Figure 3.

was found to be maximum followed by second quarter Q2 landings (7033.78). The catch of third quarter Q3 (3216.19) and fourth quarter Q4 (3226.56) are nearly equal for the period 2001–2015. The quarterly average landings Q1 (495.15) are found to be more than Q2 (468.91), but the variation in catch in Q2 (164.69) is higher than Q1 (126.61). The mean catch of Q3 (214.41) is similar to that of Q4 (215.10), but the variation in catch in Q3 (84.03) is higher than Q4 (55.60). The variation in quarter wise shrimp catch data in Chilika lagoon was shown in Figure 2. Temporal variation using image mapping of the lagoon in Figure 3 showed that catch increases in the first decade but decreases thereafter, and in 2015, catch was found comparatively low with respect to 2010–2012. The highest peak in catch was observed during 2011

Quarter wise total, mean and standard deviation of shrimps landing in Chilika lagoon for the period

Time (quarter) Total catch (MT) Mean catch (MT) Std. (MT) Q1 7427.36 495.15 126.61 Q2 7033.78 468.91 164.69 Q3 3216.19 214.41 84.03 Q4 3226.56 215.10 55.60

The best fitted SARIMA model was selected based on model selection criteria (AIC and SBC), and SARIMA (0,1,1)(0,1,1)4 model with lowest AIC (509.23) and SBC (517.26) found the best fitted model Table 2. The parameter's estimate of bestfitted SARIMA (0,1,1)(0,1,1)4 model in Table 3 showed nonseasonal moving average (component) parameter Q at lag 1 (0.71, p < 0.001) and seasonal moving average parameter (0.78, p < 0.001) at lag 4 as significant effect in the developed model. Two catches of second quarter for the year 2005 and 2011 showed the significant outliers, and catch showed significantly increase during this period. The R2 value 0.70 showed the good fit of the SARIMA (0,1,1)(0,1,1)4 model. The residual of the developed SARIMA model was tested by Ljung, and Box test (p > 0.05) showed the adequacy of the fitting of the estimated models. The developed SARIMA model was validated with the actual shrimp catch data (Table 4). The annual catch prediction error for the developed model was found below 10%,

and then starts decreasing till 2015 (Figure 3).

Time Series Analysis - Data, Methods, and Applications

Quarter wise shrimp catch (MT) in Chilika lagoon for the period 2001 to 2015.

Table 1.

2001–2015.

Figure 2.

88

Temporal variation of shrimp production of Chilika lagoon ecosystem from 2001 to 2015.


### Table 2.

Comparisons of best fitted SARIMA model with fit statistics (AIC, SBC) for quarter wise shrimp landings in Chilika lagoon for the period 2001–2015.


#### Table 3.

Best fitted SARIMA (0,1,1)(0,1,1)4 model parameter's estimate and their significance for quarter wise shrimp landings in Chilika lagoon for the period 2001–2015.


#### Table 4.

Yearly actual versus predicted shrimp catch forecast with % prediction error using SARIMA (0,1,1)(0,1,1)4 model in Chilika lagoon for the year 2014–2018.

e.g., for year 2014 (9.14%) and 2015 (1.02%). The percentage forecast error in shrimp catch for the year 2016, 2017, and 2018 was found to be 4.5, 5.9, and 7.36%, respectively (Table 4). The quarter wise forecast of shrimp catch for the period 2016–2018 with 95% upper and lower catch was shown in Table 5. The quarter wise actual catch versus predicted catch with 95% confidence limit of shrimp catch for the period of 2001 to 2018, showing an increase in catch with respect to 2015, is shown in Figure 4.

An application of SARIMAX model for total fish production forecasting is analysed for Chilika lagoon [14]. Mullets, clupeids, engraulids, perches, catfishes, sciaenids, threadfins, cichlids, tripod fishes, featherbacks, murrels, carps and

> minnows with some miscellaneous catch are the major fish catch of the lagoon with an average annual estimate of fish landing 12,000 tons per annum from 2000–2001 to 2014–2015. The fish landing was fluctuated between 10,286.34 tone (2003–2004) and 6463.92 tone (2006–2007) with average at 7725.42 tone contributing 63.65% of total fish landing of the whole lagoon. Around 64.88% of commercial catch was contributed by the yield of fish since last 15 years of post-restoration period. More than 90% fish species which contributed to the commercial landing of Chilika are of brackish or marine habitats. SARIMAX model was developed using monthly time series fish catch data, and regressor is used as a factor (combination of lagoon

> Monthly actual fish catch versus predicted (forecasted) catch by SARIMAX model for the period April 2001 to

Quarter wise actual versus predicted shrimp catch (MT) with 95% confidence limit of Chilika lagoon for the

Forecasting Shrimp and Fish Catch in Chilika Lake over Time Series Analysis

DOI: http://dx.doi.org/10.5772/intechopen.85458

Figure 4.

