Workshop Report West Africa Regional Training On the Improved NextGen Seasonal Forecasting Approach (PyCPT 2) October 2022 C O N T E N T S To cite this wor kshop report Grossi A, Robertson A, Trzaska S, Dinku T, Zougmoré R, Minoungou B, Mohamed H, 2022. West Africa Regional Training On the Improved NextGen Seasonal Forecasting Approach (PyCPT 2). AICCRA Workshop Report. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). About AICCRA Accelerating Impacts of CGIAR Climate Research in Africa (AICCRA) is a project that helps deliver a climate-smart African future driven by science and innovation in agriculture. It is led by the Alliance of Bioversity International and CIAT and supported by a grant from the International Development Association (IDA) of the World Bank. Explore AICCRA’s work at aiccra.cgiar.org Contact Us Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Email: aiccra@cgiar.org Photos: Amanda Grossi, International Research Institute for Climate and Society (IRI) Disclaimer: This workshop report has not been peer reviewed. Any opinions stated herein are those of the author(s) and do not necessarily reflect the policies or opinions of IRI, donor agencies, or partners. All images remain the sole property of their source and may not be used for any purpose without written permission of the source. This workshop report is licensed under a Creative Commons Attribution – NonCommercial 4.0 International License. © 2022 International Research Institute for Climate and Society, Columbia Climate School AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 2 Abstract From October 10-19, a nine-day training targeting West Africa (WA) was implemented in Lomé, Togo by the International Research Institute for Climate and Society (IRI) of the Columbia Climate School, in close collaboration with the AICCRA-West Africa team, the Regional Center for Training and Application in Agrometeorology and Operational Hydrology (AGRHYMET) and Meteo Togo. The workshop, which was organized as part of the World Bank’s Accelerating the Impact of CGIAR Climate Research for Africa (AICCRA) project, brought together 7 national meteorological services from the WA region, as well as its regional climate center (AGRHYMET) to improve seasonal forecasting capacities using the “NextGen” approach and its concomitant PyCPT version 2 interface (PyCPT2). In particular, the major objectives of the training were to strengthen the knowledge and understanding of national meteorological services of seasonal forecasting tools, introduce the new advances and functionalities of the Python (PyCPT2) interface for the NextGen forecasting approach, configure and run PyCPT version 2 to make the best- available forecasts in participants’ home countries, including forecast verification, and provide foundational training on best practices for forecast communication including the flexible forecast format.. Keywords West Africa; forecasting; agriculture; climate change; climate variability; capacity development; food security; Goal 2 Zero Hunger AICCRA Report title here • 3 C O N T E N T S About the Authors Amanda Grossi is a Senior Staff Associate at the International Research Institute for Climate and Society (IRI) of the Columbia Climate School. Within the AICCRA project, she is the IRI’s Regional Manager for Africa where she coordinates the IRI’s activities at the country-level in Ethiopia, Kenya, Zambia, Ghana, Mali, and Senegal. In this role, she provides critical support to the development and delivery of capacity building initiatives and digital innovations, including those associated with the IRI’s Enhancing National Climate Services (ENACTS) approach. Andrew Robertson is a Senior Research Scientist and head of the Climate Group at the International Research Institute for Climate and Society (IRI) of the Columbia Climate School. He is also an adjunct professor in Columbia Climate School where he teaches in the Climate and Society MA Program. Sylwia Trzaska is a Senior Staff Associate and climate scientist at the International Research Institute for Climate and Society (IRI) of the Columbia Climate School. Tufa Dinku is a Senior Research Scientist at the International Research Institute for Climate and Society (IRI) of the Columbia Climate School. Within the AICCRA project, he is the IRI’s Team Lead for Ethiopia, Kenya, Zambia, Ghana, and Mali and also the lead for the IRI’s Enhancing National Climate Services (ENACTS) initiative which has improved the availability, access, and use of climate data and information in more than 20 countries. Robert Zougmoré is the West Africa Lead of the AICCRA project, based at the Alliance of Bioversity International and CIAT in Dakar, Senegal. Dr. Zougmoré is an agronomist and soil scientist with a PhD in Production Ecology & Resources Conservation. Bernard Minoungou is a hydrological forecaster and modeler at AGRHYMET based in Niamey, Niger. Hamatan Mohamed is a hydrologist at AGRHYMET in based in Niamey, Niger. AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 4 Acknowledgments The Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA) project is supported by a grant from the International Development Association (IDA) of the World Bank. IDA helps the world's poorest countries by providing grants and low to zero-interest loans for projects and programs that boost economic growth, reduce poverty, and improve poor people's lives. IDA is one of the largest sources of assistance for the world's 76 poorest countries, 39 of which are in Africa. Annual IDA commitments have averaged about $21 billion over circa 2017-2020, with approximately 61 percent going to Africa. The West Africa Regional Training On the Improved NextGen Seasonal Forecasting Approach (PyCPT 2) and its associated teaching and learning materials represents a collaborative effort, made possible by the IRI’s dedicated team of climate, sectoral, and technical experts including Kyle Hall who had the vision and meticulous dedication to significantly improve the structure, functionality, and ultimate usability of PyCPT 2 alongside its integration with wider systems to position it as an open-source, community-owned and collaborative interface and approach for seasonal forecasting. It also rests firmly on the foundations and achievements by the late Lisa Goddard who led global efforts to advance near-term forecasting with relentless determination and unshakeable passion. The scale-up of the critical capacity building needed to expand the use of this high-need approach could not have been accomplished without the consistent and generous support of the Regional Center for Training and Application in Agrometeorology and Operational Hydrology (AGRHYMET) and the AICCRA- West Africa team, which co-organized the training and continuously recognizes the importance of cross- country and peer-to-peer exchanges within the region of West Africa for advancing the use and sustainability of digital innovations such as NextGen/PyCPT 2. It could also not have been accomplished without the leadership and support of Dr. Robert