Closing the yield gap between actual and potential wheat yields in Australia is important to meet the growing global demand for food. The identification of hotspots of the yield gap, where the ...potential for improvement is the greatest, is a necessary step towards this goal. While crop growth models are well suited to quantify potential yields, they lack the ability to provide accurate large-scale estimates of actual yields, owing to the sheer quantity of data they require for parameterisation. In this context, we sought to provide accurate estimates of actual wheat yields across the Australian wheat belt based on machine-learning regression methods, climate records and satellite image time series. Out of nine base learners and two ensembles, support vector regression with radial basis function emerged as the single best learner (root mean square error of 0.55 t ha−1 and R2 of 0.77 at the pixel level). At national scale, this model explained 73% of the yield variability observed across statistical units. Benchmark approaches based on peak Normalised Difference Vegetation Index (NDVI) and on a harvest index were largely outperformed by the machine-learning regression models (R2 < 0.46). Climate variables such as maximum temperatures and accumulated rainfall provided additional information to the 16-day NDVI time series as they significantly improved yield predictions. Variables observed up to and around the flowering period had a particularly high predictive power with additional information gained from data during grain filling. We further showed that, while all models were sensitive to a reduction of the training set size, a large majority had not reached saturation with a data set of 125 fields (2000 pixels). This indicates that additional training data are likely to further improve the skill of the models. We estimated that observations from 75 fields (1200 pixels) are required for the best single model to reach an R2 of 0.7. We contend that machine-learning regression methods applied to climate and satellite image time series can achieve reliable crop yield monitoring across years at both the pixel and the country scale. The resulting yield estimates meet the accuracy requirements for mapping the yield gap and identifying yield gap hotspots which could be targeted for further work by agricultural researchers and advisers.
•Seasonal climate forecasts narrow the prediction range for wheat yield.•In some instances the climate forecast provides more accurate yield predictions.•Consistent bias of low yield appears across ...the Australian wheat belt.•Weaknesses in calibration of forecast data is reducing prediction skill.
As a cropping season progresses yield forecasts become more reliable. Optimal management strategies however rely on early estimates of climate and yield. These estimates are usually derived from the historic range of climate variations applied to current crop conditions. Early in the season these predictions are wide ranging as they incorporate all past historic climate variability and hence a large range of possible yields. Dynamical seasonal climate models offer the opportunity to narrow this range contingent on the models having adequate predictive skill. This study explores the benefits of using a climate model over historical climate to predict wheat yield in the Australian cropping zone throughout the cropping season. We take an ensemble of daily outputs of temperature, radiation and rainfall from a seasonal climate model (POAMA) and apply a simple downscaling and calibration to align with 57 stations across the Australian cropping zone. These data are then used as an input to a crop model (APSIM) to translate seasonal conditions into a yield prediction. Simulations deploy historic weather data up to a date on which forecast data replace measured data. Here we used a range of dates (April to October) through the cropping season for the period 1981 to 2015, to determine where and when the forecast is skilful compared to using the full weather record up to harvest. The forecasts are categorised in three yield categories low (decile 1–3), average (4–7) and high (8–10) and determined to be ‘misleading’ if they predict low instead of high or vice versa. In the west and south of Australia less than 3 years in 20 give a misleading forecast in April, and less than 1 in 20 years by August. The forecast for east of Australia has less skill primarily due to a strong rainfall bias with the climate model not being able to simulate the correct daily rainfall patterns. Compared to the predictions gained from using the full range of historical climate, POAMA derived forecasts have a narrower prediction range than the climatology driven ones, however this comes at the expense of a higher number of misleading forecasts. Nevertheless, in June (August) the POAMA driven simulations have a greater than 65% (80%) chance of being in the correct or one category out, which was higher than using climatology in each region at the same lead time. The baseline set by this study demonstrates the potential utility of dynamic climate models to predict yield, which should only improve with on-going advances in climate modelling and techniques in downscaling.
•We simulate and compare sampling methods to inform in situ data collection.•We propose an index based on Moran's I to select efficient sampling routes.•Sequential exploration methods are highly ...efficient and flexible.•Existing data sets can be integrated into sequential exploration methods.•Our approach can be implemented in mobile applications for real-time support.
A precise knowledge of the crop distribution in the landscape is crucial for the agricultural sector to inform better management and logistics. Crop-type maps are often derived by the supervised classification of satellite imagery using machine learning models. The choice of data sampled during the data collection phase of building a classification model has a tremendous impact on a model's performance, and is usually collected via roadside surveys throughout the area of interest. However, the large spatial extent, and the varying accessibility to fields, often makes the acquisition of appropriate training data sets difficult. As such, in situ data are often collected on a best-effort basis, leading to inefficiencies, sub-optimal accuracies, and unnecessarily large sample sizes. This highlights the need for new more efficient tools to guide data collection. Here, we address three tasks that one commonly faces when planning to collect in situ data: which survey route to select among a set logistically feasible routes; which fields are the most relevant to collect along the chosen survey route; and how to best augment existing in situ data sets with additional observations. Our findings show that the normalised Moran's I index is a useful indicator for choosing the survey route, and that sequential exploration methods can identify the most important fields to survey on that route. The provided recommendations are flexible, overcome the main logistical constraints associated with in situ data collection, yield accurate results, and could be incorporated in a mobile application to assist data collection in real-time.
