- Dona Paula, Goa, India.
- +91-0832- 2450327
- iiosc2020[at]nio[dot]org
| Abstract Submission No. | ABS-2022-06-0106 |
| Title of Abstract | Operational Forecast systems for Indian oil Sardine landing |
| Authors | Swarnali Majumder*, Elizabeth Holmes, T.M. Balakrishnan Nair, Sourav Maity, Nimit Kumar, Vera L. Trainer |
| Organisation | Indian National Center for Ocean Information Services |
| Address | Indian National Center for Ocean Information Services, Pragathi Nagar Hyderabad, Telengana, India Pincode: 500090 Mobile: 7989604586 E-mail: swarnali.majumder48@gmail.com |
| Country | India |
| Presentation | Oral |
| Abstract | The Indian oil sardine (Sardinella longiceps) is an important profit-oriented fish in India. The goal of this work is to develop an operational forecast bulletin that forecasts quarterly sardine landings that assist in fisheries planning. Our study area is the southwest coast of India, where the majority of the Indian oil sardines are landed. The forecast will use quarterly fish landing data provided by the Central Marine Fisheries Research Institute (CMFRI) in Kochi, India. These landings data are based on a stratified multi-stage sample design, which considers landing centers, number of fishing days and boat net combinations in fishing operations. The forecast models will use environmental covariates which reflect ocean conditions relevant for sardine recruitment, survival and exposure to the fishery: location-specific monthly composites of sea surface temperature (SST), chlorophyll-a (CHL), upwelling (UPW), Oceanic Niño Index (ONI) and precipitation are derived from satellite. Researchers have found that a variety of environmental variables are correlated with landings of the Indian oil sardine, however these variables may be location specific and many are collinear. In a recent study (Holmes et al. 2019), we applied time series models with non-linear covariate responses to study the relationship between the ocean environment in critical months of the sardine life-cycle and future sardine landings in winter versus the summer monsoon. Holmes et al. (2019) used the life-cycle of the oil-sardine to drive the structure of the models. In this study, we take a machine-learning approach to the problem of developing a sardine landings forecast. We apply decision tree model with quarterly landings as the response and a suite of environmental covariates. Decision tree provides a machine-learning approach to variable selection and are robust to collinearity for the purpose of prediction. Comparison of these two fundamentally different approaches for developing forecasts, machine-learning via decision tree versus a biologically-informed model study, helps us understand how best forecast highly variable pelagic resources. |