Untangling crop management and environmental influences on wheat yield variability in Bangladesh: An application of non-parametric approaches

Krupnik, Timothy J.; Ahmed, Zia; Timsina, Jagadish; Yasmin, Samina; Hossain, Farhad; Mamun, Abdullah Al; Mridha, Aminul Islam; McDonald, Andrew J.

doi:10.1016/j.agsy.2015.05.007

Cited by 46 publications

(20 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The GLMM and CART analysis identified a number of agroecology‐specific factors influencing maize yield gaps. The findings are consistent with previous recommendations for site‐specific extension services on soil and crop management strategies to reduce yield gaps (Banerjee et al, ; Krupnik et al, ; Tamene, Mponela, Ndengu, & Kihara, ; Yengoh, ).…”

Section: Discussionsupporting

confidence: 91%

“…Linear regression and correlation methods have been widely used in yield gap studies to show specific factors influencing crop yields (Krupnik et al, 2015;Mackay et al, 2011;Neumann, Verburg, Stehfest, & Muller, 2010;Sawasawa, 2003). However, the heterogeneity in smallholder farms is likely to result in high spatial variability in yield gaps and their causes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Soil and management‐related factors contributing to maize yield gaps in western Kenya

Munialo

Dahlin

Onyango

et al. 2019

Food and Energy Security

View full text Add to dashboard Cite

The solution to reducing existing yield gaps on smallholder farms lies in understanding factors limiting yield in areas with agricultural intensification potential. This study applied an integrated analysis approach comprising Classification and Regression Tree (CART), generalized linear mixed model (GLMM), and factor analysis (FA), to explain soil and management‐related factors influencing maize yield gaps, in order to enhance yields. The study was conducted in Mukuyu and Shikomoli in western Kenya, sites with, respectively, high and low agroecological potential regarding soil fertility. Maize yield gaps were quantified by comparing yields on the 90th percentile of farms to yields determined in 189 fields on 70 randomly sampled smallholdings. Soil and management‐related factors were determined at early and late maize development stages. Maize yield on the 90th percentile of farms in Mukuyu and Shikomoli was 5.1 and 4.8 t/ha, respectively, and the average yield gap was 1.8 and 2.6 t/ha, representing 35% and 54% unachieved yield for Mukuyu and Shikomoli, respectively. In FA, soil was revealed to be the main factor influencing maize yield gaps at both sites, rather than management‐related variables. The CART method identified maize density, chlorophyll values, maize height, and depth to compact layer as consistent factors affecting yield at both sites, while GLMM identified soil texture (silt content) as important. According to CART, weed cover at early stages and maize density at late stages were the most limiting factor in maize production in Mukuyu and Shikomoli, respectively. Generalized linear mixed model analysis identified agroecology‐specific factors influencing maize yield gaps as soil‐available phosphorus and zinc, plus weed pressure at early maize stages in Mukuyu, and plus soil cation exchange capacity and exchangeable magnesium in Shikomoli. Through an integrated approach, it was possible to identify both consistent and agroecology‐specific factors limiting crop yields. This can increase the applicability of the findings to smallholder farms.

show abstract

Section: Discussionsupporting

confidence: 91%

Section: Introductionmentioning

confidence: 99%

Soil and management‐related factors contributing to maize yield gaps in western Kenya

Munialo

Dahlin

Onyango

et al. 2019

Food and Energy Security

View full text Add to dashboard Cite

show abstract

“…Classical approaches to model such data, such as ANOVA, regression, or (generalized linear) mixed models are often hampered by the high number of potentially available predictor variables that interact in unexpected ways, making it hard to identify important determinants or combinations of determinants of post-harvest losses. Complex nonlinear relationships are often missed without explicit pre-specification thus complicating the disclosure of new features of the data and may easily lead to spurious conclusions (Kaminski and Christiaensen 2014;Krupnik et al 2015). This is also confirmed by Flack and Chang (1987), who demonstrated that variable selection within a set of noisy predictor variables frequently resulted in selected subsets of noise variables.…”

Section: Discussionsupporting

confidence: 54%

Post-harvest management and post-harvest losses of cereals in Ethiopia

Hengsdijk

Boer

2017

Food Sec.

