Evaluating different machine learning methods to simulate runoff from extensive green roofs

Abstract

Abstract. Green roofs are increasingly popular measures to permanently reduce or delay storm-water runoff. The main objective of the study was to examine the potential of using machine learning (ML) to simulate runoff from green roofs to estimate their hydrological performance. Four machine learning methods, artificial neural network (ANN), M5 model tree, long short-term memory (LSTM) and k nearest neighbour (kNN), were applied to simulate storm-water runoff from 16 extensive green roofs located in four Norwegian cities across different climatic zones. The potential of these ML methods for estimating green roof retention was assessed by comparing their simulations with a proven conceptual retention model. Furthermore, the transferability of ML models between the different green roofs in the study was tested to investigate the potential of using ML models as a tool for planning and design purposes. The ML models yielded low volumetric errors that were comparable with the conceptual retention models, which indicates good performance in estimating annual retention. The ML models yielded satisfactory modelling results (NSE >0.5) in most of the roofs, which indicates an ability to estimate green roof detention. The variations in ML models' performance between the cities was larger than between the different configurations, which was attributed to the different climatic characteristics between the four cities. Transferred ML models between cities with similar rainfall events characteristics (Bergen–Sandnes, Trondheim–Oslo) could yield satisfactory modelling performance (Nash–Sutcliffe efficiency NSE >0.5 and percentage bias |PBIAS| <25 %) in most cases. However, we recommend the use of the conceptual retention model over the transferred ML models, to estimate the retention of new green roofs, as it gives more accurate volume estimates. Follow-up studies are needed to explore the potential of ML models in estimating detention from higher temporal resolution datasets.

The Data Synergy Effects of Time‐Series Deep Learning Models in Hydrology

“…This setup permitted simultaneous training and simulation over thousands of sites or more. However, in many other machine learning studies, following the conventional wisdom of stratification, geoscientists still tend to train separate models using data from each site (Duan et al., 2020; Herath et al., 2021; Petty & Dhingra, 2018), or each region composed of sites with similar environmental conditions (Abdalla et al., 2021; Sahoo et al., 2017).…”

Variability of green infrastructure performance due to climatic regimes across Sweden

“…The process of model calibration or dataset training was based on a backpropagation algorithm by adjusting the weights and computing the error between the output and the corresponding target value and propagating this backward through the network to adjust weights and produce the desired output (Abdalla et al, 2021). The optimum number of neurons in the hidden layers was based on a trial error process, starting with a small number of neurons, which gradually was increased until obtaining the lowest forecasting error (mean square error function).…”

Modeling the hydrological benefits of green roof systems: applications and future needs

“…25,31,37,55 In addition, in recent years, data-driven methods such as machine learning techniques have been investigated. 56 Yet, data-driven methods will not be discussed either, because field data scarcity is a common issue that managers and developers face to train and test models. Other software packages, such as MUSICX, 57 have previously been considered for green roof modeling, but as these are not as widely used and either require licenses for use or are not open source, they will not be discussed.…”

“…Stormwater control primarily relates to the mechanical process of water movement (infiltration) within GR substrate. 30,56,59 Simulation of soil water transport, thus, is Fig. 1 Components and water fluxes of a simplified GR model.…”

“…18,[34][35][36]78,92,94 The second strategy is crossvalidation, in which models are tested among different sites and climate forcings. 41,56,58 This strategy is becoming increasingly popular, because it is important for GR planning that the model can predict the performance of new implementation when data are unavailable. Nevertheless, the transferred model often fails to predict GR outflow at different sites.…”