Zougmoré of the Alliance of Bioversity International and CIAT and the West Africa Lead for the AICCRA project, who has recognized a common need for improved risk management through forecasting that extends beyond just AICCRA project core project countries, and the dedicated AGRHYMET staff such as Bernard Minoungou, Hamatan Mohamed, and Yacouba Ayouba who coordinated and managed logistics for a complex, multi-country training with grace and efficiency during an ongoing global pandemic. Last but certainly not least, a debt of gratitude is owed to the Meteo Togo for its unmatched hospitality in hosting this regional training, and its enthusiasm to carry this work forward. AICCRA Report title here • 5 C O N T E N T S Highlights 1 2 3 Enhancing the generation of From October 10-19, 2022, a A total of 24 meteorologists from 7 high-quality climate nine-day training of countries and the Regional Center information to anticipate, trainers (ToT) on the topic for Training and Application in manage, and respond to of NextGen seasonal Agrometeorology and Operational climate-related disasters and forecasting with an easy-to- Hydrology (AGRHYMET) were longer-term climate change is use Python interface to the successfully capacitated on new critical for building systems- Climate Predictability Tool advances and functionalities of the level resilience in region of (CPT) called “PyCPT 2” was NextGen/ “PyCPT 2” interface and West Africa. Capacity building conducted in Lomé, Togo. approach for seasonal forecasting, as on best-available seasonal Beyond generation of the well as best practices in producing, forecasting techniques forecast, the training also tailoring, and communicating demand- through the NextGen or emphasized best practices driven, decision-relevant forecasts. “PyCPT 2” approach is for forecast communication Detailed feedback via targeted necessary towards these ends. and tailoring for different stakeholder interviews and a plenary users. session was also gathered. 4 5 6 Through the CGIAR G The national and regional The training had The NextGen approach to meteorological service staff will participation from seven seasonal forecasting enables share the knowledge, skills, national meteorological location-specific, objective, high- and resources gained from services and AGRHYMET quality, and decision-relevant the regional training with their which are important for seasonal forecasts, amongst other colleagues in their home extending the reach of applications. Follow-up trainings countries and regions, following best-available seasonal are recommended on seasonal the training of trainers (ToT) forecasting approaches forecasting with specific attention approach. Communication to the most local levels to agricultural parameters, channels have been created to and integrating them within subseasonal forecasting, and foster ongoing collaboration decision-making processes. Python basics to sustain local and a community of practice capacity and locally-led adaptation. amongst participants for improved seasonal forecasting. AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 6 INTRODUCTION 1: Introduction Enhancing the generation of high-quality climate While there are many advantages of the information is critical for anticipating, managing, NextGen forecasting system in terms of and responding to climate-related disasters and improving the tailoring, communication, and longer-term climate change in the region of West ultimate usability of information, its enormous Africa. contribution has been in transitioning countries from subjective consensus forecasting to Seasonal forecasts in particular can help to operational objective, reproducible climate minimize and manage climate risk in the information that takes advantage of the latest agricultural sector and especially productivity available dynamical models. risks at the farm level by informing key management decisions such as planting and It also notably and uniquely puts user demand at harvesting timing, crop and cultivar choice (long the core of the design of the forecast system. versus short cycle varieties, depending on The demand defines the predictand(s) to focus rainfall, for example), timing of fertilizer or on, which in turns defines which processes or pesticide application, and much more (Klemm, mechanisms control the behaviour of the 2017). predictands, and leads to the identification of which predictor(s) to use. This same demand- The NextGen forecasting system, which is driven workflow is used to identify which forecast based on more than 25 years of research and attributes and skill metrics are needed, which in now implemented in 6 countries and two turns defines which calibration methods to regional climate centres, helps forecasters to explore, and how to present the predictions (IRI, select the best climate models for any area of 2022). interest through a process-based evaluation. To produce these forecasts, an easy-to-use Moreover, it automates the generation and Python interface to the Climate Predictability verification of tailored predictions at multiple Tool (CPT) called PyCPT is used. Under timescales (weeks to months) and at multiple continuous development, this set of libraries levels—regional, national, or even sub-national enables the user to run CPT via Python scripts (Columbia University, 2020; Acharya et al., and Jupyter Notebooks, helping automate and 2021). mass-produce CPT tasks that will normally take more time in the Windows version. PyCPT was designed to implement the next-generation of seasonal forecasts for climate services at Participants of the October 2022 West Africa Regional Training National Meteorological and Hydrological On the Improved NextGen Seasonal Forecasting Approach (PyCPT 2) pause for a group photo at the Hotel La Concorde Services (NMHSs) in the Global South. in Lomé, Togo. To capacitate NMHSs in the region of West Africa as well as its regional climate centre (AGRHYMET) on the use of this Python interface for producing forecasts, the Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA) project conducted a nine-day training on PyCPT in January 2022 with representatives from five meteorological services across West Africa and AGRHYMET (Minoungou et al., 2022). However, since this time, new advances in the Python interface and functionalities have resulted in updates to the NextGen forecasting approach (PyCPT 2) that necessitated additional training of the national meteorological services and their associated regional climate centres (RCCs) such as AGRHYMET. These advances include improvements on ease of installation, AICCRA Report title here • 7 INTRODUCTION customization, readability, reproducibility, for the NextGen forecasting approach, configure flexibility, documentation, and compatibility with and run PyCPT version 2 to make the best- other systems and Python packages such as available forecasts in participants’ home Xarray. These improvements have now countries and regions, including forecast positioned PyCPT 2 as an open-source, verification, and provide foundational