogeochemical cycles;
- processes;
- and modeling;
- BIOGEOSCIENCES;
- 0426 Biosphere/atmosphere interactions;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0476 Plant ecology;
- BIOGEOSCIENCES