Applying GRACE Satellite Data and Hydro-Meteorological Parameters For
Total Page:16
File Type:pdf, Size:1020Kb
Applying GRACE satellite data and hydro- meteorological parameters for predicting the combined terrestrial evapotranspiration index (CTEI) based on an improved SVM and a greedy regression approach Ahmed Elbeltagi ( [email protected] ) Mansoura University Faculty of Agriculture Mehdi Jamei shohadaye university Ankur Srivastava Newcastle University Iman Ahmadianfar -- Mahfuzur Rahman - Harun I. Gitari ty Jaydeo K. Dharpure e Research Article Keywords: Drought index, Climate change, Greedy regression method, GRACE, Remote sensing, Support vector machine Posted Date: February 23rd, 2021 DOI: https://doi.org/10.21203/rs.3.rs-222279/v1 License: This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License 1 Applying GRACE satellite data and hydro-meteorological parameters for 2 predicting the combined terrestrial evapotranspiration index (CTEI) based on 3 an improved SVM and a greedy regression approach 1,2,* 3 4 5 4 Ahmed Elbeltagi , Mehdi Jamei , Ankur Srivastava , Iman Ahmadianfar , Mahfuzur 5 Rahman6 , Harun I. Gitari7, Jaydeo K. Dharpure 8,* 6 1Agricultural Engineering Dept., Faculty of Agriculture, Mansoura University, Mansoura 35516, 7 Egypt 8 2College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310058, 9 China 10 3Faculty of Engineering, Shohadaye Hoveizeh University of Technology, Dasht-e Azadegan, Iran 11 4School of Engineering, The University of Newcastle, Callaghan, NSW 2308, Australia 12 5Department of Civil Engineering, Behbahan Khatam Alanbia University of Technology, 13 Behbahan, Iran 14 6 International University of Business Agricultural Technology, Dhaka 1230, Bangladesh 15 7 Department of Agricultural Science and Technology, School of Agriculture and Enterprise 16 Development, Kenyatta University, P. O. Box 43844-00100, Nairobi, Kenya. 17 8 Centre of Excellence in Disaster Mitigation and Management (CoEDMM), Indian Institute of 18 Technology Roorkee, Uttarakhand, India 19 *Corresponding author: Correspondence to Ahmed Elbeltagi ([email protected]); 20 Jaydeo K. Dharpure ([email protected]). 21 Applying GRACE satellite data and hydro-meteorological parameters for 22 predicting the combined terrestrial evapotranspiration index (CTEI) based on 23 an improved SVM and a greedy regression approach 24 Abstract 25 Drought has become relatively one of the widespread extreme events that affect socio-economic 26 sphere, ecological balance, environmental conditions, crop-yields, climate feedback mechanism, 27 and water resources management. Monitoring of the drought is one of the challenging tasks, 28 hence required to develop new techniques in prediction of this natural calamity accurately and 29 timely. In this work, Combined Terrestrial Evapotranspiration Index (CTEI) that is forced 30 through precipitation (P) and potential evapotranspiration (PET) is used for estimating drought. 31 Also, this is a novel index developed for Indus-Ganga river basin, which also uses a Gravity 32 Recovery and Climate Experiment (GRACE) terrestrial water storage anomalies (TWSA) data. 33 Thirteen input-variables comprised of GRACE-TWSA, Global Land Data Assimilation System 34 GLDAS-TWSA, Groundwater Storage Anomaly (GWSA), P, Land Surface Temperature 35 ( LST), Wind Speed (WS), Evapotranspiration (ET), Runoff, air temperature ( ), PET, net- 36 shortwave (SWN) and long-wave (LWN) along with net radiation (NR) radiations were 37 implemented for predicting CTEI. Further, thirteen different combinations were analyzed using 38 an improved Sequential Minimal Optimization (SMO) algorithm for support vector machine 39 (SVM) and a greedy linear regression method. Therefore, the objectives of this study are to 40 develop several combinations based on machine learning models used for modeling CTEI, 41 compare the accuracy and stability of these models in CTEI prediction, and select the best 42 outcomes have been provided from these models. The performed error measures and statistical 43 indicators prove that Combo 2 and 3 owing to SVM model are superior to the best combinations 44 of Greedy model. Overall based on the results, the improved SVM is superior to Greedy model in 45 estimating CTEI. SVM (Combo2: R=0.81) in spite of having a higher range of relative error, is 46 superior to Greedy model (Combo3, R=0.77) for estimating CTEI. This method can be one of 47 promising alternatives for estimating CTEI over major river basins across the world. 48 Keywords: Drought index; Climate change; Greedy regression method; GRACE; Remote 49 sensing; Support vector machine. 