Application of Generalized Additive Model for Rainfall Forecasting in Rainfed Pothwar, Pakistan

Abstract

Climatic variations affect growers of dry regions, and so the agricultural management techniques require modification according to the timing and amount of precipitation for the optimization of yield and economic output for a specified season and location. Farm manager preparedness depending on past practices can be enhanced by long-range skilled forecasting of rainfall. The well-known modes of interannual fluctuations affecting the Indian subcontinent are the Indian Ocean Dipole (IOD) and El-Niño Southern Oscillation (ENSO). Dry regions of Pakistan, i.e., Pothwar, are facing a number of key challenges in the prediction of irregular rain. Modeling skewed, zero, nonlinear, and non-stationary data are a few of the main challenges. To deal with this, a probabilistic statistical model was used in three of the dry areas of Pothwar to predict monsoon and wheat-growing season. To find out the prospects of rainfall, occurring in the system, the model utilizes logistic regression through generalized additive models (GAMs). Our study exploits climatic predictors (Pacific and the Indian Ocean SSTs demonstrating the status of the IOD and the ENSO) affecting rainfall fluctuations on the Indian subcontinent for their effectiveness in predicting seasonal rainfall (three rainfall intervals and the monsoon rains throughout the wheat-growing period). The outcome demonstrated that the observed area had the amount and fluctuation of rainfall determined by SSTs, so predictions can be carried out by intellect to overpass the gaps among average and potential wheat yield with a change in management practices, i.e., appropriate time of sowing and use of suitable genotypes. In addition, the forecasting ability score, i.e., R2, RMSE (root-mean-square error), BSS (Brier skill score), S% (skill score S), LEPS (linear error in probability space), NSE (Nash-Sutcliffe model efficiency coefficient), and ROC (receiver operating characteristics, p-value), assessed validation of model for rainfall prediction to verify the effectiveness of GAM and to formulate contrast among varying validation abilities to do cross-validation of rainfall prediction. Likewise, the forecast systems present substantial benefits in enhancing general operational management when used in agriculture production across the whole value chain.

Keywords

ENSO; Forecasting; GAMs; IOD; Management; SSTs

Published in

Title: Systems Modeling
Publisher: Springer Singapore

SLU Authors

• Ahmed, Mukhtar

  • Department of Agricultural Research for Northern Sweden, Swedish University of Agricultural Sciences
    
  • Pir Mehr Ali Shah Arid Agriculture University

UKÄ Subject classification

Agricultural Science
Meteorology and Atmospheric Sciences
Probability Theory and Statistics

Publication identifier

• DOI: https://doi.org/10.1007/978-981-15-4728-7_15
• ISBN: 9789811547270
• eISBN: 978-981-15-4728-7

Permanent link to this page (URI)

https://res.slu.se/id/publ/129851