Date:

Tuesday, September 20, 2016

Type:

MSc

Keywords:

Ground cover, light interception, LAI, maize, sorghum, millet, cotton

Mentor(s):

A.G.T. Schut, P.C.S. Traoré

Country:

Mali/Wageningen

Abstract:

This report is an exploratory research into the behavioural patterns of light interception (LI), fraction ground cover of green material (GC), leaf area index (LAI) and eBee perpendicular vegetation index (eBeePVI), which is comparable to the perpendicular vegetation index. The main goal for the project is to understand how these variables relate relative to each other in the field during the growing season, how they relate to measurements done by an Unmanned Aerial Vehicle (UAV) and how they can be used in a system to predict crop growth and yield for smallholder farmers in sub-Saharan Africa. It is important to be able to model the future yield during the growing season, and as light interception is one of the most important factors influencing plant growth it could be a good predictor for yield. For these models to work accurately, it is essential to understand the relations that govern light interception and thus biomass production for the crops. These relations could be much more useful if all crops can be defined by the same model, this would simplify the analysis of the measurements. This project focussed on data retrieved from a trial in Samanko (Mali), situated approximately 15 km from the capital, Bamako. The trial consisted of seven crops with two plots each and each plot consisted of ten quadrats. From these seven crops, six were analysed for this project: 2 sorghum varieties (Tièble and Soumba), peanut, maize, millet and cotton. The following relations were analysed in this project: LI vs. GC, LI vs. eBeePVI, GC vs. eBeePVI and LI vs. LAI. Two different models were fitted; a linear model, and the well-known negative exponential formula, the Lambert-Beer formula of attenuation. The negative exponential model explained more of the variation of the relation between LI and GC than the linear model. When the exponential function was analysed, differences between sorghum Tièble, maize, millet and cotton were not significant, peanut and sorghum Soumba were not statistically different from the previous group, but they were different between each other. For LI and eBeePVI, when considering the coefficient of determination it was less clear which relationship was best describing the data, the R2 values were similar and no decision could be made regarding this relation. The RMSE values on the other hand were much lower for the negative exponential function. Both models showed high similarities between the crops, for the linear model slopes were statistically not different. Only millet and sorghum Soumba differed in the intercept; only peanut and sorghum Soumba differed in their slope when the transformed negative exponential function was analysed. Although the R2 values for eBeePVI vs. GC relationships were similar between crop types, the R2 values for the linear function were higher in all cases; this is in accordance with the RMSE error values which were all lower for the linear function except for peanut. No differences were found between the crops for both slope and intercept in the eBeePVI vs. GC relationship. Finally, for the LI vs. LAI relation the data from the leaf surface of the quadrats was suspected of being erroneous as they delivered values outside of the range of possibilities. This relation is always described in literature as being governed by the negative exponential formula, which results in a model without intercept. To cope with the erroneous data for the LAI values, an intercept was allowed in order to be able to fit a model. No real conclusion could be drawn based on the R2 values resulting from this analysis. The RMSE did not give a conclusive answer, as the negative exponential model had lower values across all crops except for maize when no intercept was allowed, and for the linear model the RMSE values were lower in all cases except for sorghum Tièble when an intercept was allowed. Both models showed differences between the crops, mainly peanut and cotton grouped together on one side and the sorghum varieties on the other side with maize and millet in between. Ultimately, the purpose of the project for which this thesis fills in a small part, is to be able to use the current state of the fields to establish recurrent and accurate plant production predictions. Farmers, their communities and governments could anticipate by tuning their management on the information accurate predictions provides.

Address:

PPS, Radix building, Wageningen

Email: