Search for an astronomical site in Kenya (SASKYA) using climate reanalyses and high-resolution meteorological model data

Abstract

The goal of the Search for an Astronomical Site in Kenya (SASKYA) project is to identify the best possible site(s) in Kenya for astronomical optical observation, using ERA-interim climate reanalyses and high-resolution UK Met Office Africa Limited Area meteorological model (Africa-LAM) data. This initial search focusses on a selection of 13 candidate mountain peaks across Kenya. A mixture of 30 years (1981–2010) of relatively coarse-grained ERA-interim reanalyses data and 12 months’ (2011–2012) of much higher resolution UK Met Office Africa-LAM data were used to determine the best possible sites. Cloud cover, precipitable water vapour (specific humidity), vertical velocity, aerosol loadings and wind data were analysed. The results confirm that many sites in Kenya are reasonably cloud free, with estimated photometric night fractions of possibly 50 % at the best sites. Significant seasonal inter-annual and inter-decadal variations in cloud cover can be expected, however. Average precipitable water vapour (PWV) values are uncomfortably high, but periods of much lower PWV can be expected during favourable conditions in the dry seasons. Long-term vertical velocities (as a proxy to determine areas of improved “seeing” conditions) indicate that good astronomical viewing conditions are likely to be dependent on the season and wind direction across Kenya. Finally, after full consideration of the climatological data, a trade-off is expected between the best possible site in climatological terms, and the practicalities of installing remote equipment in isolated, inaccessible areas with little or no infrastructure.

Appendices

Appendix 1

We were expecting to see broad qualitative visual agreement between the output from the Africa-LAM and the local topography of Kenya. This is confirmed, for example, in Fig. 13, which shows the precipitable water vapour over Kenya, at 00z on 16 August 2011, as determined using the Africa-LAM. The high resolution of this model enables regional climatic features associated with the Rift valley to be discerned [e.g. a north–south-aligned band of higher PWV is apparent, depicted by brighter shades (red and yellow in online version)]. Very low PWV associated with the high peaks of Mount Kenya/Mount Kilimanjaro are also apparent [shown as small concentric regions of darker shades (blue and purple colours in the online version)]. The moister lowland coastal area of eastern Kenya (generally higher PWV values) is also clearly apparent. These topographic features are discernible because PWV varies greatly with height (Morland and Matzler 2007).

1.1 Africa-LAM cloud cover

Table 5 presents the average cloud cover statistics for the candidate sites, as determined from the Africa-LAM for the selected days during the period of 2010–2011. These are expressed as an “Africa LAM cloud cover (ALCC) statistic”, which is defined here as the average of the model’s bulk cloud fraction (bcc) variable over all 70 levels of the model, as defined by Eq. 4:

$$ \mathrm{ALCC}=\frac{{\displaystyle {\sum}_1^{70}\mathrm{b}\mathrm{c}\mathrm{c}}}{70} $$

(4)

where bcc is the bulk cloud fraction for each of the 70 layers of the Africa-LAM.

The data presented in Table 5 is therefore also a measure of the integrated cloudiness. As detailed earlier in Section 2, however, this vertical integration is not the same as total cloud cover because varying amounts of overlapping cloud layers will increase the integrated cloudiness value more greatly than the total cloudiness. Many other varying factors such cloud opacity, thickness, temperature, air pressure, ice/water content and droplet/crystal distribution will also have an unknown influence on the variable. Due to these factors, and also the high number of atmospheric levels in the model, a degree of caution should be applied when considering these as straightforwards percentages of cloud cover. Instead, they may be thought of as average relative values for the selected sites over the analysed nights in question.

Furthermore, because of the limited number of sample nights used in this part of the study, further caution must be applied when considering these results. It is therefore recommended to interpret these data only as rough guides as to the relative differences in cloud cover between the four best-performing sites, based on a very limited survey of 11 key nights over the 2010–2011 period.

Keeping these limitations in mind, it can be seen from Table 5 that the values for Mount Kulal, Ol Donyo Nyiro and Warges are all roughly equivalent, ranging from ∼32 to 36 %. The very low value of 16.5 % found at Mount Kenya (The Barrow) is because of its great height, as there is less atmosphere to integrate through and the Africa-LAM model may not be able to capture continual thin but local low orographic cloud at this location, despite it being a very high resolution model.

Therefore, it is of our opinion that no firm additional conclusions can be drawn from this analyses, as to whether any of the best-performing SASKYA candidate sites are considerably more or less cloudy than one another, based on the output of the Africa-LAM. On-site measurements using astronomical and meteorological site testing equipment at the candidate sites would be required to refine our knowledge of  their characteristics.

A plot of integrated bulk cloud fractions from the Africa-LAM against the ERA-interim values of total cloud cover for each candidate site might be expected to show some level of agreement between the two independently derived variables. However, this task could not be undertaken in the current study, due to the lack of any temporal overlap in the series of each dataset. It is concluded, therefore, that any further deductions on the meso-scale and local meteorology of the candidate sites would therefore need to involve a much larger dataset output from the Africa-LAM or an equivalently resolute model.

