|
NEURAL NETWORK-BASED MODELS FOR STRUMA RIVER FLOW FORECASTING
|
|
|
Snezhanka Balabanova; Vesela Stoyanova; Valeriya Yordanova
|
|
|
||
|
10.5593/sgem2023/3.1
|
|
|
1314-2704
|
|
|
||
|
English
|
|
|
23
|
|
|
3.1
|
|
|
• Prof. DSc. Oleksandr Trofymchuk, UKRAINE
• Prof. Dr. hab. oec. Baiba Rivza, LATVIA |
|
|
||
|
Accurate river flow forecasting is an extremely important issue for proper management and optimal use of water resources as well as for warnings of extreme hydrometeorological events. Rainfall-runoff simulation is essential for short and longterm forecasting of the river discharge. Determining the relationship between rainfall and runoff is one of the most important tasks faced by hydrologists. This relationship is a nonlinear and extremely complex process influenced by many factors such as watershed topology, vegetation cover, soil types, river bed characteristics, groundwater aquifers, precipitation distribution, snowmelt, rural and urban activities.
Artificial Neural Networks (ANNs) are known as powerful and flexible models and are widely used in hydrology and forecasting. This paper aims to demonstrate the research and operational application of ANN in hydrologic modeling to construct an effective operational forecasting system of stream flow and potential flood risks in the studied area. The studied area is the Struma river Basin. The availability of long historical records and a good physical understanding of the hydrologic process in the area are very important in selecting the input predictors and designing a more efficient network. Historical data from automatic stations for the period 2015 - 2022 is selected to create the networks. The six hourly precipitation, daily temperature and runoff data from eleven subwatersheds are collected and used in developing the ANN. Additional analyses of lags are performed using correlation analysis of runoff at hydrometric stations at the outlet of the watersheds and correlation analysis of runoff and accumulated precipitation data in watersheds. The statistical estimates are Nash Sutcliffe model 0.8 - 0.9, MSE - 0.04 - 0.149, MAE 0.13 -0.176 and R 0.8-0.98. Operational forecasting is based on data from the global weather model ECMWF. |
|
|
||
|
conference
|
|
|
||
|
||
|
Proceedings of 23rd International Multidisciplinary Scientific GeoConference SGEM 2023
|
|
|
23rd International Multidisciplinary Scientific GeoConference SGEM 2023, 03 - 09 July, 2023
|
|
|
Proceedings Paper
|
|
|
STEF92 Technology
|
|
|
International Multidisciplinary Scientific GeoConference SGEM
|
|
|
SWS Scholarly Society; Acad Sci Czech Republ; Latvian Acad Sci; Polish Acad Sci; Russian Acad Sci; Serbian Acad Sci and Arts; Natl Acad Sci Ukraine; Natl Acad Sci Armenia; Sci Council Japan; European Acad Sci, Arts and Letters; Acad Fine Arts Zagreb Croatia; Croatian Acad Sci and Arts; Acad Sci Moldova; Montenegrin Acad Sci and Arts; Georgian Acad Sci; Acad Fine Arts and Design Bratislava; Russian Acad Arts; Turkish Acad Sci.
|
|
|
107-114
|
|
|
03 - 09 July, 2023
|
|
|
website
|
|
|
|
|
|
9136
|
|
|
Artificial Neural networks, hydrology, river flow forecasting
|
|