How accurate are yield estimates from crop cuts? Evidence from smallholder maize farms in Ethiopia.

Agricultural statistics and applied analyses have benefitted from moving from farmer estimates of yield to crop cut based estimates, now regarded as a gold standard. However, in practice, crop cuts and other sample-based protocols vary widely in the details of their implementations and little empirical work has documented how alternative yield estimation methods perform. Here, we undertake a well-measured experiment of multiple yield estimation methods on 237 smallholder maize plots in Amhara region, Ethiopia. We compare yield from a full plot harvest with farmer assessments and with estimates from a variety of field sampling protocols: W-walk, transect, random quadrant, random octant, center quadrant, and 3 diagonal quadrants. We find that protocol choices are important: alternative protocols vary considerably in their accuracy relative to the whole plot, with absolute mean errors ranging from 23 (farmer estimates) to 10.6 (random octant). Furthermore, while most methods approximate the sample mean reasonably well, the divergence of individual measures from true plot-level values can be considerable. We find that randomly positioned quadrants outperform systematic sampling schemes: the random octant had the best accuracy and was the most cost-effective. The nature of bias is non-classical: bias is correlated with plot size as well as with plot management characteristics. In summary, our results advocate that even "gold standard" crop cut measures should be interpreted cautiously, and more empirical work should be carried out to validate and extend our conclusions.

Implications of intra-plot heterogeneity for yield estimation accuracy: Evidence from smallholder maize systems in Ethiopia.

Intra-plot heterogeneity in yield is often observed in smallholder farming systems, although its implications for yield measurement remain under-investigated. Using a unique dataset on smallholder maize production in Ethiopia, we quantify the magnitude of inter- and intra-plot heterogeneity, describe the relationship between intra-plot heterogeneity and maize productivity, and document the implications of intra-field heterogeneity on the accuracy of alternative yield estimation protocols. Our data include five common yield estimation protocols, as well as full plot harvests of 230 smallholder maize fields. We surveyed agronomic decisions, biophysical variables, and accessibility characteristics of the surveyed fields. We quantify intra-plot heterogeneity using the coefficient of variation (CV) of stand density, cob weight, and maize grain yield. A generalized linear mixed model is used to explore the relationship between these variables and the method- and heterogeneity-dependence of yield estimation accuracy. We find inter-plot CV values ranging from 32 to 56 %, 22 to 73 % and 39 to 49 % in population density, cob weight and grain yield, respectively. Intra-plot heterogeneity constituted most of this variation, with across-method mean CV values of 41 %, 82 % and 63 %, respectively, of the total variability in population density, cob weight and grain yield. A rise in intra-plot heterogeneity of 0.5 % to 0.8 % is associated with a significant increase in yield estimation error under alternative yield estimation protocols. Regression analysis shows that interactions in agronomic decisions, input intensity and plot accessibility factors dictate intra-plot heterogeneity and method accuracy in smallholder systems. Intra-plot heterogeneity is larger than inter-plot heterogeneity in the current study area. Our analysis shows that the effect of intra-plot heterogeneity on yield estimation accuracies is method-dependent and yield estimation methods that fail to capture true intra-plot heterogeneity are more error-prone. Results of such estimations should be considered with caution when used as the basis of decision-making.

Challenges and lessons for measuring soil metrics in household surveys.

While the importance of soils in agriculture cannot be overlooked, plot level soil data remain scarce in the current data landscape. Large-scale household surveys efforts are increasing in low-income countries and assessing the accuracy, scalability and cost-effectiveness of available methods is crucial. Here, we firstly explore soil data requirements for a set of objectives that include identifying a soil constraint, improving recommendation domain studies and capturing soil metrics as covariates, or as outcomes. We then expose the lessons learned from a methodological experiment in rural Ethiopia, where different approaches - farmer's self-elicitation and miniaturized spectrometers - are compared against laboratory benchmarks for a set of soil parameters: soil texture, soil pH and soil organic C. With the exception of soil particle sizes, we find that soil parameters captured through farmer's elicitation do not converge with objective metrics. Miniaturized spectrometers can provide reasonably accurate data for the identification of soil constraints - soil acidity, low organic C or sandy soils. Approximate quantitative predictions can also be delivered for soil pH (R2 = 0.72) and organic C (R2 = 0.60). The additional costs of plot sampling and analysis are in the range of $19-$23 per sample, with the additional percentage of plots with correct data equivalent to 10% for the identification of sandy soils, 75% for low organic C and 89% of acidic soils.

VARIETAL IDENTIFICATION IN HOUSEHOLD SURVEYS: RESULTS FROM THREE HOUSEHOLD-BASED METHODS AGAINST THE BENCHMARK OF DNA FINGERPRINTING IN SOUTHERN ETHIOPIA.

