Using climate forecasts for better decision making in agriculture: The case of sweet potato farming in Kitale Kenya (2018)


Source

This study was conducted as a group project in Agricultural Systems Modeling. I partnered with Agus Hasbianto in this project. 

Introduction

The objective of this to use Climate Predictability Tool (CPT), a statistical software, to make rainfall forecasts that would be vital in agricultural decision making. We aimed to apply information gained to inform sweet potato farmers about the best sowing date for the crop. Ultimately, we hope to better sweet potato yields in western Kenya. Only some results are shown here. Please note that final potato yield forecasting was not conducted. 
Study area. 


Data and Sources

To complete this project, we will need the following datasets;
  • Observed rainfall records from 1982 to 2014
  • Rainfall observations – 1982 to 2014 for the short season of October, November and December (OND) and the preceding three months (July, August and September (JAS)). This is GPCC data downloaded from the International Research Institute website.
  • Sea surface temperature observations – covering the period of 1982 to 2014 for the above two three-month periods. This is ERSST4 data downloaded from the same website as precipitation above. 

Results

The model for making rainfall forecasts in Kitale based on Pacific sea surface temperatures was skillful and made accurate above-normal rainfall predictions. The 3-state model had a Goodness Index of 0.059 and despite it having very weak skills in most parts of the country, a relatively good skill was recorded for our study area. The map is provided below.
Pearson's correlation between rainfall in Kenya and Pacific sea-surface temperatures. 

Zooming into our study area in Kitale, the model performed very well. Based on the ROC curve, the model was accurate for above normal predictions but below normal predictions were not too accurate – lower and highest tails were false predictions. 
Hits and false-alarm rates for rainfall predictions in Kitale based on Pacific Ocean SSTs.