Data for: Tailored Forecasts Can Predict Extreme Climate Informing Proactive Interventions in East Africa

This perspective discusses new advances in the predictability of east African rains and highlights the potential for improved early warning systems (EWS), humanitarian relief efforts, and agricultural decision-making. Following an unprecedented sequence of five droughts, in 2022, 23 million east Africans faced starvation, requiring >$2 billion in aid. Here, we update climate attribution studies showing that these droughts resulted from an interaction of climate change and La Niña. Then we describe, for the first time, how attribution-based insights can be combined with the latest dynamic models to predict droughts at eight-month lead-times. We then discuss behavioral and social barriers to forecast use and review literature examining how EWS might (or might not) enhance agro-pastoral advisories and humanitarian interventions. Finally, in reference to the new World Meteorological Organization (WMO) “Early Warning for All” plan, we conclude with a set of recommendations supporting actionable and authoritative climate services. Trust, urgency, and accuracy can help overcome barriers created by limited funding, uncertain tradeoffs, and inertia. Understanding how climate change is producing predictable climate extremes now, investing in African-led EWS, and building better links between EWS and agricultural development efforts can support long-term adaptation, reducing chronic needs for billions of dollars in reactive assistance.

Methods
This data set draws from four widely used sources: the Climate Hazard Center Infrared Precipitation with Stations archive (CHIRPS), the NOAA Extended Reconstruction sea surface temperature data set (version 5), seasonal SST forecasts from the North American Multi-Model Ensemble (NMME) and projected SST time-series from Phase 6 of the Climate Model Intercomparison Project (CMIP6). While all of these data are publicly available, we pull together here all the salient time series supporting the basic results of our paper. The NMME seasonal climate forecasts are based on coupled ocean-atmosphere models, intialized monthly with observed conditions. The coupled ocean-atmosphere models in the CMIP6 archive, on the other hand, are initialized in the early 19th century, and then run into the future, constrained by changes in aerosols and greenhouse gasses. The NMME provide operational forecasts. The CMIP6 provides climate change simulations.

The data are organized in a spreadsheet with tabs corresponding to figure panels. 

The Figure 1B tab contains 1981–2022 March-April-May (MAM) and October-November-December (OND) CHIRPS rainfall totals averaged over the eastern Horn of Africa (Ethiopia, Kenya and Somalia east and south of 38E, 8N). This extremely food-insecure area suffers from sequential droughts. There has also been a well-documented decline in the MAM rains beginning around 1999. This tab also contains seasonal totals expressed as 'Standardized Precipitation Index' (SPI) values. These were calculated by fitting a Gamma distribution to the MAM and OND rainfall time-series and then translating the associated quantile values to a standard normal distribution. Seasons with SPI values of less than -0.44Z or greater than +0.44Z fall within the below-normal or above-normal terciles.

The Figure 1E tab contains observed standardized 'West Pacific Gradient' (WPG) and 'Western V Gradient' (WVG) time-series for, respectively, the OND and MAM seasons. These gradients measure the difference between standardized equatorial east Pacific (NINO3.4) and standardized west Pacific SST time series. The data are standardized because relatively small temperature increases in the very warm west Pacific can be dynamically important. The observed gradient values show that warming in the west Pacific, combined with a lack of warming in the NINO3.4 region, has led to large increases in Pacific SST gradients. This sets the stage for sequential droughts in the eastern Horn.

The Figure 1F tab contains Indo-Pacific SST time-series from 152 CMIP6 climate change simulations. These simulations are based on the moderate warming Shared Socio-economic Pathway 245 scenario (SSP245). Time-series are provided for the OND equatorial west Pacific, MAM Western V region, and OND western Indian Ocean region. Observed NOAA SST time series are also provided. The human-induced warming signal is pronounced in the CMIP6 simulations. During the 2016/17 and 2020/2022 La Niña sequences, climate change contributed to exceptionally warm equatorial west Pacific and Western V SST. During the positive Indian Ocean Dipole event in 2019, climate change contributed to exceptionally warm western Indian Ocean SST. The western Indian Ocean region corresponds with the western box used to calculate the Indian Ocean Dipole (IOD). The 2019 IOD event was associated with flooding and a desert locust outbreak. The 2020–2022 period was associated with five sequential droughts in East Africa.

