NextGen approach to hydrological forecasting: Adapting PyCPT tool for hydrological forecasting Bernard Minoungou, Abdou Ali, Mandela Houngnibo, Mohamed Hamatan, Agossou Gadedjisso-Tossou, Alcade C. Segnon, Robert B. Zougmoré Keys message ● A key challenge frequently mentioned by NMHSs limiting the operationalization the NextGen approach to seasonal climate forecasting systems, especially the use the PyCPT tool was the lack of consideration of hydrologic parameters. To address this challenge. ● AGRHYMET has adapted the PyCPT tool for seasonal hydrological forecasting. The adapted version of PyCPT has been illustrated and tested for seasonal hydrological forecasting for the 2023 rainy season, with satisfactory results. ● Future development needs to integrate additional features such as multi-model ensemble and flexible forecasts. Introduction AGRHYMET Regional Climate Center for West Africa and Sahel (AGRHYMET-CCR-AOS), as part of its statutory mandate works to improve seasonal and sub-seasonal forecasting capabilities by using the NextGen approach (Houngnibo et al., 2022; Ali et al.; 2022). The NextGen forecasting system helps forecasters evaluate the performance of different global climate models, which helps determine how best to correct and combine them. It also helps forecasters select the best climate models for any region of interest through process-based evaluation, and it automates the generation and verification of forecasts suitable for multiple time scales at the regional, national, or local levels (Hansen et al., 2022). Through the Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA) project, AGRHYMET Regional Climate Centre has been capacitating National Meteorological and Hydrological Services (NMHSs) in West Africa and the Sahel on NextGen seasonal forecasting systems. The capacity development efforts focus mainly on Python interface to the Climate Predictability Tool (CPT) or PyCPT, a tool developed by the International Research Institute for Climate and Society (IRI) to implement the NextGen approach to climate forecasting (Hansen et al., 2022). The continuous improvement of the PyCPT tool has recently enabled the integration of key characteristics of the rainy season such as the onset dates of the season, dry and wet sequences, and number of dry and wet days, in addition to total rainfall. While hydrological forecasts of water availability from watersheds in major river basins are essential to support operational planning and management, the latest version of PyCPT developed by IRI does not take into account the seasonal forecast of hydrological variables. A recent survey the NMHSs on the barriers to operationalization of the NextGen approach and use the PyCPT tool indicated that a key challenge limiting the operationalization and use the PyCPT tool was the lack of consideration of hydrologic parameters (Segnon et al., 2023). To address this challenge and respond to the needs of NMHSs in the region in charge of hydrological monitoring, AGRHYMET has adapted and improving the PyCPT tool for seasonal hydrological forecasts. In this Info Note, we present the methodology adopted, achievements, and some perspectives to further improve the operationalization of NextGen approach in West Africa and the Sahel. Methodological Approach Data The primary hydrological data used to produce seasonal forecasts are the daily flows recorded at the main river stations in the region. These data are distributed across the main river basins. Data from one hundred stations distributed across the 17 NMHSs in West Africa and Sahel were used. Seasonal hydrological forecasting focuses on forecasting average river flows calculated over the high-water period. In this case study, we used data of Sea Surface Temperature (SST), wind patterns, or geopotential fields as predictors. We also used AOGCM (Atmospheric-Oceanic General Circulation Models) output. Statistical approach Like seasonal climate forecasts, hydrological forecasts are based on correlations between ocean surface temperature, precipitation, wind speed, and hydrological variables such as river flows for the high-water period. Principal component regression (PCR) is the most widely considered method. Skill assessment A variety of forecast performance scores consisting of scores based on continuous measurements, those on observed measurements, and, in some cases, forecasts are used as well. The most used criteria are listed in the diagram below (Figure 1). Figure 1: List of the skill assessment criteria (PyCPT) Implementation of hydrological seasonal forecasting on PyCPT PyCPT 2 is a set of python libraries designed to interface with CPT to facilitate operational climate forecasting and research in Python. The Python codes developed by IRI do not allow hydrological forecasts to be performed due to the fact that hydrological data are punctual. The adaptation therefore consisted of redeveloping Python codes to interface the hydrological analyzes with CPT. The codes developed take into account the following aspects: • quality control of the predictors and their filtering taking into account the rate of missing data required by CPT; • spatial verification and graphic visualization