|SMITH, HARRISON - University Of Arkansas
|NALLEY, LANIER - University Of Arkansas
|SCHMIDT, AXEL - Catholic Relief Services
|TURMEL, MARIE - Catholic Relief Services
Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/9/2023
Publication Date: 6/15/2023
Citation: Smith, H., Ashworth, A.J., Nalley, L., Schmidt, A., Turmel, M., Owens, P.R. 2023. Boundary line analysis and machine learning models to identify critical soil values for major crops in Guatemala. Agronomy Journal. 1-17. https://doi.org/10.1002/agj2.21412.
Interpretive Summary: Knowledge, data, and understanding agronomic production potential are key for advancing agriculture and society, as well as for closing yield gaps (i.e., the difference between actual and potential yield). Although, not all producers in developing nations, such as the Northern Triangle of Central America, have basic agronomic information. For instance, critical soil nutrient thresholds for developing fertilizer recommendations for major crops in countries such as Guatemala are currently lacking. The agricultural sector plays an important role in the economy in Guatemala, with agriculture accounting for 10% of their gross domestic product and 30% of total employment. Therefore, minor yield increases for major crops, such as maize, common bean, and coffee could have a major economic impact in Guatemala. The objectives for this study are to use data from 664 farms in Guatemala to 1) identify optimum soil property conditions for maize, beans, and coffee in Guatemala, 2) use crop price and boundary line estimates to develop economically sound fertilization guidelines, and 3) to analyze the drivers of yield outcomes and assess their relative importance. Our work shows how how crop sale price can be paired with crop response data to calculate yield and income gains from optimized fertility management. These results are a critical first step in developing economically sound fertility recommendations for the nation of Guatemala, and, if applied, could improve yields and increase profits and ultimately close yield gaps in this nation. Additionally, we found that weather variables within a year like precipitation and temperature may drive crop response to fertilizers (beyond rate), particularly in the dry corridor. The results presented here are a first step for data-driven and economically sound fertility recommendations in Guatemala, and contribute to a better understanding of the drivers of yield for maize, common bean, and coffee growers in this region.
Technical Abstract: Accurate and economically sound soil fertility recommendations are critical for ensuring economically sound food production for smallholder farmers. However, such recommendations are lacking in many areas due to poor availability of soil and crop response data. This study addresses this challenge by 1) using boundary line analysis to evaluate crop response to fertility applications on 644 farms in Guatemala spanning 2016-2020 to identify optimal soil property conditions for maize grain (Zea mays L.), common bean (Phaseolus vulgaris L.), and coffee (Coffea arabica L.) production in Guatemala, 2) using crop price information to develop economically sound fertilization guidelines, and 3) analyzing the drivers of yield outcomes and assess their relative importance with random forest regression models. Results demonstrate that a majority of plots currently have sub-optimal soil nutrient levels. Through yield maximizing nutrient usage, a majority of farms could see increased revenue. In addition, random forest regression underscored the central role of climate in shaping yield outcomes, highlighting the critical need for climate-smart adaptations in the region. The pairing of boundary line analysis with random forest provided unique but complementary information about the factors that drive major crop yields in Guatemalan farms. This approach is an effective method for development of fertility recommendations in Guatemala, and could be replicated in other areas where critical soil and yield information is currently lacking.