This post was written by Sarah Alexander, Paul Block, Annalise Blum, Shraddhanand Shukla, Shu Wu, Ben Zaitchik, and Ying Zhang with the Ad Hoc Blue Nile Forecast Group.

In 2018, we formed an ad hoc forecast group to provide annual outlooks for Blue Nile River flow at the beginning of each East Africa summertime rainy season. The motivation for this effort is the impending completion of the Grand Ethiopian Renaissance Dam (GERD; Figure 1). As we described in our original post on the issue, the GERD will be the largest hydropower dam in Africa. It is also the first major infrastructure project on the mainstem of the Ethiopian portion of the Blue Nile River, and the reservoir could potentially store on the order of 1.5 times the annual average flow of the river. The scope of the project and, in particular, its potential impacts on the downstream countries of Sudan and Egypt, have attracted interest and concern across the region and beyond. The reservoir filling period is expected to be particularly tense, given the potential for Ethiopia to curtail Blue Nile flows significantly in an effort to fill the reservoir and begin producing electricity as soon as possible. As recent power shortages have demonstrated, there is a pressing need for reliable electricity generation in Ethiopia, and the long-term economic impacts of the GERD are expected to be transformative.

When we produced our 2018 forecast last spring, there was some expectation that reservoir filling would commence in that year. It now appears that filling will not begin until 2020, on account of construction delays. We are providing a 2019 forecast as an additional test of the forecast approach and of the communication process.

As with last year’s forecast, we have generated predictions for June-September total rainfall for the Upper Blue Nile basin, defined as the Blue Nile River basin within Ethiopia, which is upstream of the GERD site. These forecasts are then applied to a water balance model to estimate June-December Blue Nile River streamflow at the GERD site. The forecast for 2019 shows a high likelihood of near-normal conditions (Figure 2), with close to 70% of rainfall predictions and over 60% of streamflow predictions falling in the “near normal” category (i.e., the middle tercile of historical observations).

Figure 2. Percentage of forecasts in this study predicting below normal, near normal, and above normal June-September rainfall (left) and June-December streamflow at the GERD site (right) for 2019. Forecasts include eight NMME models (ensemble mean for each model with forecasts adjusted for mean biases) and eight statistical models. NMME models were adjusted based on each model's historical mean and variability and rescaled to be consistent with the mean and variability of the climatology. The slightly higher chance of below normal flows is the result of one additional model (Blum1) falling just below the near normal category for the streamflow predictions, but just within near-normal for rainfall.

Our precipitation forecasts comprise both statistical and dynamically-based methods. Further information on these models is available in our 2018 forecast blog post. Statistical methods include previously published models, with coefficients refit to the available historical record and several original statistical forecasts developed by our group. The dynamically based forecasts are drawn from the North American Multi-Model Ensemble (NMME) seasonal forecasts. We use all ensemble members from all NMME models for which May-initialized forecasts were available at the time of writing. Figure 3 shows the results of statistical and dynamically-based forecasts separately for rainfall and streamflow. In general, the NMME dynamically-based forecasts are wetter and have a wider ensemble spread compared to the statistical forecasts, but the majority of predictions from both approaches are for near-average conditions. This forecast can be compared to our 2018 predictions, in which we forecast a higher probability of above-average flows (Figure 4).

Figure 3. (A) Boxplots of June-September rainfall climatology (1982-2017), according to CHIRPS rainfall estimates (grey), and of the 2019 forecasts for dynamically-based models participating in the North American Multimodel Ensemble (NMME; green) and for statistically-based models applied in this study (blue). Dashed lines show the upper and lower terciles of historical rainfall totals. (B) As in (A), but for the June-December forecast Blue Nile flow at the GERD site. NMME models are adjusted by model for variability and scaled to climatology mean and variability. Boxplots illustrate the interquartile range (IQR) with the horizontal line showing the mean value and points represent outliers (greater than  1.5*IQR away from the box.)

Figure 4. Same as Figure 2 but including a comparison to our forecast from 2018 (See http://blog.chg.ucsb.edu/?p=364) for details. Note that there were minor updates to methodology between 2018 and 2019, so this comparison is qualitative.

Examining each NMME model and statistical model separately (Figure 5), we see that there is considerable spread in the NMME multimodel ensemble, but that the median prediction of every NMME model is for near-normal or above-average rainfall. Spread is smaller between the statistical models, but two models do have median rainfall predictions that fall in the bottom tercile of the historical record.

We are currently experiencing a weak-to-moderate El Niño (Figure 6). El Niño events are associated with drought in the Blue Nile basin, so in a sense, it is surprising that so many of the models—both statistical and dynamically-based—predict average or above average rainfall when there is an active El Niño. This forecast optimism stems from the fact that the current El Niño is weak. Assuming that it does not intensify—and projections are that it will persist but is not likely to intensify—we expect a low risk of an El Niño-induced drought in the Blue Nile in the coming months.

We do note that the evolution of El Niño over the rainy season could have a significant impact on the performance of the prediction models included in our analysis. When we generated our original forecasts in May 2018, for example, the El Niño Southern Oscillation (ENSO) was in a neutral state, and intensification towards El Niño was not expected to occur for several months. But El Niño arrived in the middle of the summer rainy season, sooner than had been predicted. In the end, total June-September rainfall ended up being close to the long-term mean, while the forecasts that we issued in May included many members that forecast above-average conditions. We reported this shift in our July 2018 forecast update, which already showed signs that rainfall totals would be lower than had been expected in May. The arrival of El Niño likely tamped down rainfall totals, such that the beginning-of-season forecasts proved to be overly optimistic. As the 2019 rainy season evolves, it will be useful to keep an eye on the evolution of the current El Niño event and how it influences rainfall predictions in NMME and other operational forecast systems.