Estimation and easy calculation of the Palmer Drought Severity Index from the meteorological data by using the advanced
- PDF / 2,537,426 Bytes
- 14 Pages / 547.087 x 737.008 pts Page_size
- 28 Downloads / 221 Views
Estimation and easy calculation of the Palmer Drought Severity Index from the meteorological data by using the advanced machine learning algorithms Fatih Tufaner
&
Abdurrahman Özbeyaz
Received: 4 May 2020 / Accepted: 2 August 2020 # Springer Nature Switzerland AG 2020
Abstract Drought, which has become one of the most severe environmental problems worldwide, has serious impacts on ecological, economic, and socially sustainable development. The drought monitoring process is essential in the management of drought risks, and drought index calculation is critical in the tracking of drought. The Palmer Drought Severity Index is one of
the most widely used methods in drought calculation. The drought calculation according to Palmer is a timeconsuming process. Such a troublesome can be made easier using advanced machine learning algorithms. Therefore, in this study, the advanced machine learning algorithms (LR, ANN, SVM, and DT) were employed to calculate and estimate the Palmer drought Z-index
Highlights • The Palmer Drought Severity Index (PDSI) was modeled to reduce mathematical computational complexity through four machine learning algorithms (LR, ANN, SVM, and DT). • In the studied models, the meteorological variables were used as input data. • Palmer’s drought computing approach has been re-coded in the Matlab environment. And runoff (RO) and Palmer Index data were obtained by using this software. • In the study, the best correlation coefficient was obtained in the ANN algorithm with 0.98. The MSE value was 0.40 at this success. • A novel training data using meteorological variables were developed and shared online. • By using the developed training data, Palmer drought index values for other regions will be able to be calculated by researchers easily. Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10661-020-08539-0) contains supplementary material, which is available to authorized users. F. Tufaner Department of Environmental Engineering, Engineering Faculty, Adiyaman University, Adiyaman, Turkey F. Tufaner (*) Environmental Management Application and Research Center, Adiyaman University, 02040 Adiyaman, Turkey
e-mail: [email protected] A. Özbeyaz Department of Electrical-Electronics Engineering, Engineering Faculty, Adiyaman University, Adiyaman, Turkey
576
Page 2 of 14
values from the meteorological data. Palmer Z-index values, which will be used as training data in the classification process, were obtained through a specialpurpose software adopting the classical procedure. This special-purpose software was developed within the scope of the study. According to the classification results, the best R-value (0.98) was obtained in the ANN method. The correlation coefficient was 0.98, Mean Squared Error was 0.40, and Root Mean Squared Error was 0.56 in this success. Consequently, the findings showed that drought calculation and prediction according to the Palmer Index could be successfully carried out with advanced machine learning algorithms.
Keywords
Data Loading...
