The impact of pixel scale satellite rainfall uncertainty on modelled crop yield and agricultural development applications

One of five research papers commissioned in order to synthesize recent scientific research relevant to the multiple points of connection between maize intensification and the broader Sustainable Development Goal (SDG) agenda, with a particular focus on identifying actionable implications for framing effective data-for-SDG strategies. It is published as the Proceedings of: Sustainable Development Goals and Data Needs for Maize Intensification Plans, held October 25-26 2016. ABSTRACT In agricultural development and climate adaptation, initiatives are frequently constrained by an absence of comprehensive historical weather information. Satellite rainfall estimates are therefore starting to become a useful tool within agricultural development, particularly for crop simulation, farmer livelihoods research and weather based index insurance. However, satellite rainfall has several different properties to gauge information, which if not taken into account, can lead to systematic bias in output. This is particularly the case because many agricultural development applications are focused and calibrated on site-specific weather information, whereas satellite rainfall products currently have a minimum resolution of 5km and low daily-pixel skill. This paper explores how satellite rainfall estimates have been used within agricultural development applications. We then assess the impact of satellite rainfall uncertainty for crop simulation output for Ethiopian maize. This is done by driving the General Large Area Model for annual crops (GLAM) with stochastically generated TAMSAT satellite rainfall ensembles. The ensemble approach is shown to provide a satisfactory alternative to gauges or satellite bias correction.