View full text Add to dashboard Cite

Recent and systematic evidence on the magnitude of post-harvest losses in sub-Saharan Africa is scarce, hindering the identification of interventions to reduce losses. Here, we unlock standardized and systematically collected information on post-harvest management and farmer-reported post-harvest loss estimates from the Living Standards Measurement Study -Integrated Surveys in Agriculture. Using the data from Ethiopia, the objective is to disentangle factors that induce or relate to post-harvest losses in cereals. The data of approximately 2500 households and 5500 cereal records were analysed. Cereal post-harvest loss was reported by only 10% of these households. The average self-reported post-harvest loss was 24%. Rodents and other pests were most frequently reported to cause these losses. Adoption of improved storage methods was limited and most cereals were stored inside the house in bags. Random Forests (RF) was applied to gain insight into factors and conditions favouring post-harvest losses. Application of RF explained 31% of the variation in postharvest losses reported by farmers. Three major factors associated with post-harvest losses were the distance of the household dwelling to the nearest market, the distance of the household dwelling to the main road, and average annual rainfall. Losses increased the further households were located from a market or main road, and losses also tended to decrease with higher rainfall. The standardized and nationally representative survey data from Ethiopia used were a good starting point for modelling post-harvest losses but the finally available information appeared to be partial. Therefore, this paper calls for better data collection, which could help to better target interventions needed to reduce post-harvest losses.

show abstract

“…Their influence on yield can be evaluated only based on representative data collected on farms through wellprepared surveys (Ferraro et al, 2009;Zheng et al, 2009;Delmotte et al, 2011).When the data cover a wide time range-several years, for instance-with different environmental conditions and a considerable number of diversified farms, classic statistical methods might not be adequate to find cause-and-effect relationships between all variables and the outcome (yield). In this case data mining methods such as CART (Classification And Regression Trees) might be very useful (Krupnik et al, 2015;Sileshi et al, 2010;Roel et al, 2007;Breiman et al, 1984, Dacko et al, 2016. There is a great difference between the potential and real yield of winter wheat in Poland.…”

Section: Introductionmentioning

confidence: 99%

“…CART cannot be used for precise prediction. However, it allows the user to determine the most important variables and their relative performance in explaining variation in yields of a crop Krupnik et al (2015). studied the influence of crop management and environment on wheat yield variation in…”

mentioning

confidence: 99%

Use of classification and regression trees (CART) for analyzing determinants of winter wheat yield variation among fields in Poland

Iwańska

Oleksy

Dacko

et al. 2018

Biometrical Letters

View full text Add to dashboard Cite

SummaryWheat is one of the modern world’s staple food sources. Its production requires good environmental conditions, which are not always available. However, agricultural practices may mitigate the effects of unfavorable weather or poor-quality soils. The influence of environmental and crop management variables on yield can be evaluated only based on representative long-term data collected on farms through well-prepared surveys.The authors of this work analyzed variation in winter wheat yield among 3868 fields in western and eastern Poland for 12 years, as dependent on both soil/weather and crop management factors, using the classification and regression tree (CART) method. The most important crop management deficiencies which may cause low wheat yields are insufficient use of fungicides, phosphorus deficiency, non-optimal date of sowing, poor quality of seeds, failure to apply herbicides, lack of crop rotation, and use of cultivars of unknown origin not suitable for the region. Environmental variables of great importance for the obtaining of high yields include large farm size (10 ha or larger) and good-quality soils with stable pH. This study makes it possible to propose strategies supporting more effective winter wheat production based on the identification of characteristics that are crucial for wheat cultivation in a given region.

show abstract

Untangling crop management and environmental influences on wheat yield variability in Bangladesh: An application of non-parametric approaches

Cited by 46 publications

References 48 publications

Soil and management‐related factors contributing to maize yield gaps in western Kenya

Soil and management‐related factors contributing to maize yield gaps in western Kenya

Post-harvest management and post-harvest losses of cereals in Ethiopia

Use of classification and regression trees (CART) for analyzing determinants of winter wheat yield variation among fields in Poland

Contact Info

Product

Resources

About