Research Articles Determining ideal sites for a pilot experiment in Colombia to trial new forages in East Africa

The livelihood of approximately 37 million people of the East African (EA) region is highly dependent on livestock. As the pressure for livestock products heightens with growing populations, already harsh environmental conditions are predicted only to worsen (Fisher et al. 2005; Thornton et al. 2007).

In Cali, Colombia, the International Center of Tropical Agriculture
(CIAT) is leading a cutting edge new breeding programme by developing improved forage varieties specially designed to succeed in the East African market.

Cluster analysis
Characterise environmental zones of East Africa into separate environmental
clusters based on the following information associated with each pixel:
* Areas of high livestock density (Ramankutty et al., 2008)
* Areas of annual precipitation > 700 mm
* Qualitative soil data (Hengl et al., 2014)
* Environmental data: Annual consecutive days of precipitation > 1mm (CDD);
Total annual rainfall (TR); Annual maximum average rainfall over 5 days (P5D); Annual 95th percentile of precipitation (P_95); Annual average vapor pressure deficit (VPD) (Ruane et al. 2015; Funk et al. 2016).

A distance matrix was generated according to Gower’s distance to derive hierarchical clusters.

Similarity between Colombia and East Africa

We identified the centroid of each cluster, then the 40 pixels closest to it are located and averaged to obtain the reference. The comparison is developed for each indicator, calculated for time series over thirty years, within the same pixel. Distribution of Dynamic Time Warping Multivariate distances (DTWMd) where obtained for pixels of the Colombian area. The pixels displaying a DTWMd closest to zero (similarity threshold of 20%), are identified as the most similar to the reference pixel. Selected pixels from the Area in Colombia, were used to generate the map of Colombia which detailed possible areas suitable for a pilot experiment. All of the above steps are performed using the free R statistical software.

Discussion and Conclusion:

Here, we have presented a highly valuable method for measuring site similarity
for future trials that will provide vital information for targeting genotypes to
environments. We have successfully selected possible sites in Colombia for
the initial testing and first culling of improved forages, which are representative
in climate and soil conditions of the four East African clusters. The outcome of
this analysis of environments will be compared against analysis of genotype
by environment interaction from phenotypic data from each of the identified
clusters.