Accurate crop varietal identification is the backbone of any high-quality assessment of outcomes and impacts. Sweetpotato (Ipomoea batatas) varieties have important nutritional differences, and there is a strong interest to identify nutritionally superior varieties for dissemination. In agricultural household surveys, such information is often collected based on the farmer's self-report. In this article, we present the results of a data capture experiment on sweet potato varietal identification in southern Ethiopia. Three household-based methods of identifying varietal adoption are tested against the benchmark of DNA fingerprinting: (A) Elicitation from farmers with basic questions for the most widely planted variety; (B) Farmer elicitation on five sweet potato phenotypic attributes by showing a visual-aid protocol; and (C) Enumerator recording observations on five sweet potato phenotypic attributes using a visual-aid protocol and visiting the field. In total, 20% of farmers identified a variety as improved when in fact it was local and 19% identified a variety as local when it was in fact improved. The variety names given by farmers delivered inconsistent and inaccurate varietal identities. Visual-aid protocols employed in methods B and C were better than those in method A, but greatly underestimated the adoption estimates given by the DNA fingerprinting method. Our results suggest that estimating the adoption of improved varieties with methods based on farmer self-reports is questionable and point towards a wider use of DNA fingerprinting in adoption and impact assessments.

Soil water management practices (terraces) helped to mitigate the 2015 drought in Ethiopia.

While the benefits of soil water management practices relative to soil erosion have been extensively documented, evidence regarding their effect on yields is inconclusive. Following a strong El-Niño, some regions of Ethiopia experienced major droughts during the 2015/16 agricultural season. Using the propensity scores method on a nationally representative survey in Ethiopia, this study investigates the effect of two widely adopted soil water management practices - terraces and contour bunds - on yields and assesses their potential to mitigate the effects of climate change. It is shown that at the national level, terraced plots have slightly lower yields than non-terraced plots. However, data support the hypothesis that terraced plots acted as a buffer against the 2015 Ethiopian drought, while contour bunds did not. This study provides evidence that terraces have the potential to help farmer deal with current climate risks. These results can inform the design of climate change adaptation policies and improve targeting of soil water management practices in Ethiopia.

Evaluation of a miniaturized NIR spectrometer for cultivar identification: The case of barley, chickpea and sorghum in Ethiopia.

Crop cultivar identification is fundamental for agricultural research, industry and policies. This paper investigates the feasibility of using visible/near infrared hyperspectral data collected with a miniaturized NIR spectrometer to identify cultivars of barley, chickpea and sorghum in the context of Ethiopia. A total of 2650 grains of barley, chickpea and sorghum cultivars were scanned using the SCIO, a recently released miniaturized NIR spectrometer. The effects of data preprocessing techniques and choosing a machine learning algorithm on distinguishing cultivars are further evaluated. Predictive multiclass models of 24 barley cultivars, 19 chickpea cultivars and 10 sorghum cultivars delivered an accuracy of 89%, 96% and 87% on hold-out sample. The Support Vector Machine (SVM) and Partial least squares discriminant analysis (PLS-DA) algorithms consistently outperformed other algorithms. Several cultivars, believed to be widely adopted in Ethiopia, were identified with perfect accuracy. These results advance the discussion on cultivar identification survey methods by demonstrating that miniaturized NIR spectrometers represent a low-cost, rapid and viable tool. We further discuss the potential utility of the method for adoption surveys, field-scale agronomic studies, socio-economic impact assessments and value chain quality control. Finally, we provide a free tool for R to easily carry out crop cultivar identification and measure uncertainty based on spectral data.

On the Ground or in the Air? A Methodological Experiment on Crop Residue Cover Measurement in Ethiopia.

Maintaining permanent coverage of the soil using crop residues is an important and commonly recommended practice in conservation agriculture. Measuring this practice is an essential step in improving knowledge about the adoption and impact of conservation agriculture. Different data collection methods can be implemented to capture the field level crop residue coverage for a given plot, each with its own implication on survey budget, implementation speed and respondent and interviewer burden. In this paper, six alternative methods of crop residue coverage measurement are tested among the same sample of rural households in Ethiopia. The relative accuracy of these methods are compared against a benchmark, the line-transect method. The alternative methods compared against the benchmark include: (i) interviewee (respondent) estimation; (ii) enumerator estimation visiting the field; (iii) interviewee with visual-aid without visiting the field; (iv) enumerator with visual-aid visiting the field; (v) field picture collected with a drone and analyzed with image-processing methods and (vi) satellite picture of the field analyzed with remote sensing methods. Results of the methodological experiment show that survey-based methods tend to underestimate field residue cover. When quantitative data on cover are needed, the best estimates are provided by visual-aid protocols. For categorical analysis (i.e., >30% cover or not), visual-aid protocols and remote sensing methods perform equally well. Among survey-based methods, the strongest correlates of measurement errors are total farm size, field size, distance, and slope. Results deliver a ranking of measurement options that can inform survey practitioners and researchers.