Figure 5.

91

March, 2018 of Chilika lagoon (Source: [14]).

period 2001 to 2018.


#### Table 5.

Quarter wise actual versus forecasted shrimps catch with upper 95% catch (MT) and lower 95% catch (MT) prediction error using SARIMA modelling approach for the period 2014 (Q1) to March 2018 (Q4) of Chilika lagoon.

Forecasting Shrimp and Fish Catch in Chilika Lake over Time Series Analysis DOI: http://dx.doi.org/10.5772/intechopen.85458

Figure 4.

e.g., for year 2014 (9.14%) and 2015 (1.02%). The percentage forecast error in shrimp catch for the year 2016, 2017, and 2018 was found to be 4.5, 5.9, and 7.36%, respectively (Table 4). The quarter wise forecast of shrimp catch for the period 2016–2018 with 95% upper and lower catch was shown in Table 5. The quarter wise actual catch versus predicted catch with 95% confidence limit of shrimp catch for the period of 2001 to 2018, showing an increase in catch with respect to 2015, is

Yearly actual versus predicted shrimp catch forecast with % prediction error using SARIMA (0,1,1)(0,1,1)4

Year Actual shrimp catch (MT) Predicted shrimp catch (MT) % error in catch prediction

 1536.418 1676.94 9.14 1535.96 1520.201 1.02 1605.134 4.5 1627.088 5.9 1649.041 7.36

An application of SARIMAX model for total fish production forecasting is analysed for Chilika lagoon [14]. Mullets, clupeids, engraulids, perches, catfishes, sciaenids, threadfins, cichlids, tripod fishes, featherbacks, murrels, carps and

> Forecasted catch (MT)

2014(Q1) 589.113 604.435 795.914 412.956 2014(Q2) 522.827 546.637 738.091 355.183 2014(Q3) 211.991 310.37 501.812 118.929 2014(Q4) 212.487 215.502 436.937 94.0664 2015(Q1) 468.227 567.049 758.48 375.618 2015(Q2) 508.377 482.476 673.891 291.06 2015(Q3) 287.953 243.258 434.666 51.8502 2015(Q4) 271.403 227.418 418.822 36.0139 2016(Q1) 555.174 746.527 363.821 2016(Q2) 519.439 718.554 320.324 2016(Q3) 278.325 484.91 71.7396 2016(Q4) 252.196 465.991 38.4016 2017(Q1) 560.662 794.815 326.509 2017(Q2) 524.927 768.392 281.463 2017(Q3) 283.813 536.246 31.3805 2017(Q4) 257.685 518.778 93.4083 2018(Q1) 566.151 848.368 283.933 2018(Q2) 530.416 823.301 237.531 2018(Q3) 289.302 592.479 183.875 2018(Q4) 263.173 576.304 149.957

Quarter wise actual versus forecasted shrimps catch with upper 95% catch (MT) and lower 95% catch (MT) prediction error using SARIMA modelling approach for the period 2014 (Q1) to March 2018 (Q4) of Chilika

Upper 95% catch (MT)

Lower 95% catch (MT)

shown in Figure 4.

Year (quarterly)

Table 5.

lagoon.

90

model in Chilika lagoon for the year 2014–2018.

Time Series Analysis - Data, Methods, and Applications

Actual catch (MT)

Table 4.

Quarter wise actual versus predicted shrimp catch (MT) with 95% confidence limit of Chilika lagoon for the period 2001 to 2018.

minnows with some miscellaneous catch are the major fish catch of the lagoon with an average annual estimate of fish landing 12,000 tons per annum from 2000–2001 to 2014–2015. The fish landing was fluctuated between 10,286.34 tone (2003–2004) and 6463.92 tone (2006–2007) with average at 7725.42 tone contributing 63.65% of total fish landing of the whole lagoon. Around 64.88% of commercial catch was contributed by the yield of fish since last 15 years of post-restoration period. More than 90% fish species which contributed to the commercial landing of Chilika are of brackish or marine habitats. SARIMAX model was developed using monthly time series fish catch data, and regressor is used as a factor (combination of lagoon

