Investigating rainfall estimation from radar measurements using neural networks

Abstract

Rainfall observed on the ground is dependent on the four dimensional structure of precipitation aloft. Scanning radars can observe the four dimensional structure of precipitation. Neural network is a nonparametric method to represent the nonlinear relationship between radar measurements and rainfall rate. The relationship is derived directly from a dataset consisting of radar measurements and rain gauge measurements. The performance of neural network based rainfall estimation is subject to many factors, such as the representativeness and sufficiency of the training dataset, the generalization capability of the network to new data, seasonal changes, and regional changes. Improving the performance of the neural network for real time applications is of great interest. The goal of this paper is to investigate the performance of rainfall estimation based on Radial Basis Function (RBF) neural networks using radar reflectivity as input and rain gauge as the target. Data from Melbourne, Florida NEXRAD (Next Generation Weather Radar) ground radar (KMLB) over different years along with rain gauge measurements are used to conduct various investigations related to this problem. A direct gauge comparison study is done to demonstrate the improvement brought in by the neural networks and to show the feasibility of this system. The principal components analysis (PCA) technique is also used to reduce the dimensionality of the training dataset. Reducing the dimensionality of the input training data will reduce the training time as well as reduce the network complexity which will also avoid over fitting

Optimization of X-Band Radar Rainfall Retrieval in the Southern Andes of Ecuador Using a Random Forest Model

“…Thus, all together, the RF model performed remarkably with a goodness of fit up to CC = 0.83 and Eff = 0.67 at different stations and what is more importantly without any further bias adjustment using rain gauge data. The latter is a critical difference to other studies that had accomplished similar goodness of fit 0.6 < CC < 0.85 only mostly using high density rain gauge networks for bias correction as in [23] and [36] or vertical reflectivity profiles in the case of machine learning approaches as in [26,27]. The performance of the implemented RF technique in [20] (correlation coefficient of 0.82) was likewise comparable to our RF model (CC = 0.76) using a short-term dataset.…”

“…In other words, we are certain our results are unbiased regarding a point-pixel evaluation. This in turn makes difficult to compare our results with studies that used machine learning techniques with lower spatial resolution by means of S-, C-band radar data as in [26][27][28] (e.g., 1 km instead of 100 m) whose spatial resolution could minimize differences in rain gauge area coverage and radar volumetric content. While the latter is desirable, it also obscures the rainfall spatial variability which is remarkable in high mountain regions like the Andes.…”

“…While the latter is desirable, it also obscures the rainfall spatial variability which is remarkable in high mountain regions like the Andes. Furthermore, additional vertical profile of reflectivity may also contribute to the good performance (correlation coefficient ~0.8-0.9) of neural networks applications as in [26,27]. Nonetheless, also a lower performance (correlation coefficient of 0.37) was found by [28] in a test dataset from a new location 10 km apart from the radar.…”

“…Nonetheless, bias adjustment mostly depends on the rain gauge network distribution, operation and real-time availability, which are usually scarce and irregular in complex terrain. In contrast, the use of machine learning techniques for quantitative precipitation estimation has increased in recent years due to its simplicity and promising results [24][25][26][27]. This approach aims to reduce the complexity of the reflectivity correction process while properly simulate the relationship between reflectivity and rainfall.…”

A comparison of Artificial Neural Networks (ANN) and Local Linear Regression (LLR) techniques for predicting monthly reservoir levels

“…Over the years, ANN based models have been used for streamflow prediction (Han et al, 2002, Han et al, 2007Remesan et al, 2010), reservoir inflow prediction (Hassan et al, 2014), rainfall estimation (Kissi andCimen, 2012;Alqudah et al, 2013), Solar radiation estimation (Remesan et al, 2008;Shamim et al, 2010;Shamim et al, 2014), Land use classification (Srivastva et al, 2012) (Ishak et al, 2013), modelling snowmelt-runoff (Matheussen and Thorolfsson 1999;Tokar and Markus 2000), sediment transport prediction (Tayfur and Güldal, 2006), groundwater modeling (Rogers and Dowla, 1994;Lallahem and Mania, 2003;Shamim et al, 2004) ecological response to climate change (Trigo and Palutikof, 1999) and reservoir operation (Hasebe and Nagayama, 2002;Jain et al, 1999b;Raman and Chandramouli, 1996). American Society of Civil Engineering (ASCE's) task committee has also acknowledged the importance of ANNs as an important forecasting tool as elaborated in ASCE (2000a and b).…”