To better develop and protect the pelagic fishery in the northwest Indian Ocean, China’s fishing enterprises have been producing pelagic fisheries in the said area for a long time. Based on the fishing log data of light falling gear in the northwest Indian Ocean from 2016 to 2020, this study analyzed the impact of different time scales on the catch rate and fishing ground center of gravity of light falling gear fishing grounds. We also explored the relationship between different time scales and catch per unit effort (CPUE) by using the fishing ground center of gravity, the Random Forest model (RF), and the generalized additive model (GAM). The results were shown as follows: (1) From 2016 to 2020, 76,576 t were captured, and 16,496 nets were operated; (2) The gravity center of fishing ground in the Northwest Indian Ocean moved to the northeast as a whole, and the monthly fishing ground gravity center changed first to the Southern and then to the northern; (3) RF model (R² = 0.709, RMSE = 0.2034, and prediction accuracy is 55.8%), which is better than the GAM model (R² = 0.632, RMSE = 0.2242, and prediction accuracy is 37.3%). In the RF model, the importance of time variables on CPUE was in the order of week, year, operation time, and lunar phase; in the GAM model, it was week, year, lunar phase, and operation time. On the whole, the importance of the long time scale (year, week) is greater than that of the short time scale (lunar phase and operation time). (4) The RF model and GAM model show that the most critical environmental variables were SST, DO, SSS, and Chla, and the least important were SSH, Δ50, and CV50. SST, Chla, and DO significantly impact pelagic fishing and CPUE and are critical reference indexes for predicting the Northwest Indian Ocean light falling gear fishing ground. (5) The 95% confidence interval showed that the suitable interval of time, space, and environmental variables in the RF model was much smaller than in the GAM model.

为更好地开发和保护西北印度洋远洋渔业，我国渔业企业长期以来一直在该海域生产远洋渔业。本研究基于2016-2020年印度洋西北部轻型落架捕鱼日志数据，分析了不同时间尺度对轻型落架渔场捕捞率和渔场重心的影响。我们还通过使用渔场重心、随机森林模型 (RF) 和广义加性模型 (GAM) 探索了不同时间尺度与单位努力捕获量 (CPUE) 之间的关系。结果显示如下：（1）2016年至2020年共捕捞76576吨，运行16496张网；(2)西北印度洋渔场重心整体向东北移动，月渔场重心先南后北变化；（3）RF模型（R²=0.709，RMSE=0.2034，预测精度为55.8%），优于GAM模型（R²=0.632，RMSE=0.2242，预测精度为37.3%）。在RF模型中，时间变量对CPUE的重要性依次为周、年、运行时间、月相；在 GAM 模型中，它是周、年、月相和运行时间。综合来看，长时间尺度（年、周）的重要性大于短时间尺度（月相和运行时间）。(4) RF模型和GAM模型表明，最关键的环境变量是SST、DO、SSS和Chla，最不重要的是 SSH、Δ50 和 CV50。SST、Chla 和 DO 显着影响远洋捕捞和 CPUE，是预测西北印度洋轻型落网渔场的重要参考指标。(5) 95%置信区间表明RF模型中时间、空间和环境变量的合适区间远小于GAM模型。