#### Figure 5.

Monthly actual fish catch versus predicted (forecasted) catch by SARIMAX model for the period April 2001 to March, 2018 of Chilika lagoon (Source: [14]).

temperature and salinity for the same period of April 2000–2001 to March 2014– 2015 of the Chilika lagoon). The data sets for the period April 2001–March 2011 were used for training sets and April 2011 to March 2015 were taken as validation of the model. SARIMAX (1,0,0)(2,0,0)12 was found to be the best fitted model. Here, SARIMA (1,0,0)(2,0,0)12 model is also compared with the SARIMAX (1,0,0) (2,0,0)12 (SARIMA (1,0,0)(2,0,0)12 with regressors as factor) and found that the R square value is greater for SARIMAX than SARIMA and root mean square error is less for SARIMAX than the SARIMA model, which reveals that SARIMAX model is better performer than the SARIMA model for the forecasting of the total fish catch of the lagoon. The regressor, i.e., temperature and salinity used for SARIMAX development, showed significantly positive influence (p < 0.05) on the fish catch in the lagoon. Using SARIMAX model, forecasting is done for the fish production within the error catch less than 10% (for details please see [14]; Figure 5). Total fish catch forecasting also showed an increasing catch in the upcoming year till 2018 with respect to the base year 2015.

model, multiple linear regression, non-linear regression and dynamic models, Gaussian autoregressive model, etc. [34–36], but in general, ARIMA models have been widely used for better forecasting. SARIMA model is found good fit for short time series catch data and used as common forecasting model by several researchers [34, 37, 38]. This study also analysed and forecast the quarterly shrimp landing of Chilika lagoon using SARIMA time series modelling approach using data for the period 2001–2015. The best fitted SARIMA (0,1,1)(0,1,1)4 model developed for the shrimp catch in Chilika lagoon was validated with actual quarterly catch data for the period 2014–2015 and further quarter wise forecast of the shrimp catch up to the year 2018. The forecast of shrimp catch for the years 2016, 2017, and 2018 shows an increase in catch by 4.5, 5.9, and 7.36% with respect to the base year 2015 if the environmental condition of the lagoon remains the same. SARIMAX (1,0,0)(2,0,0) 12 also showed the total fish increasing catch up to year 2018 with respect to the base year 2015. These model-based fish catch forecasting for Chilika lagoon results will be useful for understanding fish catch forecast with acceptable accuracy that will enable lagoon fish managers to facilitate fisheries management by predicting the lagoon fisheries progress toward fish production. Reliable fish catch production forecasting methods are important for managers for confident decision making in fishery management. Hence, the study suggests the regular monitoring and maintenance of fish catch and water quality data to be continued for assessing fisheries trends to enable corrective steps for short- and long-term management interven-

Forecasting Shrimp and Fish Catch in Chilika Lake over Time Series Analysis

DOI: http://dx.doi.org/10.5772/intechopen.85458

Time series SARIMA (0,1,1)(0,1,1)4 model and SARIMAX (1,0,0)(2,0,0)12 model are statistical forecasting models for predicting time series quarterly catch of shrimp production and monthly total fish production in Chilika lagoon. Seasons play a significant influence in shrimp and total fish production in the lagoon. Model prediction showed shrimp and total fish catch of the lagoon will increase in the upcoming years by maintaining the present environmental and ecological conditions of the lagoon. Summer and monsoon seasons have explicit influence on shrimp production in the lagoon. Salinity and temperature of the lagoon positively influence on the total fish catch in the lagoon. Model predicts that shrimp and total fish catch will increase in the upcoming years till 2018 with respect to the base year 2015 in the lagoon. Therefore, continuous monitoring in the lagoon is essential to manage its rich and productive fishery resources, as well as its ecological integrity. This information would provide kind support to the managers of Chilika Development Authority to continue the present efforts for getting a good rich production for the livelihood of the fishermen dependent upon and sustainable management of fisheries. This time series model based study will be useful for forecasting fish production in different lagoon or aquatic system. Besides the catch and water quality monitoring, some related information regarding average monthly influx of tides in respect of amplitude and duration may be recorded for correlation with the

catch variability of important species and consideration in trend analysis.

The authors like to acknowledge the Chilika Development Authority (CDA), Bhubaneswar, Odisha, India, for providing time series fish catch data of Chilika lagoon for the present study. We are thankful to ICAR-CIFRI, Barrackpore,

tions in the lagoon.

Acknowledgements

93

5. Conclusions and future work