training on community-owned and collaborative interface good practices for forecast communication such and approach for seasonal forecasting. as the Flexible Forecast format. Thus, beyond just generating high-quality seasonal forecasts, To raise awareness and capacity on this new the training also strongly emphasized best and improved version of PyCPT (PyCPT 2), a practices for tailoring and communicating them nine-day training of trainers (ToT) targeting West based on user demands and needs. Africa was implemented in Lomé, Togo from October 10-19, 2022, by the International While the NextGen approach can generate both Research Institute for Climate and Society (IRI) seasonal and subseasonal forecasts to provide of the Columbia Climate School, in close weather-to-climate prediction products that are collaboration with the Regional Center for more seamless, the October 2022 training Training and Application in Agrometeorology and focused exclusively on building capacity around Operational Hydrology (AGRHYMET). The seasonal forecasting. workshop brought together seven national meteorological services from the West African The participating countries and RCCs are region, as well as its regional climate centre summarized below: (AGRHYMET) to improve seasonal forecasting capacities using the “NextGen” approach and its concomitant PyCPT version 2 interface (PyCPT List of Participating Countries and 2). Trainees No. Country Numb In particular, the major objectives of the training er of were to strengthen the knowledge and 1 Ghana 3 understanding of national meteorological *AICCRA core country services of seasonal forecasting tools and 2 Mali 3 approaches, introduce the new advances and functionalities of the Python (PyCPT2) interface *AICCRA core country 3 Senegal 3 *AICCRA core country The West Africa regional training on PyCPT 2 provided 4 Burkina Faso 3 not only critical knowledge and skills to produce the best- 5 Niger 3 available seasonal forecasts, but also an important forum for networking and peer-to-peer exchanges amongst 6 Nigeria 1 participating countries. Pictured here, participants from 7 Togo 3 Mali and Senegal discuss Python code to generate the 8 AGRHYMET (West Africa) 5 seasonal forecast. TOTAL 24 Towards the project’s aims of advancing gender equity and equality through participation of women, at least one of the nominees per participating country was requested to be female. Participants who had previously been trained on PyCPT 1 in January 2022 were also requested to attend where possible. A full list of participants and their affiliate institutions can be found in Box 1, while the list of trainers and support staff can be found in Box 2. The full agenda for the workshop can be found in Section 6 (Agenda). AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 8 INTRODUCTION It is also important to note that while the NextGen or “PyCPT 2” approach was used to predict scientific factors related to the seasonal forecast during this training, it can also be used for prediction in a number of applications and sectors, including:  Aedes-borne (mosquito-borne) diseases  Malnutrition  Fires (subseasonal)  Coffee yields  Subseasonal rainfall These applications of the NextGen approach in various contexts and sectors around the world, which participants were also exposed to during the training, are summarized in the research referenced under the section titled Further Reading at the end of this report. Readers are encouraged to explore these further. Participants listen to a presentation from the IRI’s Sylwia Trzaska giving an overview of the advantages of the Flexible Forecast format for communicating the seasonal forecast. AICCRA Report title here • 9 APPROACHES AND METHODS 2: Approaches and Methods The workshop aimed to train forecasters from home countries or regions, including national and regional meteorological services in forecast verification; and West Africa to, in turn, train forecasters at their home institutions (and in the case of 4) Introduce participants to the Flexible AGRHYMET, their home RCC), in best-available Forecast format (and example ENACTS seasonal forecasting approaches through the Maprooms) and best practices for NextGen (PyCPT 2) approach. forecast communication and tailoring. Towards this end, the workshop’s stated objectives were to: Throughout the workshop, a training Wiki page (website) was kept up to date with all presentations, recordings, exercises, and 1) Strengthen the knowledge and supplemental resources such as recommended understanding of national reading (IRI, 2022b). meteorological services and RCCs of seasonal forecasting tools; Because of the skills-oriented, practical nature of the training, while there were lecture and 2) Introduce the new advances and theoretical components incorporated in the functionalities of the Python interface workshop, a stronger emphasis was placed on (PyCPT version 2) for the NextGen interactive discussion, as well as group and forecasting approach; individual work to gain hands-on experience with the PyCPT 2 interface and code. 3) Configure and run PyCPT version 2 to make the best forecasts in participants’ Sylwia Trzaska from the IRI helps participants from Mali to install and configure the PyCPT package on their computers. One of the advantages of PyCPT 2 over its predecessor of PyCPT 1 is an easier installation process. AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 10 KEY RESULTS AND FINDINGS 3: Key Results and Findings All 24 participants from the NMHSs and This was done in order to capture more AGRHYMET were successfully capacitated on structured and nuanced perspectives on the how to generate and communicate the seasonal value of the training and thoughts on forecast for their home country or region using implementation going forward following the ToT. the PyCPT 2 approach. This was evidenced in a culminating group project and presentation Findings from both of these feedback whereby groups from each country or RCC mechanisms is summarized and synthesized presented the seasonal forecast generated below. using the PyCPT 2 approach. Presentations from each country or RCC can be found here. While in-depth oral feedback on the training was collected during a plenary session on the last day of the workshop, stakeholder interviews were also conducted with at least one person from each of the participating seven NMHSs and AGRHYMET during the last two days of the training. “PyCPT makes our work with seasonal forecasting much easier and simpler. Before, it would take us many days and even weeks— checking and running models one-by-one, and then putting them together. It was really a team effort. Now it’s just a click to run the models. And not only this, you can see visually and very clearly all of them and make a judgment.” —Francisca Martey, Deputy Director, Ghana Meteorology Agency (GMet) AICCRA Report title here • 11 K E Y R E S U L T S A N D F I N D I N G S Findings from Key Stakeholder Interviews forecast in a timelier manner. This was indicated with Participants as important for more