Ghana field book results – 2012

Submitted by charlotte.schilt on Thu, 10/11/2018 - 15:31

A total of 318 farms were interviewed to determine the impact of the 2012 N2Africa field trials, implemented in Ghana. There were 122 farms included in the survey of soybean results, where the highest average yield of 1744 kg ha-1 was achieved using a treatment of combined inoculant and P fertiliser. 59 farmers who participated in groundnut trials were interviewed, where a treatment of P and K fertilisation obtained the largest average yield of 549 kg ha-1. A further 99 farmers who participated in cowpea trials were interviewed, where a treatment of inoculant and P fertilisation obtained the largest average yield of 1193 kg ha-1. The Northern region produced the highest average yields for soybean and cowpea, but there were no farmers in the region participating in groundnut trials. The largest groundnut yields were obtained in the Upper East region. The majority of the surveyed farmers had a role as Lead Farmer in the project. Only few surveyed farmers were Satellite farmers.

Determining the impact of agronomic indicators on yield variability was hindered by data entry errors. Key indicators such as plot treatment, farm size and type of mineral fertiliser used contained conflicting values or errors, which impacted analyses. One agronomic indicator which showed potential to explain variability was weeding frequency. Further analyses using this indicator, in combination with planting date, revealed the optimal planting strategy for each crop in the 2012 season. For soybean and cowpea trials, planting in July and weeding three times produced the largest average yields. Groundnut trials benefitted less from increases in weeding frequency and the optimal strategy was weeding once or twice, and planting in July.

The analyses of socioeconomic indicators did identify trends, such as increasing yields with increases in family size and the occurrence of family members working outside the family farm, but a lack of depth in the available variables limited the ability to determine what mechanisms were driving the observed trends.

Although some trends were observed which could potentially elucidate yield variability the errors in data processing and the lack of depth for those indicators which appeared to be explaining variability, impacted the ability to accurate decouple any actual relationships from observed trends. Going forward greater care must be taken to minimise data entry errors. Something which may help data collectors, are observed ranges for indicators such as farm size and yields. This way the data collector can already understand what values are typical for a question, and when the received answer is substantially outside observed ranges they can ask the question again and potentially minimise errors.

Further, there are some indicators which are redundant during data analyses, often because they are exploring similar topics which have been covered by other questions. Those indictors which show promise, such as family size and working elsewhere, could be expanded to include further details such as the amount of income generated from working elsewhere and where that income is spent, to help elucidate any relationships from the observed trends in yield variability. Therefore the data collection could be improved by removing any questions which explore similar topics and adding questions to those variables which appear to be explaining yield variability.

Kirt Hainzer
Not indicated


Linus Franke