Multi-temporal variability forecast of particulate organic carbon in the Indonesian seas.

Abstract

The current global condition characterized by high levels of CO2 is altering the carbon cycle and elemental biogeochemistry, resulting in subsequent global warming, climate change, ocean acidification, and the indirect response of deoxygenation. The features of Indonesia's coastal ecosystems and continental shelf waters also contribute to spatio-temporal ocean carbon variability. For instance, the level of particulate organic carbon (POC) will change annually, and thus, over a decadal period, ocean dynamics may affect the temporal variability of POC. Motivated by such conditions, future forecasting is needed to envision the productivity of Indonesian seas by predicting vital parameters such as POC. This research aimed to forecast the temporal variability of POC in Indonesian waters. The Seasonal Autoregressive Integrated Moving Average (SARIMA) forecasting model was used by considering the lowest value of the Akaike information criterion (AIC) and the mean absolute percentage error/MAPE (threshold < 10%). Using the highest correlation coefficient (threshold: 0.75), we obtained the best fit for forecasting POC temporal variability. Hindcast POC data (2002-2020/2021) was used to train the forecasting model. The result shows that forecasting of POC temporal variability can be conducted up to 2030. The validity of prediction is ensured for less than 5years forward after 2020 with correlation coefficients of 0.65 and 0.83 for seasonal and monthly POC, respectively. The hindcast and forecast estimates of POC in the Indonesian seas show a decreasing trend. The present study emphasizes the different forecasting results obtained using the different approaches of annual versus inter-annual variability. A sustained research effort is still required to assess POC forecasting for its potential benefits in marine system monitoring and assessment.