regular contexts such as seasonal agricultural planning for farmers, but also in emergency contexts or that of climate The objectives of the interviews carried out with extremes such as droughts of floods. each of the NHMSs and AGRHYMET were as follows: As Bello Ahmad of the Nigeria Meteorological Agency (NiMet) explained, “Nigeria is a very big 1. Obtain a brief explanation of the value country with a population of 200 million people— of the PyCPT 2 methodology and a very wide and diverse people. Almost all the training, in participants’ own words. sectors of the Nigerian economy rely on seasonal forecasting to prepare and plan early 2. Capture participant feedback about for their activities. Most notably, the agricultural which aspects and parts of the training sector and that of water resources rely on this to participants found most/least useful; properly plan for what’s to come.” 3. Capture participant feedback on any He went on to explain the forecast’s value challenges for implementing the PyCPT especially in times of climate extremes, which 2 methodology in their own are becoming more frequent on the continent. meteorological services going forward, “Just recently, we have been having high- and thoughts on next steps for intensity rainfall. People have lost lives, property, operationalization of its use nationally and even farms. Our Director General came out and/or sub-nationally; to say that 2022 has been the worst year in terms of extreme flooding. 31 out of 36 districts 4. Capture participants’ perspectives on (almost 90%), and 1.4 million people have been specific aspects of the training, affected by extreme flooding. It’s important to including the importance of user-centred have good information and early warning so we and tailored communication of the can act before this comes.” Bello also explained forecast and the flexible forecast format. that this kind of flooding also disrupts transport, including of inputs and supplies for agriculture, which was the case this year due to flooding on Importance of NextGen/PyCPT 2 (Context and the Niger River. When it comes to early warning Need for PyCPT2 Seasonal Forecasting and early action for emergencies, Bello and Methodology) others stressed, having the forecast out even a little bit earlier can make all off the difference. Timeliness of the Forecast Francisca Martey, Deputy Director from the All 8 participants interviewed indicated that the Ghana Meteorological Agency (GMet) echoed NextGen/PyCPT 2 forecasting approach was of this reality in her own country. “The seasonal great value to them, though the aspects forecast is used for planning for individuals, highlighted differed. Almost all participants municipalities, and even the government as a emphasized the time-saving aspect of the whole,” she explained. It is used for planning PyCPT 2 approach due to its enabling of everywhere in Ghana—by agriculturalists, forecasters to quickly assess past model hydrologists, health workers, and even for performance, and then correct and combine tourists—especially for managing risk. Will there different global climate models to come up with be flooding? Will we get enough rains for verified and tailored predictions at the seasonal agriculture or electricity? Should people around timescale. rivers or dams be evacuated if spillage is expected? What will the government do?” While some participants noted that this time- saving aspect was very important on the supply She also extolled the time-saving benefits side given that their NMHS (like many in Africa) alongside ease of visualization and what this has tends to be understaffed and that this efficiency meant for her team at GMet. “PyCPT makes our would allow them to focus their energy on other work much easier and simpler,” she explained. important things like model development or “Before, it would take us many days and even research for their countries, others underscored weeks—checking and running models one-by- that this time-saving aspect would have benefits one, and then putting them together. It was for the end user in that they could receive the really a team effort. Now it’s just a click to run AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 12 K E Y R E S U L T S A N D F I N D I N G S the models. And not only this, you can see Objectiveness, Quality, and Reproducibility of visually and very clearly all of them and make a the Forecast judgment.” Some participants highlighted that the improved Location-Specific, Tailored Forecasts documentation through PyCPT 2 was very valuable to them due to the reproducibility and replicability of the forecast arising from a more All participants noted that the ability to have systematic and streamlined process. location-specific forecasts was critical for agricultural planning and disaster risk management (DRM), and many noted that As Bello Ahmed of the Nigerian Meteorological having spatially and temporally complete Agency described, “PyCPT 2 has made life datasets in their countries (4 km resolution) easier; it has made forecasting easier. If you through the Enhancing National Climate compare it to the previous version, you’ll see Services (ENACTS) initiative and data that things have been made very flexible and (Nsengiyumva et al., 2021) was important customized so that you can concentrate on towards this end. producing the actual forecast. You just have to tell it the models you want, and the data comes straight from the IRI Data Library. Apart from From a regional perspective, forecasts tailored that, on assembly, PyCPT 2 allows you to for specific locations and users has been assemble the forecast very easily and also extremely valuable. As Bernard Minoungou of assess the skill [quality] easily. This is a very AGRHYMET attests, “The value of PyCPT at our important aspect of the PyCPT tool and level has really been the higher-resolution approach: With that, we can make better and forecasts, at a scale of 4-5 kilometres, that more skillful forecasts using PyCPT2.” enable us to give location-specific information. This is very important for many applications but especially in agricultural decision-making.” “The value of PyCPT at our level has really been the higher- resolution forecasts, at a scale of 4-5 kilometres, that enable us to give location- specific information. This is very important for many applications but especially in agricultural decision- making.” —Bernard Minoungou, Hydrological Modeller, Regional Centre for Training and Application in Agrometeorology and Operational Hydrology (AGRHYMET) AICCRA Report title here • 13 K E Y R E S U L T S A N D F I N D I N G S The documentation enabled by a systematic Value of Improved Communication and Tailoring forecasting approach helps to create historical of the Forecast, including the Flexible Forecast climate records and datasets that ultimately lead Format to improved predictions of various climate events. Beyond the generation of high-quality, user- centred, objective seasonal forecasts, the Transparency and Customizable Forecasts training also emphasized