Bias Correction of Climate Modeled Temperature and Precipitation Using Artificial Neural Networks

Abstract

Climate studies and effective environmental management require unbiased climate datasets. This study develops a new bias correction approach using a three-layer feedforward neural network to reduce the biases of climate variables (temperature and precipitation) over northern South America. Air and skin temperature, specific humidity, and net longwave and shortwave radiation are used as inputs to the network for bias correction of 6-hourly temperature. Inputs to the network for bias correction of monthly precipitation are precipitation at lag 0, 1, 2, and 3 months, and also the standard deviation of precipitation from 3 × 3 neighbors around the pixel of interest. The climate model data are provided by the Community Climate System Model, version 3 (CCSM3). Results show that the trained artificial neural network (ANN) can improve the estimation error and correlation of the variables for both calibration and validation periods even when there is a low temporal consistency between the time series of the model data and targets. The developed model is also able to modify the probabilistic structure of the variables although the quantile-based information is not directly considered in the network. The ANN model outperforms linear regression, which is used for comparison purposes. The new method can be used to produce bias-corrected climate variables that can be used as forcing to hydrological and ecological models.

Regression‐based regionalization for bias correction of temperature and precipitation

“…They are defined as reaching 70, 80, and 90% of the performance of the trained model using observations at every pixel. Pixels are identified as improved validating pixels when their biases (Bias VlP ) are smaller than: where Bias PbyP is the bias between the bias‐corrected temperature and the target, when the biases are corrected pixel by pixel (Moghim and Bras, ). Figure shows the delineation of the study domain and the corresponding training pixels for each region in March for 70, 80, and 90% performances.…”

“…The study domain was delineated and the training pixels were determined for each month and season using the LR model. To evaluate the regionalization ability of the ANN model, we train (calibrate) a three‐layer feedforward neural network developed by Moghim and Bras () at the defined training pixels and apply the trained model to reproduce bias‐corrected temperature and precipitation at all pixels within the delineated regions with a desired accuracy (at least 80% performance level of pixel‐by‐pixel correction). The regionalization ability of the ANN and LR models to improve the results in terms of mean squared error (MSE), Bias, correlation ( ρ ), and Kolmogorov–Smirnov test (KS) for temperature and precipitation is expressed as where SS (skill score) refers to the percent improvement and A to the statistic (MSE, Bias, ρ , or KS).…”

“…Moghim and Bras () developed a data driven approach using an artificial neural network (ANN) trained at each pixel using targets (observations) over northern South America. Results revealed that the trained model can improve all statistics including the bias (errors) and probabilistic structure of the climate model outputs relative to the observations.…”

“…Their results showed that although the performance of the distributionbased method is better than the mean-based method, they cannot remove the biases of precipitation over five river basins in North America due to low correlation between time series of modelled and observed precipitation. Moghim and Bras (2017) developed a data driven approach using an artificial neural network (ANN) trained at each pixel using targets (observations) over northern South America. Results revealed that the trained model can improve all statistics including the bias (errors) and probabilistic structure of the climate model outputs relative to the observations.…”

Prediction of the Seasonal Changes of the Chloride Concentrations in Urban Water Reservoir

“…At first, artificial neural networks were used in economic sciences [16] and in meteorology [17]. Currently, more and more often used in prediction of changes in surface waters, demersal waters and hydrological changes [18][19][20][21][22][23][24][25][26][27][28].…”