Next Article in Journal
Floods and Countermeasures Impact Assessment for the Metro Colombo Canal System, Sri Lanka
Previous Article in Journal
Merging Real-Time Channel Sensor Networks with Continental-Scale Hydrologic Models: A Data Assimilation Approach for Improving Accuracy in Flood Depth Predictions
Article Menu

Export Article

Open AccessArticle
Hydrology 2018, 5(1), 10; https://doi.org/10.3390/hydrology5010010

Comparing Machine Learning and Decision Making Approaches to Forecast Long Lead Monthly Rainfall: The City of Vancouver, Canada

1
Department of Civil Engineering, McMaster University, Hamilton, ON L8S 4L8, Canada
2
Department of Land, Air, and Water Resources, University of California, Davis, CA 95616, USA
*
Author to whom correspondence should be addressed.
Received: 21 December 2017 / Revised: 10 January 2018 / Accepted: 17 January 2018 / Published: 22 January 2018
View Full-Text   |   Download PDF [2044 KB, uploaded 22 January 2018]   |  

Abstract

Estimating maximum possible rainfall is of great value for flood prediction and protection, particularly for regions, such as Canada, where urban and fluvial floods from extreme rainfalls have been known to be a major concern. In this study, a methodology is proposed to forecast real-time rainfall (with one month lead time) using different number of spatial inputs with different orders of lags. For this purpose, two types of models are used. The first one is a machine learning data driven-based model, which uses a set of hydrologic variables as inputs, and the second one is an empirical-statistical model that employs the multi-criteria decision analysis method for rainfall forecasting. The data driven model is built based on Artificial Neural Networks (ANNs), while the developed multi-criteria decision analysis model uses Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) approach. A comprehensive set of spatially varying climate variables, including geopotential height, sea surface temperature, sea level pressure, humidity, temperature and pressure with different orders of lags is collected to form input vectors for the forecast models. Then, a feature selection method is employed to identify the most appropriate predictors. Two sets of results from the developed models, i.e., maximum daily rainfall in each month (RMAX) and cumulative value of rainfall for each month (RCU), are considered as the target variables for forecast purpose. The results from both modeling approaches are compared using a number of evaluation criteria such as Nash-Sutcliffe Efficiency (NSE). The proposed models are applied for rainfall forecasting for a coastal area in Western Canada: Vancouver, British Columbia. Results indicate although data driven models such as ANNs work well for the simulation purpose, developed TOPSIS model considerably outperforms ANNs for the rainfall forecasting. ANNs show acceptable simulation performance during the calibration period (NSE up to 0.9) but they fail for the validation (NSE of 0.2) and forecasting (negative NSE). The TOPSIS method delivers better rainfall forecasting performance with the NSE of about 0.7. Moreover, the number of predictors that are used in the TOPSIS model are significantly less than those required by the ANNs to show an acceptable performance (7 against 47 for forecasting RCU and 6 against 32 for forecasting RMAX). Reliable and precise rainfall forecasting, with adequate lead time, benefits enhanced flood warning and decision making to reduce potential flood damages. View Full-Text
Keywords: real time rainfall forecasting; large scale climate signals; artificial neural networks; multi-criteria decision analysis real time rainfall forecasting; large scale climate signals; artificial neural networks; multi-criteria decision analysis
Figures

Figure 1

This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. (CC BY 4.0).
SciFeed

Share & Cite This Article

MDPI and ACS Style

Zahmatkesh, Z.; Goharian, E. Comparing Machine Learning and Decision Making Approaches to Forecast Long Lead Monthly Rainfall: The City of Vancouver, Canada. Hydrology 2018, 5, 10.

Show more citation formats Show less citations formats

Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Metrics

Article Access Statistics

1

Comments

[Return to top]
Hydrology EISSN 2306-5338 Published by MDPI AG, Basel, Switzerland RSS E-Mail Table of Contents Alert
Back to Top