The Figure 2A tab contains observed and predicted 1982–2022 MAM and OND Pacific gradient time series (WVG and WPG). The forecasts are based on six models from the North American Multi-model Ensemble (NMME). The OND forecasts are based on NMME predictions made in May. The MAM forecasts are based on NMME predictions from September. The data have been accessed via the IRI data library. Six individual standardized SST forecasts for the NINO3.4 and west Pacific regions are extracted for each model and then combined using a weighted average proportional to each model's skill (R2). The NINO3.4 and west Pacific SST are then used to calculate the WVG and WPG forecasts. Observed WVG and WPG values are based on NOAA Extended reconstruction version 5 SST.

The Figure 2B tab is very similar to 2A but contains the west Pacific OND and MAM time series. While SST observations and CMIP6 simulations indicate more frequent extremely warm SSTs (tabs 1E and 1F), these can be predicted surprisingly well, offering opportunities to anticipate associated climate extremes.

The Figure 3A tab contains the CMIP6 simulation data supporting panel 3A. The standardized WPG and WVG time series are provided for 152 CMIP6 SSP245 simulations, and the individual changes in event frequencies have been calculated for each simulation. These changes contrast WPG and WVG event frequencies in 2020–2030 versus 1920-1979.  An increase in event frequency is a very robust result, due to the very robust warming in the west Pacific. This latter warming can be verified via the data in the Figure 1F tab if desired. Note that a few CMIP6 models only had one simulation. Results for these models were not listed in the inset in Fig. 3A, due to space limitations.

本综述探讨了东非降雨可预报性领域的最新进展，并着重阐述了其在优化早期预警系统（Early Warning Systems, EWS）、人道主义救援工作以及农业决策方面的应用潜力。在经历前所未有的连续5次干旱后，2022年有2300万东非民众面临饥荒，救援需求超过20亿美元。本文更新了气候归因研究成果，表明此类干旱由气候变化与拉尼娜（La Niña）事件共同作用引发。随后，本文首次阐述了如何将基于归因的认知与最新动态模型相结合，实现提前8个月的干旱预报。随后，本文探讨了预报应用面临的行为与社会障碍，并综述了相关研究，以分析早期预警系统能否提升农牧咨询服务与人道主义干预效果。最后，结合世界气象组织（World Meteorological Organization, WMO）发布的《全民早期预警》计划，本文最终提出一系列建议，以支持可落地且具权威性的气候服务工作。公信力、紧迫性与准确性，有助于克服资金有限、权衡不明及惰性带来的各类障碍。明晰气候变化如何催生当前可预报的气候极端事件、投资由非洲主导的早期预警系统，以及强化早期预警系统与农业发展工作的联动，均可助力长期气候适应，减少长期以来对数十亿美元被动救援的依赖。

研究方法
本数据集源自四类广泛使用的数据源：气候灾害中心红外降水与站点融合数据集（Climate Hazard Center Infrared Precipitation with Stations, CHIRPS）、美国国家海洋和大气管理局扩展重建海表温度数据集（第5版，NOAA Extended Reconstruction sea surface temperature data set version 5）、北美多模式集合预报系统（North American Multi-Model Ensemble, NMME）的季节尺度海表温度预报数据，以及气候模式比较计划第6阶段（Climate Model Intercomparison Project Phase 6, CMIP6）的海表温度预估时序数据。尽管所有数据源均公开可获取，但本文整合了支撑本研究核心结论的所有关键时序数据集。北美多模式集合预报系统的季节尺度气候预报基于海气耦合模式，以实时观测条件为初始场每月开展初始化。而气候模式比较计划第6阶段存档中的海气耦合模式，则以19世纪早期为初始时刻启动模拟，并依据气溶胶与温室气体变化强迫延续至未来时段。北美多模式集合预报系统提供业务化预报产品，气候模式比较计划第6阶段则用于开展气候变化模拟试验。

