Heidi Horan

Yield gap analysis of rainfed wheat demonstrates local to global relevance

Australia has a role to play in future global food security as it contributes 0·12 of global wheat exports. How much more can it contribute with current technology and varieties? The present paper seeks to quantify the gap between water-limited yield potential (Yw) and farmer yields (Ya) for wheat in Australia by implementing a new protocol developed by the Global Yield Gap and Water Productivity Atlas (GYGA) project. Results of past Australian yield gap studies are difficult to compare with studies in other countries because they were conducted using a variety of methods and at a range of scales. The GYGA project protocols were designed to facilitate comparisons among countries through the application of a consistent yet flexible methodology. This is the first implementation of GYGA protocols in a country with the high spatial and temporal climatic variability that exists in Australia. The present paper describes the application of the GYGA protocol to the whole Australian grain zone to derive estimates of rainfed wheat yield gap. The Australian grain zone was partitioned into six key agro-climatic zones (CZs) defined by the GYGA Extrapolation Domain (GYGA-ED) zonation scheme. A total of 22 Reference Weather Stations (RWS) were selected, distributed among the CZs to represent the entire Australian grain zone. The Agricultural Production Systems sIMulator (APSIM) Wheat crop model was used to simulate Yw of wheat crops for major soil types at each RWS from 1996 to 2010. Wheat varieties, agronomy and distribution of wheat cropping were held constant over the 15-year period. Locally representative dominant soils were selected for each RWS and generic sowing rules were specified based on local expertise. Actual yield (Ya) data were sourced from national agricultural data sets. To upscale Ya and Yw values from RWS to CZs and then to national scale, values were weighted according to the area of winter cereal cropping within RWS buffer zones. The national yield gap (Yg = Yw-Ya) and relative yield (Y% = 100 × Ya/Yw) were then calculated from the weighted values. The present study found that the national Yg was 2·0 tonnes (t)/ha and Y% was 47%. The analysis was extended to consider factors contributing to the yield gap. It was revealed that the RWS 15-year average Ya and Yw were strongly correlated (R = 0·76) and that RWS with higher Yw had higher Yg. Despite variable seasonal conditions, Y% was relatively stable over the 15 years. For the 22 RWS, average Yg correlated positively and strongly with average annual rainfall amount, but surprisingly it correlated poorly with RWS rainfall variability. Similarly, Y% correlated negatively but less strongly (R = 0·33) with RWS average annual rainfall, and correlated poorly with RWS rainfall variability, which raises questions about how Australian farmers manage climate risk. Interestingly a negative relationship was found between Yg and variability of Yw for the 22 RWS (R = 0·66), and a positive relationship between Y% and Yw variability (R = 0·23), which suggests that farmers in lower yielding, more variable sites are achieving yields closer to Yw. The Yg estimates appear to be quite robust in the context of estimates from other Australian studies, adding confidence to the validity of the GYGA protocol. Closing the national yield gap so that Ya is 0·80 of Yw, which is the level of Yg closure achieved consistently by the most progressive Australian farmers, would increase the average annual wheat production (20·9 million t in 1996/07 to 2010/11) by an estimated 15·3 million t, which is a 72% increase. This indicates substantial potential for Australia to increase wheat production on existing farmland areas using currently available crop varieties and farming practices and thus make a substantial contribution to achieving future global food security.

Assessing climate risks in rainfed farming using farmer experience, crop calendars and climate analysis

Climate risk assessment in cropping is generally undertaken in a top-down approach using climate records while critical farmer experience is often not accounted for. In the present study, set in south India, farmer experience of climate risk is integrated in a bottom-up participatory approach with climate data analysis. Crop calendars are used as a boundary object to identify and rank climate and weather risks faced by smallhold farmers. A semi-structured survey was conducted with experienced farmers whose income is predominantly from farming. Interviews were based on a crop calendar to indicate the timing of key weather and climate risks. The simple definition of risk as consequence × likelihood was used to establish the impact on yield as consequence and chance of occurrence in a 10-year period as likelihood. Farmers’ risk experience matches well with climate records and risk analysis. Farmers’ rankings of ‘good’ and ‘poor’ seasons also matched up well with their independently reported yield data. On average, a ‘good’ season yield was 1·5–1·65 times higher than a ‘poor’ season. The main risks for paddy rice were excess rains at harvesting and flowering and deficit rains at transplanting. For cotton, farmers identified excess rain at harvest, delayed rains at sowing and excess rain at flowering stages as events that impacted crop yield and quality. The risk assessment elicited from farmers complements climate analysis and provides some indication of thresholds for studies on climate change and seasonal forecasts. The methods and analysis presented in the present study provide an experiential bottom-up perspective and a methodology on farming in a risky rainfed climate. The methods developed in the present study provide a model for end-user engagement by meteorological agencies that strive to better target their climate information delivery.

Multimodel ensembles improve predictions of crop–environment–management interactions

A recent innovation in assessment of climate change impact on agricultural production has been to use crop multimodel ensembles (MMEs). These studies usually find large variability between individual models but that the ensemble mean (e-mean) and median (e-median) often seem to predict quite well. However, few studies have specifically been concerned with the predictive quality of those ensemble predictors. We ask what is the predictive quality of e-mean and e-median, and how does that depend on the ensemble characteristics. Our empirical results are based on five MME studies applied to wheat, using different data sets but the same 25 crop models. We show that the ensemble predictors have quite high skill and are better than most and sometimes all individual models for most groups of environments and most response variables. Mean squared error of e-mean decreases monotonically with the size of the ensemble if models are added at random, but has a minimum at usually 2-6 models if best-fit models are added first. Our theoretical results describe the ensemble using four parameters: average bias, model effect variance, environment effect variance, and interaction variance. We show analytically that mean squared error of prediction (MSEP) of e-mean will always be smaller than MSEP averaged over models and will be less than MSEP of the best model if squared bias is less than the interaction variance. If models are added to the ensemble at random, MSEP of e-mean will decrease as the inverse of ensemble size, with a minimum equal to squared bias plus interaction variance. This minimum value is not necessarily small, and so it is important to evaluate the predictive quality of e-mean for each target population of environments. These results provide new information on the advantages of ensemble predictors, but also show their limitations.