50 1 Introduction 51 One of the most significant catastrophic phenomena in nature is drought that has 52 potentially severe outcomes on agriculture, socio-economic, and ecosystems (Wilhite 2000; Mo 53 2011). Drought-related disasters have been a large increase by changing climate worldwide over 54 the past decades, although with different intensities (Allen et al. 2011). In addition, with respect 55 to the global warming phenomenon, the frequency and effect of drought are gradually growing; 56 hence, it is very important to make a deep study on the drought phenomenon. With the purpose of 57 quantifying drought, many researchers used hydrological and meteorological factors affecting 58 drought to create various drought indices and compare drought frequency, duration, and intensity 59 and its effect on temporal and spatial scales (Marcos-Garcia et al. 2017; Oloruntade et al. 2017). 60 In this regard, several drought indices were introduced as a primary factor for predicting 61 drought such as Palmer Drought Severity Index (PDSI) (Palmer 1965), Standardized Precipitation 62 Index (SPI) (McKee et al. 1993), and Standardized Runoff Index (SRI) (Shukla and Wood, 2008; 63 Paul et al., 2018),Vegetation Drought Response Index (VegDRI) (Brown et al. 2008), 64 Standardized Stream flow Index (SSI) (Shukla and Wood 2008; Vicente-Serrano et al. 65 2010),Standardized Precipitation Evapotranspiration Index (SPEI) (Vicente-Serrano et al. 2010), 66 Multivariate Standardized Drought Index (MSDI) (Hao and AghaKouchak 2013), Gravity 67 Recovery and Climate Experiment (GRACE)derived Drought Severity Index (DSI) (Feng et al. 68 2017; Zhao et al. 2017),water storage deficit index (WSDI) (Sun et al. 2018), Combined 69 Climatologic Deviation Index (CCDI) (Sinha et al. 2019), and Combined Terrestrial 70 Evapotranspiration Index (CTEI) (Dharpure et al. 2020). 71 Among the above-stated drought indices, the SPI, PDSI, and SRI are the most common 72 indicators applied to drought characterization. The SPI uses the long-term precipitation dataset 73 (Wang et al. 2016), while the PDSI considers evaporation, precipitation, soil moisture, runoff, 74 and other variables (Zargar et al. 2011) to estimate long-term drought periods. Besides, the SPEI 75 is a drought indicator based on the PDSI and SPI for expressing meteorological drought. Based 76 on the SPEI, Wang and Chen (2014) evaluated the application of this indicator in China and 77 indicated that it has a good capability to present drought occurrence. Zhao et al. (2017) 78 demonstrated that the SPEI was suitable for monitoring drought in both short- and long-term in 79 China. Rivera et al. (2017) presented that the SSI can precisely show the extreme hydrological 80 drought in the Central Andes, Argentina. Tirivarombo et al, (2018) proved that the SPEI can 81 better describe drought than the SPI in the Kafue basin due to consider the evapotranspiration 82 factor. Dharpure et al.(2020) presented the CTEI by combining the GRACE and meteorological 83 variables such as precipitation (P) and potential evapotranspiration (PET). They indicated that 84 this index has several advantages compared with the other drought indices. For instance, it can 85 decrease simulation, numerical modeling, and parameterization. Also, the CTEI yielded a 86 measurement of climatic parameters and water storage variations to determine hydrological 87 drought occurrences. 88 Predicting drought to decrease the damage made by the drought phenomenon is an 89 inevitable issue. The most commonly utilized prediction models are artificial neural network 90 (ANN) Borji et al., (2016),support vector machines (SVMs) (Ahmad et al. 2010; Weng 2012). 91 Khan et al. (2020)developed wavelet-based hybrid ANN-autoregressive integrated moving 92 average (ARIMA) models for predicting the SPI and the Standard Index of Annual Precipitation 93 (SIAP). They indicated that the proposed model can provide very promising accuracy to predict 94 the drought events. Li et al. (2020) used three data models (i.e., SVM, random forest (RF), and 95 ARIMA) to predict the SPEI in the Guangzhong area in China. The authors showed that the RF 96 and SVM models have better performance to predict the SPEI than the ARIMA model. Three 97 data-driven models (i.e., SVM, ANN, and k-Nearest Neighbor (KNN)) were investigated for 98 predicting drought over Pakistan (Khan et al., 2020). According to the results, the SVM and ANN 99 were more accurate than the KNN to predict the droughts. 100 The main goal of the present study was to introduce a novel data-driven model for 101 predicting extreme droughts over India based on the Combined Terrestrial Evapotranspiration 102 Index (CTEI). This study employs the CTEI as the drought index because of its incorporation of 103 the GRACE data and meteorological variables (i.e., P and PET), thus better show the drought 104 than the other indices (Dharpure et al. 2020). In this study, two novel models called SMO-SVM