Machine learning and wet prairies: An additional tool for land managers

Link To Share This Poster: https://cdmcd.co/mXy8rB

Enrique Gomezdelcampo, School of Earth, Environment and Society, Bowling Green State University, Bowling Green, OH and Priyanka More, South Carolina Department of Natural Resources, Columbia, SC

Background/Question/Methods

There is a vital connection between the hydrology and the ecology of a wet prairie. Temporal water table variation is one of the main determinants of a healthy wet prairie. Particularly, when restoring wet prairies it is important to forecast how the water table will fluctuate on different seasons and at different temporal resolutions. Simulations of water table levels are effectively computed using physical based models. However, these models require detailed parameterization and are computationally intensive. Machine learning using artificial neural networks (ANNs) provides a simpler and efficient method for short-term water table forecast when relatively long-term records are available to train a neural network.
How good is an ANN forecast with respect to a physical model forecast? Is temporal high resolution data needed for a successful AAN forecast? How can ANN be used as a tool for wet prairie restoration?
A nonlinear autoregressive neural network with exogenous inputs (NARX) was used to train the model and to forecast water table levels at an hourly and a daily time step. The NARX model used a Levenberg-Marquardt learning algorithm along with a combination of input delay, feedback delay, and hidden layers. The data to train the model came from three years (May 2015-May 2018) of hourly water table data, hourly/daily precipitation, air temperature, relative humidity, and wind speed data from the Oak Openings Region of northwest Ohio.
Results/Conclusions
The NARX model provided successful short-term forecasts (6 months) for hourly and daily temporal resolutions. The R² for the hourly testing period was 0.85 and for the daily model was 0.91. The NARX model was not able to predict a sudden increase in water table levels due to a large snowmelt event in 2018, but it is not surprising as the model was not trained using snow melt events or a snow depth variable. As any data-based model, NARX is highly influenced by the quantity and quality of the input data set. Regardless, of its current limitations, land managers could use the NARX model to better understand the behavior of water table in wet prairies, one of the main drivers of this natural community.