Turbidity Datasets for Lake Turkana based on Landsat

Based on optical medium-resolution Landsat8-OLI and Sentinel2-MSI (A and B) satellite imagery the turbidity of Lake Turkana was assessed by VISTA between 2014 and 2017 using the reflective band 5 of OLI and 8A of Sentinel (band centre: 864 nm). A large number of analysed images had to be dismissed because of cloud cover. An overview of the used satellite scenes is provided in Table 1 (in the appendix).
The remotely sensed data was corrected for atmospheric effects and transformed to surface reflectance data using the VISTA processing chain before applying a regression algorithm. This algorithm is based on an empirical relationship, derived in another project, between surface reflectance of the water and the turbidity – expressed as Formazin Nephelometric Units (FNU, defined in ISO 7027). This can be used as approximation for relative development of turbidity. For a more accurate evaluation of the remote sensing data local measurements would be beneficial. The realtionship was then applied to an overall number of 16 cloud-free satellite images (Landsat & Sentinel 2): 𝑇𝑢𝑟𝑏𝑖𝑡𝑖𝑦=(𝑅𝑒𝑓𝑙𝑒𝑐𝑡𝑎𝑛𝑐𝑒𝐵𝑎𝑛𝑑5−1.5398)0.1254
Results:
Using all 16 satellite scenes time series analysis was conducted visually. The series is shown in Figure 1. Unfortunately, Landsat 8 derived imagery only covers the western part of the lake while Sentinel 2 covers the whole study area but shows interpolation issues between the two UTM zones (37 and 38). As this issue belongs to the data treatment conducted by the Copernicus Services no correction approach could be applied.
Although the temporal and spatial coverage of the data is relatively sparse, some notes can be made about the spatial patterns presented in Figure 1: The delta region of the Omo River is clearly visible in imagery acquired during late summer and autumn (June to October) as turbidity values are notably higher near the approximated position (up to 120 FNU) of the delta at the northern part of the lake than in the south (20 to 40 FNU). The sharp gradient causes turbulent mixing patterns between the water of the Omo River and the lake. Apart from the effect of the delta the spatial patterns of the turbidity of Lake Turkana show gentle differences (10 to 20 FNU) that are present in the whole study area and seem to be caused by local circulation patterns.
Due to the restricted mapping capacity a set of six test pixels was randomly chosen in the western part of the lake to ensure coverage by Landsat in order to show the temporal patterns of the turbidity (see Figure 2). In Figure 2 the spatially averaged turbidity of 9 pixels is drawn for each of the six subplots over the time. Turbidity values range from near 0 to 120 FNU with clearly visible annual patterns, however, no clear shift between the single years could be identified: Minimum values of turbidity can be observed during spring and increase to maximum values during late summer and
2
autumn. While the maximum values differ notably between the subplots (40 up to nearly 120 FNU) all minimum values can be found near 0 FNU. These findings coincidence well with climate conditions in the headwaters of the Omo River catchment. Annual rainfall is mainly received from convective systems during the time between June to September. These storm events cause higher inflows to Lake Turkana from Omo River with large amounts of sediment and dissolved solids.
Regarding the effects of dam projects along Omo River no definitive conclusions can be drawn. The Gibe I and Gibe II dam were set in service in 2004 and 2010, respectively (Velpuri and Senay, 2012). Thus, the effect of these two projects on the turbidity of Lake Turkana cannot be assessed from the analysis presented in this document.
The Gibe III reservoir project that was finalized in December 2016 falls within the time series analysis, however, the filling of the dam is reported to have begun earlier. According to the company responsible for constructing the dam, Salini Impregilo (https://ethiopia.salini-impregilo.com/en/projects/gibe-iii-hydroelectric.html; accessed at 20th December 2017), impounding started in January 2015. Thus, the effect of the reservoir is probably not visible with the presented analysis.