Date of Award

2024-12-01

Degree Name

Master of Science

Department

Computational Science

Advisor(s)

Natalia Villanueva Rosales

Abstract

Water management is important for residents in semi-arid urban areas due to increasing demand, water scarcity, and rising costs. It is estimated that in semi-arid regions, 40-70% of the household water consumption is used in landscaping. Therefore, urban landscaping water use can substantially contribute to water conservation. This work aims to estimate the water needs of urban landscaping vegetation to inform residents in semi-arid regions.Evapotranspiration indicates water and energy exchange between the atmosphere, soil, and vegetation. This interaction depends on solar radiation, evaporation, transpiration, and other biophysical parameters. Evapotranspiration has become a reference for water management in agriculture (e.g., crop irrigation). However, the evapotranspiration standards do not include urban landscaping vegetation. This project aims to create an evapotranspiration estimation model for urban landscaping vegetation in semi-arid regions using Machine Learning (ML).

To achieve this goal, four Deep Learning (DL) models were implemented: Multilayer Perceptron (MLP), Long-Short Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolutional-LSTM (Conv-LSTM). These models were trained using meteorological data from thirty-tree stations, including evapotranspiration. The DL models were compared against four benchmark Machine Learning (ML) regression models: linear regression (LR), XGBRegressor (XGBR), support vector regressor (SVR), and random forest regressor (RFR). The performance of the eight ML models to estimate evapotranspiration was evaluated using regression metrics, including R-squared (R2), mean squared error (MSE), root mean squared error (RMSE), and mean absolute percentage error (MAPE). From the DL models, the MLP network produced the best data fit (R2 = 0.9694). The ML regression models show a slightly better performance than MLP, with a difference of R2 less than 0.01. Results also show that the RFR model performed better during the summer months, and the MLP model performed better the rest of the year. These results suggest that more than one model can be used to optimize the evapotranspiration estimation depending on the season. In addition, a proof-of-concept is presented to illustrate how the proposed evapotranspiration models can be used to calculate the water-loss amount of an urban landscaping plant (i.e., sunflower) due to transpiration and evaporation. This process required identifying a plantâ??s species coefficient from the literature and evapotranspiration data from an additional weather station in the region. The evaluation metrics using the additional weather station data showed a performance of R2 = 0.9156 for the MLP model and an R2 = 0.9320 using the RFR model. The plant's water-loss amount can be used to identify the plant's daily water needs and inform irrigation schedules. This work is a first step towards supporting water conservation strategies in household urban landscaping, a critical need in semi-arid regions.

Language

en

Provenance

Recieved from ProQuest

File Size

55 p.

File Format

application/pdf

Rights Holder

Damian Lorenzo Gallegos Espinoza

Download

Share

COinS