本数据集以电子表格形式组织，工作表标签与论文附图面板一一对应。
工作表"图1B"包含1981–2022年3-4-5月（MAM）与10-11-12月（OND）的气候灾害中心红外降水与站点融合数据集降雨总量，数据为非洲之角东部区域（埃塞俄比亚、肯尼亚以及东经38°以东、北纬8°以南的索马里）的平均降雨量。这片粮食保障极度薄弱的区域连续遭受干旱侵袭，且有充分文献记录显示，1999年左右开始的3-4-5月降雨呈现显著减少趋势。该工作表同时包含以标准化降水指数（Standardized Precipitation Index, SPI）表示的季节降雨总量。其计算方法为：首先对3-4-5月与10-11-12月的降雨时序数据拟合伽马分布，再将对应的分位数转换为标准正态分布值。当标准化降水指数小于-0.44或大于+0.44时，对应季节降雨分别处于偏少或偏多的三分位区间。

工作表"图1E"包含分别对应10-11-12月与3-4-5月季节的观测标准化西太平洋梯度（West Pacific Gradient, WPG）与西V梯度（Western V Gradient, WVG）时序数据。此类梯度指标表征标准化赤道东太平洋（NINO3.4）与标准化西太平洋海表温度时序数据的差值。对数据进行标准化处理的原因在于：西太平洋暖水区的小幅升温可对大气环流产生重要的动力影响。观测梯度数据显示，西太平洋升温叠加尼诺3.4（NINO3.4）区域升温停滞，导致太平洋海表温度梯度显著增大，为非洲之角东部的连续干旱埋下伏笔。

工作表"图1F"包含152组气候模式比较计划第6阶段气候变化模拟试验得到的印太海表温度时序数据。此类模拟基于中等升温情景共享社会经济路径245（Shared Socio-economic Pathway 245, SSP245）。数据集包含10-11-12月赤道西太平洋、3-4-5月西V区域以及10-11-12月西印度洋区域的海表温度时序数据，同时提供美国国家海洋和大气管理局观测海表温度时序数据。气候模式比较计划第6阶段模拟结果中，人为变暖信号显著。在2016/2017年与2020/2022年连续拉尼娜事件期间，气候变化导致赤道西太平洋与西V区域海表温度异常偏高；2019年正印度洋偶极子事件期间，气候变化导致西印度洋海表温度异常偏高。西印度洋区域对应用于计算印度洋偶极子（Indian Ocean Dipole, IOD）的西侧区域。2019年印度洋偶极子事件与洪涝灾害及沙漠蝗灾暴发相关，而2020–2022年则对应东非连续5次干旱事件。

工作表"图2A"包含1982–2022年3-4-5月与10-11-12月太平洋梯度时序数据（西V梯度与西太平洋梯度）的观测值与预报值。预报结果源自北美多模式集合预报系统的6个模式成员。10-11-12月的预报基于5月份发布的北美多模式集合预报系统预测结果，3-4-5月的预报则基于9月份的预测结果。本数据集通过国际气候与社会研究所（IRI）数据库获取。针对每个模式，提取尼诺3.4区与西太平洋区域的6组独立标准化海表温度预报结果，再依据各模式的预报技巧（决定系数R²）进行加权平均融合。随后利用尼诺3.4区与西太平洋海表温度计算西V梯度与西太平洋梯度的预报值。观测的西V梯度与西太平洋梯度值基于美国国家海洋和大气管理局第5版扩展重建海表温度数据集。

工作表"图2B"与"图2A"结构类似，但仅包含西太平洋区域10-11-12月与3-4-5月的时序数据。尽管海表温度观测数据与气候模式比较计划第6阶段模拟结果均显示，极端暖海表温度事件发生频率有所提升（详见工作表1E与1F），但此类事件的预报效果却出人意料地出色，为提前应对相关气候极端事件提供了可能。

工作表"图3A"包含支撑图3A面板的气候模式比较计划第6阶段模拟数据。提供152组气候模式比较计划第6阶段共享社会经济路径245情景模拟的标准化西太平洋梯度与西V梯度时序数据，并针对每组模拟计算了事件频率的变化量。此类变化量对比了2020–2030年与1920–1979年的西太平洋梯度与西V梯度事件发生频率。由于西太平洋显著升温，事件频率增加是极为稳健的模拟结果。若有需要，该升温趋势可通过工作表"图1F"中的数据验证。需注意的是，部分气候模式比较计划第6阶段模式仅开展了1组模拟，受版面限制，此类模式的结果未在图3A的插图中列出。