Appendix 2: integrated specific humidity

In a similar way to that undertaken when analysing cloud cover output from the Africa-LAM (as described in the preceding section), specific humidity statistics were determined for the same four best-performing SASKYA candidate sites using output from the Africa-LAM. Unfortunately, it was not possible to determine the precipitable water vapour (PWV) for each site using the Africa-LAM model, as the specific heights of each model layer (for the determination of each layer’s thickness and thus mass weighting) were not specified in the Africa-LAM humidity files. Thus, a new variable, which was called the integrated specific humidity (ISH), was defined as the vertical summation of the specific humidity through the 70 levels of the Africa-LAM model, averaged and multiplied by the mean total mass of the atmosphere above the candidate site, as follows:

$$ \mathrm{I}\mathrm{S}\mathrm{H}=\kern0.5em \frac{{\displaystyle {\sum}_{z=1}^{z=70}q}}{70}\kern0.5em \overline{M} $$

(5)

where q is the specific humidity (g_{water vapour}/kg_air), and $\overline{M}$ is the average mass of the atmosphere above each candidate site (calculated by determining the long-term 1981–2010 average atmospheric air pressure at each candidate site using the ERA-interim reanalysis).

Thus, differences between ISH and PWV will occur where the atmospheric layers in the model are of varying thickness and mass. Lacking a suitable alternative in this instance, however, meant that the computation of the ISH was deemed the best option for us, given the circumstances. The resulting ISH values for each site are shown in Table 6 in units of kg_{water vapour}/m², which is also equivalent to millimetres, the same unit as for PWV.

Looking closely at Table 6, it can be seen that the ISH values are considerably higher than the mean PWV presented for each site in Table 4 (more than double in some instances), based the ERA-interim dataset. This is probably because of the major shortcomings in the method of determining ISH (compared to PWV), as outlined above (as it is unlikely that the ERA-interim model is incorrect in the earlier instance). The largest cause of error is probably due to the incorrect mass weightings per layer of atmosphere. Therefore, it is recommended that the absolute values presented in Table 6 should be ignored in absolute terms, but instead be considered in relative terms only. In this light, some useful information can still be extracted, namely that standard deviations of ISH amount to only a small fraction of the total ISH signal at each site (unlike what we found for cloud amounts in the previous section), the Mount Kenya site has ISH values of less than half that of most of the other sites (Mount Kulal, Warges), and Ol Donyo Nyiro mountain is somewhat drier than either Mount Kulal or Warges (as also found out in Section 2 using the ERA-interim analysis). These confirm the findings already presented in Table 4 using the ERA-interim reanalysis.

Furthermore, a plot of Africa-LAM ISH values against ERA-interim PWV values (Fig. 14) for all of the original candidate sites shows fairly good agreement (r ² = 0.51, significant at 90 % confidence level). This correlation mainly exists, however, because PWV decreases rapidly with height (Morland and Matzler 2007) and our sites cover a range of altitudes. Therefore, a much larger sample size of output statistics (preferably PWV, not ISH) from the Africa-LAM or an equivalently resolute model, compared to ERA-interim data for the same period and presented as anomalies of PWV from their mean, might be a better way of comparing the two such datasets—it is recommended that any future studies attempt this.

Fig. 13

Precipitable water vapour (specific humidity; kg_water/kg_Air) map of Kenya, at 00z on 16 August 2011, from the UK Met Office Limited Area Model (LAM) for Africa (Met Office, ×200), run at 12 km resolution with 70 vertical levels

Fig. 14

A plot of Africa-LAM integrated specific humidity (ISH) determined from 33 instances of the model during 2010–11, against long-term (1981–2010) ERA-interim precipitable water vapour (PWV) for the four candidate sites of Mount Kulal, Ol Donyo Nyiro, Warges and Mount Kenya (The Barrow)

Table 5 Africa-LAM cloud cover statistics (ALCC)

Table 6 Africa-LAM integrated specific humidity (ISH)

Appendix 3

The times and days of the Africa-LAM data that were selected for data download were as follows:

22UTC, 15th June 2011	23UTC, 15th June 2011	00UTC, 16th June 2011
22UTC, 15th July 2011	23UTC, 15th July 2011	00UTC, 16th July 2011
22UTC, 15th August 2011	23UTC, 15th August 2011	00UTC, 16th August 2011
22UTC, 15th September 2011	23UTC, 15th September 2011	00UTC, 16th September 2011
22UTC, 17th October 2011	23UTC, 17th October 2011	00UTC, 18th October 2011
22UTC, 25th December 2011	23UTC, 25th December 2011	00UTC, 26th December 2011
22UTC, 15th January 2012	23UTC, 15th January 2012	00UTC, 16th January 2012
22UTC, 14th February 2012	23UTC, 14th February 2012	00UTC, 15th February 2012
22UTC, 15th March 2012	23UTC, 15th March 2012	00UTC, 16th March 2012
22UTC, 15th April 2012	23UTC, 15th April 2012	00UTC, 16th April 2012