the decision-relevance of forecasts and the importance of various communication channels, terminology, Another positive aspect brought up by approaches, and formats for ensuring this. participants was that the PyCPT 2 is not a black box; it is a transparent and open-source tool and approach that can be modified and customized One of these approaches is the Flexible according to needs. Forecast presentation that aims to overcome obstacles to using seasonal climate forecasts for decision making. The Flexible Forecast is an “The fact that we are now able to move from online presentation that rectifies the main CPT to have the flexibility to use multiple models criticisms of the tercile convention by presenting and to improve on the tercile forecast with the downscaled forecasts as full probability probabilistic one (the Flexible Forecast) is a big distributions in probability-of-exceedance format improvement,” said Kofi Opoku Nana of the along with the historical climate distribution Ghana Meteorological Agency (GMet). (Hansen et al., 2022). “The Flexible Forecast can help answer many questions at the local level. This aspect of PyCPT 2 has a bearing on the development of the country and how information can be best communicated in the future. It is very valuable.” — Lamine Diop, Meteorologist, National Agency of Civil Aviation and Meteorology of Senegal (ANACIM) AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 14 K E Y R E S U L T S A N D F I N D I N G S As Bernard Minoungou of AGRHYMET collaboration on common issues. So these kinds explained, “Besides high-resolution, location- of initiatives are really encouraged.” specific forecasts, the other major advantage of PyCPT is that it gives the flexibility and Almost all interviewed participants were able to possibility to present the forecast in terms of cite specific peers that they were able to learn exceeding or not exceeding a certain threshold from, form new or strengthened relationships of rainfall, which is important for many different with, or by whom they were personally inspired kinds of users.” or motivated. Value of Peer-to-Peer Learning and South-South “I appreciated the collaborative nature of this Collaboration training,” explained Afoussatou Diarra Ba. “Because while our countries are very different, All participants interviewed extolled the benefits we have common challenges and similar climatic of conducting the NextGen/PyCPT 2 training as impacts, and similar predictors. If we didn’t do a region rather than individually with each the training together, we would not see this.” national meteorological service, citing the benefits of networking, building an in-region Expanding and Improving Use of community of practice and capacity, and peer-to- NextGen/PyCPT 2 peer exchanges. Some participants also noted a better ability to focus fully on the training topic, The interviewees from each of the 7 NHMS and because when trainings are held in their home RCC were asked how their institutions could be office or city, they are often called back for day- better supported to expand the use of the to-day activities or distracted by regular duties. NextGen/PyCPT 2 approach. And, because the workshop was a ToT, each was asked if they On the pedagogical benefits, Bernard foresaw any challenges or barriers to Minoungou of ICPAC attested, “The regional implementing and cascading the training to their approach to training is good because there are colleagues in their home countries. people with many backgrounds and levels of expertise. Doing the training together helped us to share experiences and encourages “I appreciated the collaborative nature of this training. Because while our countries are very different, we have common challenges and similar climatic impacts, and similar predictors. If we didn’t do the training together, we would not see this.” — Afoussatou Diarra Ba Meteorologist, Mali-Météo AICCRA Report title here • 15 K E Y R E S U L T S A N D F I N D I N G S All participants indicated that they felt confident 3. Need for foundational training in Python to cascade the training as a ToT to their home countries or regions. Participants insisted that basic training in Python was necessary for true Findings from Plenary Feedback Session ownership of the NextGen/PyCPT system and approach. Many During the plenary feedback session on the final participants lauded the software for its day of the workshop, it was noted that only 13 of flexibility and the ability of forecasters the 24 participants (just over half) were actually like themselves to modify the code and familiar with Python before the training, and only put in their own thoughts, ideas, and one considered himself an expert. This made it functions, as well as customize the all the more impressive that all participants were outputs such as with the colours. able to ultimately produce their seasonal However, participants generally felt that forecasts using PyCPT 2 by the end of the having more formal proper training in training, largely due to peer-to-peer support. Python would help them to maximize their use of the approach and be able to In terms of countries with high-resolution, not just use the code already created for spatially and temporally complete datasets them but to also create their own code through ENACTS, Nigeria and Burkina Faso to meet the needs of their users. A 3-7 were the only two countries in the training that day introductory training on Python was did not have ENACTS data at national level therefore requested, possibly as an (though they are included in regional add-on to one of the 2023 trainings to implementation). However, for the rest of the minimize costs of convening. It was countries that did have ENACTS data, new noted that foundational skills in Python Python script is still needed to be able to would help the forecasters with PyCPT incorporate and run it with PyCPT. but also other roles and responsibilities at their NHMS, as this is becoming a Overall, while there was rich and nuanced more in-demand language and skill. feedback from the plenary discussion, the following four ideas and requests were echoed “It’s a pity that more people don’t have by most participants and came through most more knowledge of Python,” explained strongly: Boureima Salifou Soumaila of National Meteorological Directorate of Niger 1. Need for a follow-up NextGen seasonal (DMN). “As scientists, we need to be up forecasting training in 2023 focusing on to date on these important languages.” agricultural parameters 4. Need to share the PyCPT approach Participants overwhelmingly requested more widely that the PyCPT 2 package be updated to include important parameters that are Participants noted that while the training relevant for the agricultural sector, included 7 countries and AGHRYMET, including onset, cessation, length of the there are 17 in the region of West season, number of wet days, and Africa, and therefore people need to be number of dry days. thinking creatively and collaboratively on how to best raise awareness and 2. Need for a follow-up training NextGen skills on this approach beyond their sub-seasonal forecasting training in countries. One suggestion toward doing 2023 this was to include a session on this topic at the regional climate outlook Participants were unanimous in the forum in West Africa known as need for NextGen/PyCPT forecasting at Prévisions Climatiques Saisonnières en the sub-seasonal timescale (2-4 Afrique Soudano-Sahélienne weeks), and participants from most (PRESASS). countries indicated that the demand from their users is for forecasts at 5. Need for ongoing collaboration amongst monthly timescales. the trainees and a community of practice around best-available AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 16 K E Y R E S U L T S A N D F I N D I N G S forecasting approaches and techniques address will go to all colleagues present through NextGen/PyCPT. at the training.  