Modeling salt and water dynamics in the root zone
Abstract
The problem of irrigation with treated wastewater is both related to the crop effects of sub-optimal water quality and to soil salinity management in the root zone.
First, a numerical method is proposed for solving deterministic Richards' equation with a sink term describing root water uptake; the problem is treated in a Gardner framework, and different root uptake functions are considered. Richards equation with such a forcing term is first semidiscretized in time, and therefore integrated forward in space, according to the TMoL (transverse method of lines) approach. Results are successfully compared with MATLAB solver for parabolic PDEs. The second work refers to a machine learning approach for analysing a one-year dataset and for forecasting the behavior of salt concentration and soil water content. Soil data of salt concentration and water content are collected in a tomato crop in Stornarella (southeastern Italy). We link this dataset together with weather measurements collected over the same time window, and tackle the problem of measurement forecasting at the larger soil depth. We resort to random forest analysis, for which the entire measurement dataset is used to train and test the forecast model which reads data from 20 cm level to predict, with suitable confidence, salt concentration and water content at a depth of 50cm. Our machine learning model provides a reliability of 99%, with no overfitting and large scalability in terms of predicting variables. Future works concern the use of yield data for improving the analysis and its usability in an agronomical context. Moreover, mixing the two approaches is the challenge to be tackled in future papers.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMH135.0011B
- Keywords:
-
- 1804 Catchment;
- HYDROLOGY;
- 1829 Groundwater hydrology;
- HYDROLOGY;
- 1830 Groundwater/surface water interaction;
- HYDROLOGY;
- 1871 Surface water quality;
- HYDROLOGY