Drought assessment based on standardized precipitation index and its forecasting using neural networks
| dc.contributor.advisor | Singh, Praveen Vikram | |
| dc.contributor.author | Rai, Priya | |
| dc.date.accessioned | 2019-01-21T10:14:08Z | |
| dc.date.available | 2019-01-21T10:14:08Z | |
| dc.date.issued | 2018-06 | |
| dc.description.abstract | Drought is one of the most important environmental problems affecting the planet. Drought means scarcity of water, which disturbs the entire ecosystem and adversely affects various sectors of society, e.g. agriculture, hydropower generation, water supply and industry. Occurrence of drought and its forecasting are critical components of hydrology which play a major role in risk management, drought preparedness and mitigation. The identification, characterization and monitoring of droughts are of great importance in water resources planning and management. Based on its nature, drought may be divided into four different categories, viz., metrological, agricultural, hydrological and socioeconomical. Of these, the meteorological drought signifies the paucity of rainfall over a region for a considerable period of time. Drought occurs in nearly all climatic zones of the world at one time or other, but this creeping phenomenon mostly affects tropics and adjoining regions. This study was conducted using the rainfall data for 44 years (1971-2014) of Parbhani district of Marathwada region with specific objectives to find the best fit probability distribution, determine occurrence of meteorological drought based on Standardized Precipitation Index (SPI) for different time scales, develop Multilayer Perceptron Neural Network (MLPNN) models and assess the performance of these models for meteorological drought forecasting. The data were used to determine best fit probability distribution among Normal, log Normal, Gamma distribution and Weibull distribution based on KS goodness of fit statistic at 1% and 5% level of significance. The rainfall data were used to calculate drought and wet conditions in the region based on SPI values considering time scales of 1-, 3-, 6-, 9-, 12- and 24-month for the best fit distribution. The forecasting of the occurrence of the drought based on SPI was done at considered time scales using Multilayer Perceptron Neural Network (MLPNN) models. The lag found for different time scales i.e., SPI-1, SPI-3, SPI-6, SPI-12 and SPI-24 were 1, 2, 4, 7, 8 and 12 respectively. The NeuroSolutions Software was used with Levenberg-Marquardt learning rule and activation function was hyperbolic tangent with range -1 to 1. The performance of developed models was assessed using statistical indices such as root mean squared error (RMSE), Coefficient of determination (R2) and Coefficient of efficiency (CE). The Normal and Gamma distribution were found to be the best fitted on the rainfall data, however, Gamma distribution had an edge over the Normal distribution. During the total period of 528 months (1971-2014), 32 extreme drought, 69 severe drought, 82 extreme wet and 160 severe wet events occurred. The multilayer perceptron neural network models can be used only for forecasting of long duration SPI. The network architecture 7-8-1, 8-11-1 and 12-9-1 can be used for SPI-9, -12 and -24 months respectively for the study area. | en_US |
| dc.identifier.uri | http://krishikosh.egranth.ac.in/handle/1/5810091867 | |
| dc.keywords | drought resistance, precipitation, neural networks | en_US |
| dc.language.iso | en | en_US |
| dc.pages | 135 | en_US |
| dc.publisher | G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand) | en_US |
| dc.research.problem | Neural Networks | en_US |
| dc.sub | Soil and Water Engineering | en_US |
| dc.subject | null | en_US |
| dc.theme | Drought | en_US |
| dc.these.type | M.Tech | en_US |
| dc.title | Drought assessment based on standardized precipitation index and its forecasting using neural networks | en_US |
| dc.type | Thesis | en_US |