of the location of hydrometric stations: this visualization is involved in the analysis of the spatial coherence of the forecasts developed; • performance of statistical analyzes on punctual variables such as river discharge; • Visualization of the performance of the forecasts developed; • Visualization of deterministic and probabilistic forecasts. The adapted PyCPT for seasonal hydrological forecasts is illustrated by the workflow in Figure 2. The workflow largely follows the one proposed by IRI. The aim is to automate and/or simplify multi-model ensemble forecasts and their updates. At the current stage of development, ensemble forecasts and flexible forecasts have not yet been integrated. Deterministic skill metrics Pearson correlation Spearman correlation Two_alternative_forced_ choice Roc_area_below_normal Roc_area_above_normal Probabilistic skill metrics (in sample) Generalized_roc Rank_probability_skill_ score Figure 2: Implementation of hydrological seasonal forecasting on PyCPT workflow Application The 2023 hydrological forecast over West Africa using the adapted PyCPT The adapted PyCPT tool was used to produce seasonal forecasts over West Africa. The forecast configuration is summarized in the table below. Models CanSIPSIC3, CCSM4, GEOSS2S, CFSv2 MOS Method PCR Predictor (GCM Data) Rainfall (simulated by GCMs) for the rainy season Predictand (Observed Data) Observed discharge Training period 1982-2016 Target Season for predictor Major rainy season over the region (AMJJASO) Predictand Domain 17 ECOWAS and CILSS countries Predictor Domain 17 ECOWAS and CILSS countries Assessment Spearman correlation, Pearson correlation, 2AFC, ROC Above and ROC Below Step1: Forecast setup Creation of folders Selection of forecasts variables Choice Predictor and Predictand domain, Filtering predictand based on missing values rate Step2: Stastical analysis and skill assessment Download predictors Run PCR analysis Plotting and assess skills Step 3: Forecasts Generation Calibrated forecasts from individual models Constructing MME & Skill Scores Generating Flexible Forecasts Saving results Principal component regression was carried out by relating the mean discharge for the high- water period to the precipitation from the general circulation models. The performance criteria obtained are shown in Figure 3. Most of the models considered in this study showed an acceptable performance. Figure 3: Performance of models used for forecasting The probabilistic seasonal hydrological forecasts are illustrated in Figure 4 for the four models considered in this study. Overall, flows are expected to be equivalent to or higher than the average of the reference period. Figure 4: Prévisions saisonnières hydrologiques de la saison pluvieuse 2023 Lessons learned and perspectives The work carried out by the AGRHYMET Regional Center, in response to the needs expressed by players in the sub-region, has resulted in the enhancement of PyCPT for hydrological applications. The current development has been tested through case studies in West Africa, and the results obtained are promising. It is therefore necessary to continue development by integrating other functionalities, in particular ensemble and flexible forecasts. The next steps of the work will consist of: ● Continue adapting PyCPT for hydrology by integrating additional functionalities; ● Bulk of the code moved from notebook to libraries, making updates easier; ● Develop a prototype sub-seasonal version; ● Elaborate and share a synthetic practical guide for the next steps; ● Organize online courses (E-learning) to enable users to overcome bottlenecks when using the tool; ● Assess the capacity of national stakeholders to take ownership of the system and organize more face-to-face training sessions. References Ali, A., Houngnibo, M.C., Agali, A., Gadédjisso-Tossou, A., Mohamed, H., Segnon, A.C., Zougmoré, R.B. 2022. Subseasonal forecasts in West Africa: Current status and prospects for operationalization. AICCRA Info Note. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available at: https://hdl.handle.net/10568/127021 Houngnibo, M.C., Ali, A., Gadédjisso-Tossou, A., Mohamed, H., Agali, A., Minoungou, B., Quenum, N., Segnon, A.C., Zougmoré, R.B. 2022. Towards a new approach for Seasonal Climate Forecasting in West Africa. AICCRA Info Note. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available at: https://hdl.handle.net/10568/127020 Hansen, J., Grossi, A., Trzaska, S., Dinku, T., Baethgen, W., Ali, A. 2022. Scaling Out the Next Generation of Seasonal Climate Forecasts in Africa. AICCRA Info Note. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available at: https://hdl.handle.net/10568/125770 Segnon, A.C., Obossou, E., Ali, A., Houngnibo, M.C., Minoungou, B., Mohamed, H., Zougmoré, R.B. 2023. Towards operationalization of NextGen seasonal forecasting systems in West Africa: Stocktaking of the regional capacity building initiative AICCRA Info Note. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA). Available at: https://hdl.handle.net/10568/135206 https://hdl.handle.net/10568/127021 https://hdl.handle.net/10568/127020 https://hdl.handle.net/10568/125770 https://hdl.handle.net/10568/135206