Drought Early Prediction Models

High variability of climate has seen over Central India such as very hot summer, very cold winter and intense monsoon. However, summer monsoon is important for the agriculture of central India and its agriculture is always affected by drought because MP receives most of precipitation during summer monsoon season (JJAS). Central India receives almost 95% of rainfall during JJAS season only. However, this region is vulnerable for extreme events like droughts and floods, therefore, this region experiences regular and long-term droughts that causes the most severe losses to the agricultural economy. Therefore, there is a strong need of a model or system to predict drought in advance with less resources. Among the many drought indicators such a s SPI, SPEI, PDSI and aridity index being used for monitoring, most used are the standardized precipitation index (SPI), standardized precipitation and evapo-transpiration index (SPEI), Palmer drought severity index (PDSI), soil moisture percentile, standardized runoff index and standardized soil moisture index.

Based on the distinguished values of SPI, the impact of drought can be assessed for appropriate repercussion to reduce impacts. Moreover, the impact on the agricultural production have seen because of the soil moisture deficiencies even for a short period during times of maximum crop water use. several studies had discussed the importance of drought monitoring which consider some or all of the above factors such as SPI, SPEI, PDSI etc. Based on multiple indicators the Global Integrated Drought Monitoring and Prediction System (GIDMaPS) provides drought information in real time used a probabilistic framework for drought forecasting based on accumulated rainfall to make predictions 20 days in advance.

There are a comparison study of SPI and SPEI. Their comparison study is concluded with the some impotent points. First, the fluctuation value and continuity are slightly different in the SPI and SPEI with the increasing timescale. Second, at various timescale the characteristics of the SPI and SPEI found quite different. Third and last, largest diffrence between SPI and SPEI was found for the shortest timescale. In the conclusion they also said the SPEI is more suitable than the SPI for drought monitoring but its region dependent. Meteorological drought variability in the Indian region using SPEI have examined trends and behaviour of meteorological drought events over India using SPEI indices and found that the increase in drought over eastern part of India was attributed due to decrease in rainfall, whereas because of the decrease in PET the frequency of drought occurrence over western arid region is decreasing presented an application of a seasonal prediction model using SPEI as a drought indicator. Moreover, SPEI is also used in many studies to identify drought.

Using traditional statistical or physical models (general circulation models) for forecasting drought at various time scale, the extraordinary development have seen in the past. Many data-driven machine learning algorithms for drought prediction have also been proposed for the monitoring and prediction of drought. Traditional statistical auto-regressive (AR) models to artificial neural networks (ANNs) have been used and proposed for forecasting drought and its key parameters. Data-driven modeling is established based on empirical relationships among drought indices (SPI, SPEI etc.) and their predictors such as SST and rainfall. ANN is the data-driven modeling technique that has attracted immense attention in time-series forecasting. In the hydrological forecasting, ANNs have been used all around the world with a significant degree of reliability applied ANNs to forecast rainfall using various predictors in Indonesia predicted summer monsoon rainfall in China using ANN did a comparison study of two different drought indices are the effective drought index (EDI) and the SPI for Iran used ANN for the prediction of soil moisture. Furthermore, dry regions with a significant probability of drought occurrence was regularly chooses for ANN's data-driven predictions forecasted SPI in Ethiopia and forecasted EDI in Africa by ANN. They used the hydro-meteorological parameters and climate signals as predictors for the forecasting of drought. In the dry region of eastern Australia and Vietnam, applied ANNs to forecast SPEI. ANN is quite successful due to the simplicity and easiness of handling the data and its implementation, and it's performance towards various drought forecasting studies.

On top of these, the use of climate indices that represent large-scale atmospheric and oceanic drivers of precipitation plays a important role in upgrade forecasting performance. Nevertheless, reliable or significant predictions always require a set of large in time scale datasets that can be used to train the ANN models. The problem with these enormous datasets are often incomplete and inaccessible in developing and undeveloped nations. Therefore, a trained drought forecasting model is available with the existing data to forecast future drought could be help full for poor or developing countries that can not afford the high cost of data.