当前位置: X-MOL 学术J. Hydrol. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Reconstructed centennial precipitation-driven water storage anomalies in the Nile River Basin using RecNet and their suitability for studying ENSO and IOD impacts
Journal of Hydrology ( IF 5.9 ) Pub Date : 2024-11-04 , DOI: 10.1016/j.jhydrol.2024.132272
Jielong Wang, Joseph Awange, Yunzhong Shen, Ling Yang, Tengfei Feng, Yongze Song

While the Gravity Recovery And Climate Experiment (GRACE) and its Follow-On (GFO) missions have offered valuable observations for monitoring total water storage anomalies (TWSA), their short record constrains our ability to study the complete range and long-term variability of TWSA in the Nile River Basin (NRB). Previous studies reconstructing TWSA in this region either relied on specific hydrological models or did not consider spatial correlations among the TWSA grids. Here, we employ RecNet, a deep learning model capable of providing independent TWSA observations without relying on hydrological models while considering spatial correlations, to reconstruct precipitation-driven TWSA in the NRB from 1923 to 2022. The reconstructed data are validated by comparisons with the Global Land Data Assimilation System (GLDAS), the WaterGAP Global Hydrology Model (WGHM), the water balance budget, long-term runoff data, and GRACE-REC (i.e., a global reconstruction dataset freely available online). Subsequently, the suitability of the reconstructed data for studying El Niño Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) impacts within the NRB is assessed. Dividing the NRB into four sub-regions, i.e., the Lake Victoria Basin (LVB), the Bahr el Jebel and Bahr el Ghazal basins (BJBG), Ethiopian Highlands region (EH), and the Lower Nile River Basin (LNRB), it is shown that RecNet successfully reconstructs precipitation-driven TWSA over BJBG and EH, achieving correlation coefficient (CC), normalized root mean square error (NRMSE), and Nash–Sutcliffeefficiency (NSE) of 0.94/0.11/0.88 and 0.96/0.09/0.91 during the testing period, respectively. Additionally, RecNet’s reconstruction shows better agreement with GLDAS and WGHM than GRACE-REC, correlating well with runoff and the water balance budget in these regions. The relatively poor performance in the LVB and LNRB regions could be attributed to the substantial influence of Lake Victoria and the arid climate, respectively. Correlation analysis and wavelet coherence analysis identify significant coherence between ENSO/IOD and the reconstructed TWSA in BJBG and EH, with CC values of −0.68/0.34 and −0.82/0.56, respectively. This study provides centennial reconstructed TWSA data that could be useful in climate change/variability studies and water resource management within the NRB.

中文翻译:


使用 RecNet 重建的尼罗河流域百年降水驱动的蓄水异常及其研究 ENSO 和 IOD 影响的适用性



虽然重力恢复和气候实验 (GRACE) 及其后续 (GFO) 任务为监测总储水异常 (TWSA) 提供了有价值的观测结果,但它们的短暂记录限制了我们研究尼罗河流域 (NRB) TWSA 的完整范围和长期变化的能力。以前在该地区重建 TWSA 的研究要么依赖于特定的水文模型,要么没有考虑 TWSA 网格之间的空间相关性。在这里,我们采用了 RecNet,这是一种深度学习模型,能够在考虑空间相关性的同时提供独立的 TWSA 观测,而无需依赖水文模型,以重建 1923 年至 2022 年 NRB 中降水驱动的 TWSA。通过与全球土地数据同化系统 (GLDAS)、WaterGAP 全球水文模型 (WGHM)、水平衡预算、长期径流数据和 GRACE-REC(即在线免费提供的全球重建数据集)进行比较来验证重建的数据。随后,评估重建数据对研究 NRB 内厄尔尼诺南方涛动 (ENSO) 和印度洋偶极子 (IOD) 影响的适用性。将 NRB 划分为四个子区域,即维多利亚湖流域 (LVB)、Bahr el Jebel 和 Bahr el Ghazal 流域 (BJBG)、埃塞俄比亚高原地区 (EH) 和尼罗河流域下游 (LNRB),表明 RecNet 成功地在 BJBG 和 EH 上重建了降水驱动的 TWSA,实现了相关系数 (CC)、归一化均方根误差 (NRMSE)、 测试期间的 Nash-Sutcliffeefficiency (NSE) 分别为 0.94/0.11/0.88 和 0.96/0.09/0.91。 此外,RecNet 的重建显示出与 GLDAS 和 WGHM 的一致性优于 GRACE-REC,与这些地区的径流和水平衡预算密切相关。LVB 和 LNRB 区域相对较差的表现可分别归因于维多利亚湖和干旱气候的巨大影响。相关性分析和小波相干性分析确定了 ENSO/IOD 与 BJBG 和 EH 中重建的 TWSA 之间的显著相干性,CC 值分别为 -0.68/0.34 和 -0.82/0.56。本研究提供了百年重建的 TWSA 数据,可用于 NRB 内的气候变化/变率研究和水资源管理。
更新日期:2024-11-04
down
wechat
bug