Scholarlink

<p>One of the current challenges in climate science is seasonal forecasting and the assessment of model performance across different regions. The challenges in Central Africa (CA) are partly related to its large spatial extent, where model performance can vary considerably. Nevertheless, reliable seasonal climate forecasts are useful for decision-making in sectors such as agriculture, energy production, and extreme event management. Given the limited number of studies evaluating the performance of the SEAS4 and SEAS5 models in CA, this study examines the quality of their seasonal precipitation forecasts over the retrospective period (1981–2010) for the four main seasons in the region: December–February (DJF), March–May (MAM), June–August (JJA), and September–November (SON). The evaluation combines analyses of seasonal climatology with deterministic and categorical verification metrics. The results indicate that the SEAS4 and SEAS5 models generally capture the spatial distribution of seasonal rainfall climatology and reproduce the bimodal and unimodal rainfall regimes over Central Africa. Forecast skill decreases with lead time. At long lead times, the Pearson Correlation Coefficient (PCC) decreases to approximately 0.25 and 0.32 during MAM, 0.35 and 0.49 during DJF, and 0.40 and 0.46 during JJA for SEAS4 and SEAS5, respectively, while the Taylor Skill Score (TSS) remains below 0.45. In contrast, short-lead forecasts show better categorical performance, with the Probability of Detection (POD) generally exceeding 0.55 and Accuracy ranging from 0.60 to 0.80 for SEAS5 and from 0.40 to 0.60 for SEAS4 across the four seasons. Overall, SEAS5 shows slightly higher skill than SEAS4, particularly at short lead times, whereas both forecasting systems exhibit reduced performance at longer lead times. These results provide an evaluation of the relative performance of the two forecasting systems over Central Africa and highlight both their strengths and limitations for seasonal rainfall prediction in the region.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluation of ECMWF SEAS4 and SEAS5 Seasonal rainfall forecast skill over Central Africa

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Armand Feudjio Tchinda,
  • Pascaline Liaken Dickmu,
  • Derbetini Appolinaire Vondou

摘要

One of the current challenges in climate science is seasonal forecasting and the assessment of model performance across different regions. The challenges in Central Africa (CA) are partly related to its large spatial extent, where model performance can vary considerably. Nevertheless, reliable seasonal climate forecasts are useful for decision-making in sectors such as agriculture, energy production, and extreme event management. Given the limited number of studies evaluating the performance of the SEAS4 and SEAS5 models in CA, this study examines the quality of their seasonal precipitation forecasts over the retrospective period (1981–2010) for the four main seasons in the region: December–February (DJF), March–May (MAM), June–August (JJA), and September–November (SON). The evaluation combines analyses of seasonal climatology with deterministic and categorical verification metrics. The results indicate that the SEAS4 and SEAS5 models generally capture the spatial distribution of seasonal rainfall climatology and reproduce the bimodal and unimodal rainfall regimes over Central Africa. Forecast skill decreases with lead time. At long lead times, the Pearson Correlation Coefficient (PCC) decreases to approximately 0.25 and 0.32 during MAM, 0.35 and 0.49 during DJF, and 0.40 and 0.46 during JJA for SEAS4 and SEAS5, respectively, while the Taylor Skill Score (TSS) remains below 0.45. In contrast, short-lead forecasts show better categorical performance, with the Probability of Detection (POD) generally exceeding 0.55 and Accuracy ranging from 0.60 to 0.80 for SEAS5 and from 0.40 to 0.60 for SEAS4 across the four seasons. Overall, SEAS5 shows slightly higher skill than SEAS4, particularly at short lead times, whereas both forecasting systems exhibit reduced performance at longer lead times. These results provide an evaluation of the relative performance of the two forecasting systems over Central Africa and highlight both their strengths and limitations for seasonal rainfall prediction in the region.