Meta-analysis for evaluation of fertilizer effect under upland rice cultivation in Africa

Related Research Project
Africa rice farming system


Rainfed upland is one of the dominant ecosystems for rice production in sub-Saharan Africa, of which the yield is very low (approx. 2.1 t ha-1). To overcome such low productivity, fertilizer is necessary requirement, and its effective use is key for improvement of rice yield and farmer’s livelihood. Previous studies often reported that fertilizer application did not result in sufficient yield increase. It’s partly because the yield gain with fertilizer could be affected by biophysical factors such as precipitation and soil texture; however, their interaction remained poorly understood. Hence, the meta-analysis was performed for the dataset fertilizer trials conducted in 8 countries in sub-Saharan Africa, where the environmental backgrounds varied considerably, in order to quantify the effect of soil texture and precipitation on yield gain with fertilizer. 

The dataset was composed of 151 paired observations of control and fertilizer treatment from fertilizer trials using NERICA 4 from 8 countries (Table 1). Soils were classified into low clay soil (≤ 20% clay content) and high clay soil (>20% clay content). Yield gain with fertilizer application (YG), that is yield difference between fertilizer treatment and control, was evaluated with fertilizer rates of nitrogen, phosphorous and potassium and with precipitation. Regression analysis showed that YG was not correlated with fertilizer rates of phosphorous and potassium (p > 0.05, data not shown). Precipitation has closely correlated with YG, irrespective of soil types (p < 0.001, Fig. 1 left). Nitrogen was correlated with YG in high clay soil, (p < 0.001, Fig. 1 right,) but not in low clay soil (p > 0.05). Our Bayesian estimation clarified that YG will increase by 0.168 and 0.145 t ha-1 for low and high clay soil with 100 mm increase of precipitation (Table 2). YG was also expected to increase by 0.653 t ha-1 in high clay soil with 100 kg increase of N fertilizer rate, but in low clay soil, 95% posterior credibility intervals included zero, indicating that YG did not always increase with N fertilizer rate. Overall, these results recommend the fine-tuning of N fertilizer input based on soil type and expected precipitation. 

The obtained results will allow the policy maker or private sector to predict the fertilizer-suitable areas for technical dissemination and commercial sale at the region and/or national level. However, the trends at regional level could not be directly applicable to individual fields or to region where the yields was constrained by other nutrient deficiency such as phosphorous etc. In addition, further assessments was required for the variety other than NERICA 4.    

Figure, table

  1.  Table 1. Description of fertilizer trials using upland rice (NERICA 4) in 8 countries in Africa 

    1) Database:
    2) Tanaka et al. (2017).


  2. Fig. 1. The relationship of yield gain with fertilizer (YG) with precipitation (left) and N fertilizer application rate (right) 


  3. Table 2 Posterior probability distribution of the effect of Precipitation and N fertilizer rate on yield gain with fertilizer under different soil types 

    1) Credible interval is the range containing a 95% percentage of probable values estimated by Bayesian statistics.


    Figure and tables rerprinted/modified with permission from Asai et al. (2021).




Research project
Program name


Term of research

FY 2021–2025

Responsible researcher

Asai Hidetoshi ( Crop, Livestock and Environment Division )

KAKEN Researcher No.: 30599064

Kawamura Kensuke ( Social Sciences Division )

KAKEN Researcher No.: 90523746

Publication, etc.

Asai et al. (2021) Field Crops Research 272: 108284

* Affiliation at the time of implementation of the study.