An accompanying Google Amongst the goals of West Africa Group (Click here) was created, where regional training on improved seasonal any participant feel free to start forecasting approaches through conversations about any topics they PyCPT2 is to create a strong network of choose. actors within Africa with equally strong capacities to advance the generation of high-quality climate information and  Participants were encouraged to sign up services. Participants of the training are for GitHub by clicking here, join the IRI therefore encouraged and expected to PyCPT project group, and then use act as ambassadors and resources of the Discussion Board functions this PyCPT 2 approach within their own under “Issues” on the top menu bar to countries, but others as well. To do so, raise any issues or technical problems the connections made during the they may have as they implement training must be sustained, cultivated, PyCPT2. The IRI team encouraged and even expanded. participants to interact with each other on this board to help each other solve In an effort to support the continued issues first amongst each other (peer- interactions of this group even after the to-peer) before coming to the IRI. This conclusion of the training, therefore, the is because the AICCRA project like to AICCRA-West Africa team in build a collaboration with the IRI implemented strong community of practice and the following actions, shared with capacity within the East and Southern participants on the last day of the Africa regions, though it will assist in training: answering questions there as they arise as well.  An email list (pycpt- wa@iri.columbia.edu) for all participants  Participants were encouraged to was created. An email sent to this email continue accessing and even sharing the training resources found on the Wiki “I appreciate that IRI and AICCRA have tried to create a community around PyCPT 2 seasonal forecasting that will allow us to do so much. The community that has been created around this tool will really allow it to flourish. I would not be surprised to see a lot of people around this table really leading PyCPT on this continent. I really see it as ours now.” —Bello Ahmed, Meteorologist Nigerian Meteorological Agency (NiMet) AICCRA Report title here • 17 K E Y R E S U L T S A N D F I N D I N G S page for the training here. This serves Since the training has concluded, these digital as a hub for all presentations, exercises, communities of practice have been used by research, and other materials shared participants to connect with each other to help during the training. with any challenges that have arisen during implementation, and especially to share new  Shortly after the training, an online Users’ Guide to PyCPT2 for Seasonal developments or answer any questions Climate Forecasts was shared with participants may have relating to Python codes. participants as a digital resource. For the PyCPT email list for West Africa connecting all participants, Bernard Minoungou and Mandela Houngnibo from the Regional Center for Training and Application in Agrometeorology and Operational Hydrology (AGRHYMET) were designated as focal points, whereas for the sister East and Southern Africa (ESA) group, Tamirat Bekele of the International Livestock Research Institute (ILRI) was previously designated as focal point to moderate and answer questions. Both focal points from ESA and WA respectively sit within each other’s email lists and Google Groups to promote cross- pollination of ideas and learning between regions and across the continent. Amadou Diakite of Mali-Météo begins his presentation on skill maps for the forecast that he just generated for the first time using the PyCPT 2 approach. At the end of the training, each country had to present on the forecasts and models developed using the PyCPT 2 approach, as well as its implementation and automation plans going forward. AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 18 ** 4: Conclusions and Recommendations While all participants were able to successfully generate the forecast using the new and improved approach for their respective countries or regions, and a strong network of actors to advance the generation of high-quality climate information and services was reinforced, additional activities and actions are recommended following in-depth feedback and consultations with participants to sustain these activities and promote the co-generation of high- quality forecasts within the region: 1. A follow-up NextGen seasonal forecasting training focusing on agricultural parameters should be pursued in early 2023 as a region. This training should build upon the previous two NextGen seasonal forecasting trainings to build capacities on seasonal forecasting as it relates to important parameters for the agricultural sector, including rain onset, cessation, length of the season, number of wet days, and number of dry days. 2. A follow-up training focusing on NextGen/PyCPT forecasting at the sub- seasonal timescale (2-4 weeks) most relevant for agricultural decision-making The female forecasters who took part should be pursued in 2023 as a region. in the West Africa Regional Training on 3. Either separately or as an add-on to one the NextGen (PyCPT 2) approach pause of the two aforementioned for a photo at the end of the workshop. recommended NextGen regional Ensuring women have equal access to forecasting trainings for 2023, a technical education and capacity minimum 3-day basic training in Python building is an essential catalyst to should be pursued. This training is advancing climate services that serve necessary to help participants to everyone. maximize their use of the approach, and more specifically to be able to not just use the code already created for them but to also create their own original The nine-day training of trainers (ToT) targeting code to meet the needs of their users, in East and Southern Africa capacitated 24 line with the community-owned and participants representing 7 national generated vision for NextGen. meteorological services from the West Africa 4. Participants will explore ways to share region, as well as its regional climate centre the NextGen/PyCPT 2 