Interpreting yield variation in commercial production of crops: An alternative to strategies for bridging the yield gap
As part of the modeling workshops started by the Decision and Policy Analysis (DAPA) research area few months ago, last week Daniel Jiménez gave a presentation about the experiences of the Site –specific agriculture team “Interpreting Yield variation in commercial production of crops” over the last 8 years. The implementation of a range of methods was illustrated to researchers from DAPA and agrobiodiversity research areas.
In the beginning of the presentation, Daniel introduced the concepts of his research and the challenges of working with information on commercial production of crops. Furthermore, it was explained the characteristics of this information, and some of the strategies followed by the team to deal with information that unlike to the data that can be obtained from field-trials and yield ceilings; can reach thousands of records, and is difficult to process only by means of parametric- approaches. Figure 1 illustrates that in many cases, when collecting commercial production most farmers present low yields.
Three case studies
Three cases studies were presented, for each approach, the advantages and disadvantages were shown.
1. Andean blackberry based on ANNs:
The first case study was based only on non-parametric models. Multilayer perceptron (MLP) and Self-Organizing Maps (SOM) were used as computational models in the identification and visualization of the most important variables for modeling the production of Andean blackberry. Analysis of Andean blackberry (Rubus glaucus) production models obtained by means of artificial neural networks exploiting information collected by small-scale growers in Colombia and publicly available meteorological data.
Parametric and non-parametric regression models were used to identify those environmental factors most closely associated with the product/fruit yield, then these factors were used to define three Homogeneous Environmental Conditions (HECs).
Parametric models were used to analyze the data within a hierarchical framework of clusters of HECs, locations and farms. HECs were used as a proxy for the climatic conditions, location, as a proxy for cultural factors associated with a geographic region, and farm as proxy for management skills. Interpretation of commercial production information: A case study of lulo (Solanum quitoense), an under-researched Andean fruit
3. Plantain and Avocado
Over the last three years, the site-specific team had the opportunity of collecting information on plantain and avocado commercial production. The information included: crop yield, GPS, climate, soils and crop management. The team has compiled and analyzed as much information as possible over a wide range of heterogeneous and socioeconomic conditions where these crops are grown (Figure 4 ). Parametric and semi-parametric approaches that suit to handle the analytical requirements of this commercial information have been presented. Hierarchical clustering on principal components to cluster climate, CATPCA to cluster soil data, and generalized linear models to find the likely effect on production of implementing the management practices found as determinant for the production of these crops. It is noteworthy that over the last 5 years the team has provided scientific evidence of yield gap between farms in similar environmental conditions. Hence the implementation or their findings can be regarded as an alternative to the methods studied to bridge the yield gap; through an approach that seems to be a revolutionary way of putting agronomy into practice based on the analysis of information on commercial production of crops.
These discussion spaces/ workshops have generated very positive results, not only from a scientific point of view, but allowing researchers from different CIAT’s areas to share their experiences and receive feedback from other groups.
By Hugo Andrés Dorado and Daniel Jiménez