Countries: Mali, Burkina Faso, Ghana, Niger, Nigeria, Ethiopia, Kenya, Uganda, Tanzania and Zambia

Introduction

Africa is home to important centers of origin, diversity and cultivation of millets. Annually, millet is grown over an area of 19.5 Mha  producing 13.6 Mton in Africa (the average for the period from 2015 to 2020; FAOSTAT). Pearl millet is a climate hardy crop which is grown in harsh conditions, but as a subsistence crop. Harvested from an area of 19.5 Mha in the semi-arid regions of Africa pearl millet contributes 17% area to cereal production. In Sub-Saharan Africa, millet biodiversity constitutes both a unique ecological heritage and a critical food security component among millions of small-scale farmers. Around 68 percent of all millet lands in Africa is in the ten Sub-Saharan countries (Table 1).

Table 1. Annual millet harvest area in sub-Saharan Africa
Country	Annual millet harvest area (1000 ha)*	Percentage of total  millet lands in Africa
Burkina Faso	1221	6.3
Ethiopia	460	2.4
Ghana	170	0.9
Kenya	114	0.6
Mali	2075	10.6
Niger	6877	35.2
Nigeria	1855	9.5
Uganda	140	0.7
Tanzania	298	1.5
Zambia	44	0.2
Total	13254	67.9
* The data are from FAOSTAT for the period from 2015 to 2020

 

 

The crop model

The Python Crop Simulation Environment of WOFOST (WOrld FOod STudies) was used for the implementation (https://pcse.readthedocs.io/en/stable/). This model takes into account phenological development, leaf development, light interception, CO₂ assimilation, root growth, transpiration, respiration and partitioning of assimilates (de Wit et al., 2019). Daily weather data, crop parameters, soil parameters, and management data are needed to run the model and simulate water-limited potential yield (Yw). The source of different data to do the simulations for each country were presented in Table 2.

Table 2.  Sources of data to simulate rainfed millet potential yield and for calculating the yield gap. Note: GYGA CA = GYGA country agronomist
Country	Sowing window	Daily weather data	Cultivar thermal time requirment**	Soil data	Actual yield
Burkina Faso	GYGA CA	Propagated data*	calculated	AFSIS***	National Statistical database
Ethiopia	GYGA CA	Measured and Propagated data	calculated	AFSIS	GYGA CA
Ghana	GYGA CA	Measured and Propagated data	calculated	AFSIS	GYGA CA
Kenya	GYGA CA	Propagated data	calculated	AFSIS	FAO
Mali	GYGA CA	Propagated data	calculated	AFSIS	FAO
Niger	GYGA CA	Measured and Propagated data	calculated	AFSIS	GYGA CA
Nigeria	GYGA CA	Measured and Propagated data	calculated	AFSIS	GYGA CA
Tanzania	GYGA CA	Propagated data	calculated	AFSIS	GYGA CA
Uganda	GYGA CA	Propagated data	calculated	AFSIS	FAO
Zambia	GYGA CA	Measured and Propagated data	calculated	AFSIS	GYGA CA
* Details are in Van Wart et al. (2015) ** Calculated based on the sowing, flowering and maturity timing information from the GYGA country agronomist using weather data and cardinal temperatures *** AFrica Soil Information Service (Leenaars J.G.B., 2018)

 

Reference weather stations

The simulation was performed for 81 weather stations in the ten countries (Table 3). These were identified as reference weather stations for millet in SSA countries based on the GYGA protocol (https://www.yieldgap.org/web/guest/methods-overview). SPAM (Spatial Production Allocation Model; https://www.mapspam.info/) maps, together with expert knowledge from agronomists and experts from these countries, were used to identify the millet harvested area  (You et al., 2009, 2014). Following van Bussel et al (2015), a total of 81 buffer zones were selected for rainfed millet in SSA countries (Table 3, Fig. 1)

 

Figure 1. The location of the reference weather stations for rainfed millet in Sub-Saharan African countries

 

Weather data

Weather data used in simulation included daily solar radiation, maximum and minimum temperatures, precipitation, vapor pressure deficit, and wind speed. Weather data for selected weather stations were subjected to quality control measures to fill in missing data and identify and correct erroneous values. In the case of those stations with only few years of weather data (Table 3), long-term weather data were generated based on correlations between measured weather data and NASA-POWER maximum and minimum temperatures (Van Wart et al., 2015). In the case of solar radiation and rainfall, the data from NASA-POWER were used without any correction (NASA, 2022). In the case of buffers without measure weather data at all, named the virtual stations (Table 3), uncorrected NASA-POWER data were used for all meteorological variables needed for crop modeling.

 

Soil information

Soil data was obtained from the AFSIS^[1] (Leenaars J.G.B., 2018). Soil moisture content at field capacity, soil moisture content at wilting point, not infiltrating fraction of rainfall, initial soil water, and maximum rootable depth of the soil were obtained as soil parameters for each weather station.

Management data

The simulations were done under water-limited conditions with no limitation for nutrients. So, the sowing date within the buffers of the weather stations was the main management input data needed to run the model. For this purpose, the common sowing windows of millet for each buffer were retrieved through agronomists from the countries (Fig. 2). These sowing windows and an algorithm were used to estimate the sowing date for each year for each weather station. The algorithm calculates the amount of cumulative rainfall for seven consecutive days into the sowing window. The last day of this period resulting in more than 20 millimeter cumulative rainfall was considered as the sowing date. If there was no consecutive seven days with at least 20 millimeter rainfall into the sowing window, the last day of the sowing window was as the sowing date.

Crop parameters

The crop parameters consist of parameter names and the corresponding parameter values that are needed to parameterize the components of the crop simulation model. These are crop-specific values regarding phenology, assimilation, respiration, biomass partitioning, etc. The same crop parameters were used for the simulations in all the weather stations except the parameters for phenology. The phenological parameters including the thermal time requirement from emergence to flowering and from flowering to harvesting were calculated based on the observed data for sowing, emergence, flowering and harvesting using the cardinal temperature and weather data at each weather station (Table 3).

 

Figure 2. The sowing window to plant rainfed millet at the weather stations in Sub-Saharan Africa countries; The name and information of each station are presented in Table 3.