Document Type : Research Paper
Authors
1 Assistant Professor, university of Tabriz
2 Faculty Member of Tabriz University
3 Department of Water Engineering, University of Guilan
4 Ph.D. student of Tabriz University
Abstract
Introduction
Floods are a natural occurrence that causes casualties, livestock losses and damage to buildings, facilities, gardens, fields and natural resources every year. Therefore, rainfall estimates have long been considered by researchers in various fields, and along with the advancement of science and the emergence of new technologies, many advances have been made in the methods of rainfall estimation and evaluation and validation to achieve the best method. In the last twenty years, there has been a lot of progress in rainfall estimation methods. This advancement is due to the possibility of using a lot of information from different parts of the world, better understanding of atmospheric phenomena, exchanges and atmospheric rotations, improving the performance of models, progress in various surveillance tools such as radar and satellite and computer power. The methods used to estimate precipitation, especially in the short term, have shortcomings and are generally based on numerical forecasting models or the use of empirical analyzes, which are usually not very accurate for multi-hour intervals, so the use of satellite data It has been recommended as a supplement to address this problem, and doing so could greatly help increase the accuracy of numerical models for rainfall estimates.
Methodology
The study used the physical properties of a cloud of five waves between 2011 and 2015.
The data of the second generation of MSG meteorological satellite has good coverage on different regions of Iran. The satellite has 12 channels on the region and produces accurate products. Some of these products are in line with the physical properties of the cloud used in this study. These products are produced daily every 15 minutes and include cloud peak pressure (CTP), cloud peak temperature (CTT), cloud light depth (COT), thermodynamic cloud phase (CPH), and the volume of water in the cloud. Density (CWP) are the effective radius of cloud droplets (REFF) and cloud type (CT). Was obtained.
The criterion for the accuracy of the calculations was the two MAE statistics
Equation 1:
Equation 2:
Results and discussion
In this study, TRMM satellite data was considered as control data. After receiving TRMM images in MATLAB software environment, programming was performed and precipitation data were extracted from NETCDF files. After extracting TRMM satellite data, Meteosat satellite products were prepared through the CMSAF database and their data were extracted using MATLAB software code. In the study of waves, the coefficient of determination in the GPR model was 0.72 in the experimental section and 0.77 in the training section. In the TD model, the determination coefficient is calculated in the experimental section 0.64 and in the training section 0.87. However, in the neural network model, the coefficient of determination is 0.68 in the experimental section and 0.72 in the training section. The results show a good relationship between the components studied.
Investigating the Effects of Cloud Physical Properties: One of the methods for determining the effectiveness of each of the physical properties of the cloud in estimating rainfall is the sensitivity analysis method. After calculating the coefficient of determination and the error coefficient, the sensitivity of each of the physical properties in estimating the precipitation was performed by the method of calculating the sensitivity analysis. Sensitivity analysis was calculated for all waves. Calculations show that the cloud type is most effective, followed by the effective radius of the cloud droplets and then the optical depth of the cloud in the second and third positions, respectively. Among the physical properties studied, the lowest effect is related to the cloud phase.
To investigate the relationship between the physical characteristics of the cloud and the amount of precipitation, five waves of pervasive precipitation were selected between 2011 and 2015. Rainfall data from the region's stations were extracted. In order to validate the TRMM data, a comparison was made between the precipitation data of the selected stations and the precipitation of this satellite. Metoost satellite products were used to extract the physical properties of the cloud. After extracting the data, the physical properties of the cloud were matched to the time scale of the data and evaluated using TRMM satellite rain as a control.
Conclusion
The selection criteria were such that the waves lasted for at least two days and covered the entire area. On the day of the operation, the precipitation information of the meteorological stations of the region was obtained and also the precipitation information of TRMM satellite was extracted. In order to validate the data of TRMM satellite, the information of meteorological stations was compared with TRMM precipitation and obtained the necessary correlation. In order to get a better result, the matching of numbers was done in terms of time scale.
In the next step, using the meteosat satellite products, the physical properties of the cloud were obtained for all waves. Data were extracted at all stages for each pixel. Then the data correlation matrix was performed with three models of GPR, TD and MLPBR, the results of which are given in Table One. Due to the use of different models as well as the study of 8 physical properties of the cloud, the results show a high relationship between the components of the study, so that the coefficient of determination in the GPR model for the experimental and training sections was 0.7 and 0.77, respectively. These coefficients for the TD model in the experimental and training sections are 0.64 and 0.87, respectively. In the artificial network model (MLPBR), the coefficients obtained in the experimental and training sections are 0.68 and 0.72, respectively. The numbers obtained indicate a relatively good relationship between the components. Sensitivity analysis was performed. Sensitivity analysis results show that the cloud type feature has the greatest effect on precipitation and then the effective radius of cloud droplets and then cloud light depth are in the second and third positions, respectively. Among the physical properties studied, the lowest effect is related to the cloud phase.
Keywords
Main Subjects
)