Runoff forecasting for hydropower purposes in general
Most hydropower producers in Norway have implemented forecasting procedures, with the purpose of optimizing the hydropower production from natural runoff and from water stored in the reservoirs. In Scandinavia, the Swedish HBV-model is the most often used rainfall-runoff model for such forecasting. At Glommens and Laagens Water Management Association, GLB, such HBV-models have been calibrated for simulation of the hydrological response of 34 individual subcatchments. These catchment models are tied together by a basin model, which routes the calculated local discharges through the reservoirs and further down the main river to the sea, where the total catchment area reaches 41.200 km2. The concept includes automatic updating of the HBV-models by use of Kalman filtering techniques.
This is a complete river basin model, which also includes physical limitations on the operation of the 26 reservoirs and watercourse diversions. It is a fast and effective tool for evaluation of the consequences of a particular reservoir operation on any of the 44 power stations in the basin. It can of course also be used for analyzing the effect of changes in the natural runoff from a particular subcatchment on the river discharges further downstream.
In spite of the described advantages one should, however, not forget that this model is mainly intended and calibrated for use under more normal conditions than experienced during the flood of 1995. It is not primarily a flood forecasting tool, but an aid for economical optimization of the hydropower production. On the other hand, forecasting for both flood damage control and hydropower production usually aims at minimizing the unintended loss of flood water from the reservoirs, thereby reducing both the flood peaks and the bypassing of discharges outside the hydropower stations.
At the end of April 1996 snow storage in the river basins of Glomma and Laagen had accumulated to 130-150 percent of normal. In the beginning of May temperatures rose and snowmelt slowly began. After one week, however, temperatures fell again and snowmelt stopped. Fresh snow even fell at altitudes as low as 500 meters and the weather forecasts indicated low temperatures for at least one further week. On this basis GLB already in the beginning of May predicted possible high water levels in June to cause damage around Lake Mjøsa, Norways largest lake. Such predictions, however, are not the same as forecasts, since they are based on quantitative weather forecasts for only one week, and thereafter on a statistical analysis of historical data. They represent a limited selection out of a large number of possible scenarios. This explains the rather large variation of water levels predicted for Lake Mjøsa in the beginning of June 1995.
Water levels predicted for Lake Mjøsa:
On 22 May temperatures in the whole catchment started to rise again. They increased with as much as 10 degrees Celsius during six days and snowmelt was rapidly initiated up to altitudes of 1000 meters. Melt intensities in parts of the river basin reached levels around 20 mm per day for several days. The meteorological situation during this week was characterized by an almost stagnant frontal zone, with locally high precipitation intensities, and at noon 31 May the Norwegian Meteorological Office issued a warning of extreme precipitation. A few days later, on 2 June, the combination of simultaneous snowmelt and precipitation in the upper parts of the basin produced a sharp flood peak. Flood forecasts during these days were sometimes quite uncertain, since they were depending on the precision of the highly variable meteorological forecasts.
Other factors influencing the quality of the flood forecasts are the goodness of the hydrological models calculating runoff from precipitation and snowmelt, and the hydraulic models describing the routing of the flood peak down the river and through lakes and reservoirs, like Lakes Mjøsa and Øyeren. During an extreme flood, like the one of 1995, effects not considered by the models may also occur. Examples of these kinds of effects are the breakdown of dikes, which withdraw large volumes of water from the river, when they could have made large damage further downstream. After the peak of the flood some of this water will gradually be returned to the river again.
The flood of 1995 is estimated to represent return intervals varying from 50 to 200 years, depending on the location within the basin. The water level of Lake Mjøsa, for example, was the second highest in 130 years.
Historical flood levels in Lake Mjøsa since 1789:
Due to the generally large amount of snow in the river basin, GLB already on 7 April asked the Norwegian Water Resources and Energy Administration, NVE, for permission to extraordinary pre-release water from one reservoir, Lake Osen, to prepare space for the expected snowmelt volumes (NVE is the national water management authority granting water rights). Later, on 8 May, GLB asked for another extraordinary permission to pre-release water from Lake Mjøsa, Norways largest lake. The onset of snowmelt was now significantly delayed and higher snowmelt intensities than normal could therefore be expected. The possibility of simultaneous snowmelt contributions from the lowlands and the mountains also increased every day. During the coming weeks GLB and NVE were in frequent contact concerning the growing danger of an extreme flood. Several new permissions for extraordinary regulation had to be granted. In many reservoirs the release of water was maintained at high levels to ensure the flood dampening capacity of the reservoirs. At the most critical stage around the flood peak this was a continuous act of balancing. The strategy was to release water at a high level until the flood peak was expected, and then to decrease releases to a minimum, thereby cutting off the peak. Because the flood peak itself was a combination of snowmelt and precipitation, decisions had to rely heavily on meteorological forecasts.
The combined result of the pre-releases and the other reservoir regulation during the most critical days was a significant downstream reduction of the flood peak. The effect of the reservoirs was unexpectedly large, considering the fact that the total degree of regulation of the river basin is only 16 percent; that is 16 percent of the mean annual runoff within the basin can be stored in the available reservoirs.
The 1995-flood in River Glomma at Elverum:
At Elverum in the Østerdalen Valley runoff was reduced by appr. 800 m3/s (20-25 %), and at Losna in the Gudbrandsdalen Valley by 4-500 m3/s (15-20 %). Without the effect of the reservoirs, the flood at Elverum would have reached levels above the extreme flood of 1789. As a consequence of the reduction of runoff from the upper parts of the basin the water levels in Lake Mjøsa and Lake Øyeren also was significantly reduced. For Lake Øyeren the combined effect of the regulations upstream, excavations made at the natural outlet after the 1967-flood, the reopening of two bypass tunnels, and the increased capacity of the power station, was a reduction of the water level with as much as 4 meters.
The effect of flood reducing measures in Lake Øyeren:
The flood of 1995 clearly shows that the regional Water Managment Association has an important role to play during extreme floods, and that it has a large responsibility for damage prevention. The flood also showed that optimal utilizaion of reservoirs often depends on the possibilities of permitting deviations from normal regulation procedures. Close cooperation between the reservoir owners and the authorities is therefore very important. The primary goal has to be overall maximum damage reduction, without consideration of corporate prestige and interests. The flood of 1995 also showed that it often will be the regional Water Management Association (in the 1995-case: GLB) that has the best knowledge about the river basin and the possibilities for damage reduction.
After the flood an official Flood Mitigation Commission, was appointed by the Norwegian Government in July 1995, with the aim to evaluate possible improvements of disaster prevention related to floods. The Commission so far has concluded that "quantitative forecasting in regulated rivers should be performed as a cooperation between the reservoir owner and the authorities". It has also pointed out that only one forecast should be issued during extreme floods, to avoid different and confusing bases for the planning of actions to be taken. Regarding the effect of reservoirs, the Commission has looked at possible new regulations in the Glomma Basin which could increase the flood reduction potential, since the present degree of regulation is only 16 percent.
Different aspects of the 1995-flood have been reported at several conferences during 1996:
A documentary report on the flood, including many photographs, has also been published by GLB, both in English and in Norwegian. The title of the report is "The 1995-flood in the Glomma and Laagen River Basins". Copies of this report or any of the articles can be ordered by sending a request by e-mail to firstname.lastname@example.org or email@example.com.