approach within (AGRHYMET), with high-need knowledge and their own networks, including at the skills to improve their seasonal forecasting regional climate outlook forum in West capabilities using the NextGen or “PyCPT 2” Africa known as Prévisions Climatiques approach. Saisonnières en Afrique Soudano- Sahélienne (PRESASS). AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 20 ** As participants continue to navigate through Participants are also encouraged to maximize generating the seasonal forecast using the use of community communication channels NextGen/PyCPT 2 approach in their home created to continue fostering collaboration and countries and regions, and as they cascade the in-region capacity: knowledge and skills gained at the regional training to their own meteorological service staff  The email list (pycpt- in line with the ToT approach, they should wa@iri.columbia.edu); an email sent to ensure to leverage and maximize the use of the this email address will go to all teaching and learning materials made available colleagues present at the training. to them, and to which they can actively contribute. These include:  An accompanying Google Group (Click  The Wiki page for the training here, here), where any participant feel free which serves as a hub for all to start conversations about any topics presentations, exercises, research, and they choose. other materials shared during the training that participants can refer to  The IRI PyCPT project group on Github and guide other trainees towards. here, with the accompanying  The Users’ Guide to PyCPT2 for Discussion Board here (“Issues”) to Seasonal Climate Forecasts, a raise any issues or technical problems comprehensive but living digital related to the code they may have as resource and how-to guide for they implement PyCPT2. installing, understanding, and using PyCPT 2. “In many countries, we don’t see women at technical meetings like this. It was wonderful to see so many women present at this training. I thank IRI and AICCRA for that.” —Fatoumata Sangho Diabate, Meteorologist, Mali-Météo **************************************************************** ** Use of these resources and communication channels is intended to foster a strong regional community of practice for promoting the use of best-available forecasting approaches through NextGen/PyCPT 2. Similar digital resources and a sister community of practice has also been set up for the East and Southern Africa region following their own NextGen/PyCPT 2 training, and focal points for the digital resources sit within each other’s groups to foster cross-continental collaboration. “I appreciate that IRI and AICCRA have tried to create a community around PyCPT 2 seasonal forecasting that will allow us to do so much,” said Bello Ahmed of the Nigerian Meteorological Agency, capturing the sentiment of the group. “Before, PyCPT was so closed and limited in what we could do in terms of expansion.” “But now, we have the opportunity to do whatever we want,” he said. “The community that has been created around this tool will really allow it to flourish. We have seen a lot of software that has been developed beyond imagination due to the communities around them. I would not be surprised to see a lot of people around this table really leading PyCPT on this continent. I really see it as ours now.” AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 22 ** 5: List of Participants and Trainers Box 1 List of Trainees No. Name Gender Organization/ Contact Structure 1 Thomas Bere M ANAM (Burkina Faso) bererthomas@gmail.com 2 Boukaré Compaore M ANAM (Burkina Faso) compaoreboukare@gmail.com 3 Asséta Bougma F ANAM (Burkina Faso) sikireasseta@gmail.com 4 Fatoumata Sangho Diabate F Mali-Météo (Mali) mdsangho80@gmail.com 5 Afoussatou Diarra Ba F Mali-Météo (Mali) afoussatou@gmail.com 6 Amadou Diakite M Mali-Météo (Mali) amadiakite094@gmail.com 7 Mamadou Lamine Diop M ANACIM (Senegal) mlaminediop@gmail.com 8 Asse Mbengue M ANACIM (Senegal) mbengueass91@gmail.com 9 Ndèye Amy Sal F ANACIM (Senegal) ndeyeamy.sal@anacim.sn 10 Francisca Martey F GMet (Ghana) ciscasowah@yahoo.com 11 Kofi Opoku Nana M GMet (Ghana) opokunn@gmail.com 12 Maureen Ahiataku F GMet (Ghana) maureenahiataku@gmail.com 13 Nafissa Dignon Bertin F DMN (Niger) nafissadignon85@gmail.com Aminou 14 Touné Nazirou M DMN (Niger) fatimid.oran@gmail.com 15 Boureima Salifou Soumaila M DMN (Niger) salifousb3@gmail.com 16 M'Poh M’Koyi M Meteo Togo (Togo) felixmpoh@gmail.com 17 Kossi Tchaa Agniga M Meteo Togo (Togo) agningakossi@gmail.com 18 Gamba Nana Dare F Meteo Togo (Togo) darenana@hotmail.com 19 Ahmad Bello Abdullahi M NiMet (Nigeria) a.bello@nimet.gov.ng 20 Bernard Minoungou M AGRHYMET bernard.minoungou@cilss.int 21 Ibrah Seidou Sanda M AGRHYMET ibrah.seidousanda@cilss.int 22 Mohamed Hamatan M AGRHYMET mohamed.hamatan@cilss.int 23 Lucie Namodj F AGRHYMET lucie.namodji@cilss.int 24 Narcisse Quenum M AGRHYMET narcisse.quenum@cilss.int There were a total of 24 trainees, 9 of whom were women (~38%) and none of whom were youth (under the age of 35 years). **************************************************************** ** Box 2 List of Trainers and Support Staff No. Name Gender Organization/ Position/Title Email Structure 1 Andrew Robertson M IRI Senior Research Scientist awr@iri.columbia.edu 2 Sylwia Trzaska F IRI Senior Staff Associate syl@iri.columbia.edu 3 Tufa Dinku M IRI Senior Research Scientist tufa@iri.columbia.edu 4 Amanda Grossi F IRI Senior Staff Associate amanda@iri.columbia.edu 5 Asher Siebert M IRI Senior Staff Associate asiebert@iri.columbia.edu *remote 6 Kyle Hall M IRI Staff Associate kjhall@iri.columbia.edu *remote 7 Yacouba Ayouba M AGRHYMET Accounts Manager ayouba.yacouba@cilss.int AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 24 * 6: Agenda Day Content Facilitator Morning:  Workshop Opening AICCRA team and Météo  Introduction of Participants Togo (host) October 10  Workshop Objectives and Outline Andrew Robertson (Monday)  Introduction to NextGen/PyCPT 2 Sylwia Trzaska Afternoon:  Introduction to Python  PyCPT 2 Software installation Morning:  PyCPT 2 Software installation (cont.) Andrew Robertson  Lab: navigating different components of PyCPT 2 October 11 (including CPT-IO, CPT-DL python libraries) Sylwia Trzaska (Tuesday) Afternoon: Asher Siebert (remote)  Seasonal forecasting with GCMs Kyle Hall (remote)  Lab: navigating GCM and observed datasets in PyCPT 2 Morning:  Introduction to PyCPT 2 in Jupyter Notebook  Lab: navigating Jupyter Notebook  Introduction to CPT-CORE library Afternoon: Andrew Robertson  Making Deterministic and Probabilistic multi-model October 12 forecasts with PyCPT 2 Sylwia Trzaska o MOS with PyCPT 2: CCA and PCR Asher Siebert (remote) (Wednesday) (including plotting and interpretation of Kyle Hall (remote) EOF and CCA modes) o Deterministic and Probabilistic skill scores in PyCPT 2 o Interpreting Deterministic and Probabilistic multi-model forecasts with PyCPT 2  Lab: Country Seasonal rainfall forecast using GCM rainfall Morning: Andrew Robertson October 13 Sylwia Trzaska (Thursday)  Lab: Country Seasonal rainfall forecast using GCM rainfall (cont.) Asher Siebert (remote) AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 26 **  Lab: Preparation of country projects Kyle Hall (remote) (sensitivity studies) Afternoon:  Presentations of countries plans for sensitivity studies  Lab: country projects  Introduction to the SPI and Creating drought Andrew Robertson October 14 forecasts with PyCPT 2 Sylwia Trzaska (Friday)  Lab: country projects (cont.) Amanda Grossi Sylwia Trzaska October 15  Lab: country projects (cont.) Amanda Grossi (Saturday) October 16 FREE DAY (no workshop) (Sunday) Morning:  Presentation of countries’ project results Sylwia Trzaska October 17 Afternoon: Amanda Grossi (Monday)  Presentations of countries’ project results Asher Siebert (remote) (continued)  Discussion October 18 Sylwia Trzaska  Lab: Write-up of countries’ results for technical report (cont.) Amanda Grossi (Tuesday)  Wrap-up of the technical report Sylwia Trzaska October 19  Discussion: Feedback from the training Amanda Grossi (Wednesday)  Discussion: Next steps AICCRA team and Météo Togo (host)  Concluding Remarks **************************************************************** * References Acharya, N., Ehsan, M. A., Admasu, A., Teshome, A., and Hall, K. J. C. (2021). On the next generation (NextGen) seasonal prediction system to enhance climate services over Ethiopia. Clim. Serv. 24, 100272. Available at: https://doi.org/10.1016/j.cliser.2021.100272 Columbia University. (2020). The Next Generation of Climate Forecasts. Available at: https://iri.columbia.edu/wp-content/uploads/2020/06/Fact-Sheet_Next- Gen_small.pdf Hansen J, Dinku T, Robertson A, Cousin R, Trzaska S, Mason S. (2022). Flexible forecast presentation overcomes longstanding obstacles to using probabilistic seasonal forecasts. Frontiers in Climate 4:908661. Available at: https://doi.org/10.3389/fclim.2022.908661 International Research Institute for Climate and Society (IRI). (2022a). The NextGen Approach — Guide to PyCPT 2 for Seasonal Climate Forecasts. Available from: https://iri-pycpt.github.io/PyCPT2-Seasonal-Forecast-User-Guide/intro.html International Research Institute for Climate and Society (IRI). (2022b). IRI Wiki Pages, NextGen / AICCRA West Africa (WA) Regional Improved NextGen Seasonal Forecasting (PyCPT 2) Training, October 10-19, 2022, Lome, Togo. Available from: https://wiki.iri.columbia.edu/index.php?n=NextGen.AICCRA-Lome-Oct2022 Klemm T, McPherson R. (2017). The development of seasonal climate forecasting for agricultural producers. Agricultural and Forest Meteorology. 232. Available at: https://doi.org/10.1016/j.agrformet.2016.09.005 Minoungou B, Houngnibo M, Nsengiyumva G, Lona I, Namodji L, Halidou T, Hamatan M, Dinku T, Ali A, Zougmore R. 2022. West Africa regional training on ENACTS-related capacity for National Meteorological Services and the Regional Climate Centre. AICCRA workshop report. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available at: https://hdl.handle.net/10568/120345 Nsengiyumva G, Dinku T, Cousin R, Khomyakov I, Vadillo A, Faniriantsoa R, et al. (2021). Transforming Access to and Use of Climate Information Products Derived from Remote Sensing and In Situ Observations. Remote Sensing. Multidisciplinary Digital Publishing Institute;13:4721. Available at: https://doi.org/10.3390/rs13224721 AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 28 Further Reading ** NextGen for rainfall characteristics: Martinez, C. et al. (2022) ‘Seasonal prediction of the Caribbean rainfall cycle’, Climate Services, 27(June), p. 100309. doi: 10.1016/j.cliser.2022.100309. NextGen for Aedes-borne diseases: Muñoz, Á. G. et al. (2020) ‘AeDES: a next-generation monitoring and forecasting system for environmental suitability of Aedes-borne disease transmission’, Scientific Reports. Nature Publishing Group, 10(1), p. 12640. doi: 10.1038/s41598-020-69625-4. DiSera, L. et al. (2020) ‘The Mosquito, the Virus, the Climate: An Unforeseen Réunion in 2018’, GeoHealth. John Wiley & Sons, Ltd, 4(8). doi: 10.1029/2020GH000253. NextGen for undernutrition: White, C. J. et al. (2022) ‘Advances in the application and utility of subseasonal-to-seasonal predictions’, Bulletin of the American Meteorological Society, pp. 1–57. doi: 10.1175/bams-d-20- 0224.1. NextGen for subseasonal rainfall: Domeisen, D. I. V. et al. (2022) ‘Advances in the subseasonal prediction of extreme events: Relevant case studies across the globe’, Bulletin of the American Meteorological Society, pp. 1473– 1501. doi: 10.1175/bams-d-20-0221.1 NextGen for subseasonal fire prediction: Fernandes, K., Bell, M. and Muñoz, Á. G. (2022) ‘Combining precipitation forecasts and vegetation health to predict fire risk at subseasonal timescale in the Amazon’, Environmental Research Letters, 17(7). doi: 10.1088/1748-9326/ac76d8. NextGen as an example of applications of NMME prediction: Becker, E. J. et al. (2022) ‘A Decade of the North American Multimodel Ensemble (NMME): Research, Application, and Future Directions’, Bulletin of the American Meteorological Society, 103(3), pp. E973–E995. doi: 10.1175/bams-d-20-0327.1. NextGen as part of IRI’s “climate services ecosystem generation”: Goddard, L. et al. (2020) ‘Climate Services Ecosystems in times of COVID-19’, WMO at 70 - Responding to a Global Pandemic. WMO Bulletin 69(2), pp. 39–46. Available at: https://public.wmo.int/en/resources/bulletin/climate-services-ecosystems-times-of-covid-19 NextGen for Coffee: Pons, D. et al. (2021) ‘A Coffee Yield Next-Generation Forecast System for Rain-fed Plantations: the Case of the Samalá Watershed in Guatemala’, Weather and Forecasting, 36(6), pp. 2021–2038. doi: 10.1175/WAF-D-20-0133.1. Others: Muñoz, A. et al. (2019) ‘NextGen: A Next-Generation System for Calibrating, Ensembling and Verifying Regional Seasonal and Subseasonal Forecasts’, AGUFM, 2019, pp. A23U-3024. Available at: https://ui.adsabs.harvard.edu/abs/2019AGUFM.A23U3024M/abstract **************************************************************** * Muñoz, Á. G. et al. (2019) ‘Can We Predict “Climate Migrations”? The 2018 Guatemalan Case’, in American Geophysical Union, Fall Meeting 2019, abstract #GC13E-1213. Available at: https://ui.adsabs.harvard.edu/abs/2019AGUFMGC13E1213M/abstract AICCRA Climate Risk Management in Agricultural Extension Refresher Training • 30 Accelerating Impacts of CGIAR Climate It is led by the Alliance of Bioversity ** Research for Africa (AICCRA) is a project International and CIAT and supported that helps deliver a climate-smart by a grant from the International African future driven by science and Development Association (IDA) of the innovation in agriculture. World Bank. Citation: Grossi A, Robertson A, Trzaska S, Dinku T, Zougmoré R, Minoungou B, Mohamed H, 2022. West Africa Regional Training On the Improved NextGen Seasonal Forecasting Approach (PyCPT 2). AICCRA Workshop Report. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available online at aiccra.cgiar.org With support by: ****************************************************************