\label{sec-1-2}
\label{sec-1-2-1} Ghana’s Akosombo dam is located at southern end of Lake Volta; see Figure \ref{map}. This hydro power plant produces much of the electricity for the country \citep{ecg_2014_2014}. Since the dam can not function when reservoir levels are below critical levels, factors affecting the water supply to Lake Volta are a major social concern for the country. Studies suggest that a dominant factor in determining water levels in Lake Volta is climactic variability \citep{leemhuis_volta_2009}, with concerns including both drought years and lagged effects from repeated dry years, years of below average rainfall \citep{bekoe_impact_2013}. While a limited amount of the precipitation that falls over the Volta basin reaches the dam, a non-linear relationship between run off and precipitation has been observed, with moderate; i.e., within the range of typical interannual variability, variation in rainfall creating large shifts in river flow and consequently reservoir levels \citep{andreini_volta_2000}. The largest contributor to Lake Volta water levels is estimated to be the Oti river. The Oti is located on the eastern edge of the Volta basin; see Figure \ref{map}, and feed mostly by runoff from the westerly side of the Akwapim-Togo mountain range \citep{andreini_volta_2000}. This river system receives less precipitation than the Black Volta or White Volta rivers, which occupy most of the drainage basin, but has the largest percentage of this rainfall converted to runoff into Lake Volta.
The Akosombo dam is an import component of Ghana’s energy strategy \cite{ecg_2014_2014} and Lake Volta as well as the greater basin is heavily depended on by subsistence agriculture and fisheries activities across the region \citep{zwieten_review_2008}. It is believed that the viability of these activities is affected both by long term climatic trends and global modes of interannual variability, and that these risks are poorly reflected in current management decisions \citep{owusu_changing_2008}. Current difficulties caused by variability are expected to be compounded by rising populations and increased activity \citep{owusu_changing_2008}.
\label{sec-1-2-2} Observational records for the Volta basin region are limited, but less sparse than for many other locations in West Africa. This makes the region an ideal place for our study, since we have to face head on the issues of limited data, while still having enough material to ‘talk about’ in order to illustrate the problematic situation. Data is available from satellite observation networks, as well as meteorological stations (as discussed below). How well these observations represent experienced conditions, however, is often unknown. There is enough information to ‘get a sense’ of what occurs in the region, but spatial inhomogeneity, inconsistent sampling, and unverified reliability, limits any rigorous assessment of the observational record.
\label{sec-1-2-3}
It is an open question at what scales model simulations are informative. The layout and intricacies of the Volta Basin makes it possible to conduct a relative testing of appropriate scales.
Being at the very downstream end of the Volta basin, Akosombo dam is a result of all combined upstream activity. As such, it can (indirectly) inform about unobserved behavior upstream, as well as allowing us to relate multiple grid cells (the Volta Basin) to a single impact (dam levels). Lake Volta and thus the Akosombo dam are fed by the Volta basin and thus depend on the rainfall from across the Volta Basin area. However, within this area the Oti sub-basin is understood to be the greatest contributor to lake and dam levels. Furthermore, rainfall falling in the Oti sub-basin is reflected in lake and dam levels with a more immediate effect than that of the other sub-basins. This is because of the understanding that rainfall over the Akwapim-Toto mountain range, located in the Oti sub-basin, is quickly transported down slopes and into the river basin. This implies that while it is possible to consider the relationship between rainfall over the entire area and dam levels, it is further possible to break this down to a finer scale analysis and look at how well RCM simulations perform when it comes to the different sub-components of the basin.