ABSTRACT
Burkina Faso lies around the Niger River. In the Northern part
of country, food insecurity is a worry for inhabitants, authorities, and rural
development partners. Therefore, knowledge of food production level by local
yield forecasting before harvest is important to the local Early Warning
Systems (EWS) in areas where there are significant climate change as shown by
famers and National Direction of meteorology. It's also a way to know food
deficient localities and to define in time the way and opportunity for
intervention. In this Complementary Master thesis work, we intend to forecast
yield of two basis food cereals using meteorological, soil and crops
information of three Burkina Faso's province: Passoré, Yatenga and Soum.
The different data enumerated above permitted to propose crop yield forecasts.
The evaluation of the models given by the generalization error shows that, in
the province of Passore, none forecasting model is reliable to predict yield of
millet or sorghum. It's the same for the sorghum yield prediction in the
province of Yatenga. The main model performance parameters are: R2=19.63 %,
R2p=18.62%, RMSE=221.97 Kg/ha and RRMSE=35.35%. The models to forecast yield of
millet and sorghum in the province of Soum are much more reliable. Similar good
results are found for the millet yield forecasting in the province of Yatenga.
In this last case performance parameters show that: R2=65.89%, R2p=62.43%,
RMSE=132.70% and RRMSE=22.53%. These reliable models use two to three
explicative's variables. These variables have an agronomic meaning and they are
early enough in several cases to serve for actual prediction. So, they may be
of potential usefulness in local Early Warning System. The comparison of yield
forecasting results based on one hand on weather station data and on the other
hand on data taken from the Tutiempo internet web site shows that meteo station
data is still better for yield forecasting process.
Key words: Burkina, forecast,
yield, local EWS, R2p, RRMSE, variable, Early.
LISTE DES SIGLES ET ABREVIATIONS
CRC : Croix Rouge Canadienne
DGPER1 : Direction
Générale de la Promotion de l'Economie Rurale
DRMAHRH/N: Direction Régionale du
Ministère de l'Agriculture, de l'Hydraulique et des Ressources
Halieutiques du Nord
FAO: Food and Agricultural Organization
FEWSNET: Famine Early Warning Systems Network
INERA : Institut de l'Environnement et de la
Recherche Agricole
INSD: Institut National des Statistiques et de
la Démographie
MAHRH : Ministère de l'Agriculture de
l'Hydraulique et des Ressources Halieutiques
MCTC : Ministère de la Culture du
Tourisme et de la Communication
MRA : Ministère des Ressources
Animales
MT : Ministère des Transports
MED : Ministère de l'Economie et du
Développement
NDVI: Normalized Difference Vegetative Index
R2 : Coefficient de
détermination entre les rendements historiques et les variables
explicatives
R2p : Coefficient de
détermination en phase de validation
RMSE: Root Mean Square Error
RRMSE: Relative Root Mean Square Error
SAP : Système d'Alerte Précoce
SISA : Système d'Information sur la
Sécurité Alimentaire
TFE : Travail de Fin d'Etudes.
1 Ce service portait le nom de Direction
Générale de la Prévision et des Statistiques Agricoles
(DGPSA).
| 