Abstract
In agriculture, sustainability is framed as an aspiration to achieve multiple goals including positive production, environmental and social outcomes. These aspirations include: increasing ...production of nutritious food; minimising risk and maximising resilience in response to climate variability, fluctuating markets and extreme weather events; minimising impacts on global warming by reducing emissions; efficiently using limited resources; minimising negative on-site and off-site impacts; preserving biodiversity on farm and in nature; and achieving positive social outcomes reflected in farmers’ incomes (revenue and profit). Here we used cropping systems simulation to assess multiple (11) sustainability indicators for 26 crop rotations to quantify their sustainability throughout Australia’s subtropical cropping zone. Results were first expressed via a series of maps quantifying the minimal environmental impacts of attributes such as N applied, N leached, runoff and GHG emissions of the 26 crop rotations while identifying the locations of the optimal rotation for each attribute. Inspection of these maps showed that different rotations were optimal, depending on both location and the attribute mapped. This observation demonstrated that an 11-way sustainability win-win across all attributes was not likely to happen anywhere in the cropping zone. However, rotations that minimised environmental impacts were often among the more profitable rotations. A more holistic visualisation of the sustainability of six contrasting sites, using sustainability polygons, confirmed that trade-offs between sustainability indicators are required and highlighted that cropping in different sites is inherently more or less sustainable, regardless of the rotations used. Given that trade-offs between the various sustainability attributes of crop rotations are unavoidable, we plotted trade-off charts to identify which rotations offer an efficient trade-off between profit and other sustainability indicators. We propose that these maps, sustainability polygons and trade-off charts can serve as boundary objects for discussions between stakeholders interested in achieving the sustainable intensification of cropping systems.
•Insufficient N accounts for a 40% loss in wheat yields in Australia.•Tillage; time of sowing; summer fallow weeds; and low seedling density also limit yields.•Frost and heat stress account for ...losses of 16%–26% relative to water-limited yield potential.•Multiple limiting factors interact differently in different seasons and environments.•Emerging management practices can boost the yield frontier by 30%.
Closing the yield gap is essential for global food security and for farmers who face increasing costs of production. Recent work showed that Australia’s wheat growers are achieving about half their water-limited yield. While quantifying the yield gap is a necessary first step towards closing them, the next step is to understand which factors constrain rainfed grain growers from achieving their water-limited yields. Here we conducted in silico experiments over 15 years at 50 weather stations to ascertain the impact on grain yield of suboptimal practices against the ‘best management practice’ rules that were used to calculate the benchmark water-limited yields. Average national losses per suboptimal practice were: the average N fertiliser application rate – 40%; conventional tillage – 33%; suboptimal weed control during the summer fallow – 26%; low seedling density – 12%; and a two week delay in sowing – 7%. Combining two suboptimal practices does not necessarily lead to an additive effect on yield. Other factors that contribute to the yield gap include biotic stresses such as plant diseases, insects and other pests, in-crop weeds and extreme weather events (e.g. floods, strong winds and hail). In addition to calculating the impact of causes of the yield gap we investigated the opportunity to lift the water-limited yield by adopting an emergent new management practice of sowing on an optimised site specific date that is earlier than the traditional sowing window as described for the currently accepted best practice. We found that this emergent practice, matched with slower maturing varieties and additional N inputs as required, has the potential to increase wheat yields nationally by 30%. Frost and heat stress accounted for losses of 16% to 26% depending on the stress function used. Allowing for the impact of frost and heat stress reduced the yield potential of both the current and emergent water-limited yields yet it did not reduce the advantage of the emergent practice.
Global food security requires eco-efficient agriculture to produce the required food and fiber products concomitant with ecologically efficient use of resources. This eco-efficiency concept is used ...to diagnose the state of agricultural production in China (irrigated wheat–maize double-cropping systems), Zimbabwe (rainfed maize systems), and Australia (rainfed wheat systems). More than 3,000 surveyed crop yields in these three countries were compared against simulated grain yields at farmer-specified levels of nitrogen (N) input. Many Australian commercial wheat farmers are both close to existing production frontiers and gain little prospective return from increasing their N input. Significant losses of N from their systems, either as nitrous oxide emissions or as nitrate leached from the soil profile, are infrequent and at low intensities relative to their level of grain production. These Australian farmers operate close to eco-efficient frontiers in regard to N, and so innovations in technologies and practices are essential to increasing their production without added economic or environmental risks. In contrast, many Chinese farmers can reduce N input without sacrificing production through more efficient use of their fertilizer input. In fact, there are real prospects for the double-cropping systems on the North China Plain to achieve both production increases and reduced environmental risks. Zimbabwean farmers have the opportunity for significant production increases by both improving their technical efficiency and increasing their level of input; however, doing so will require improved management expertise and greater access to institutional support for addressing the higher risks. This paper shows that pathways for achieving improved eco-efficiency will differ among diverse cropping systems.
Fallows are widespread in dryland cropping systems. However, timely information about their spatial extent and location remains scarce. To overcome this lack of information, we propose to classify ...fractional cover data from Sentinel-2 with biased support vector machines. Fractional cover images describe the land surface in intuitive, biophysical terms, which reduces the spectral variability within the fallow class. Biased support vector machines are a type of one-class classifiers that require labelled data for the class of interest and unlabelled data for the other classes. They allow us to extrapolate in-situ observations collected during flowering to the rest of the growing season to generate large training data sets, thereby reducing the data collection requirements. We tested this approach to monitor fallows in the northern grains region of Australia and showed that the seasonal fallow extent can be mapped with >92% accuracy both during the summer and winter seasons. The summer fallow extent can be accurately mapped as early as mid-December (1–4 months before harvest). The winter fallow extent can be accurately mapped from mid-August (2–4 months before harvest). Our method also detected emergence dates successfully, indicating the near real-time accuracy of our method. We estimated that the extent of fallow fields across the northern grains region of Australia ranged between 50% in winter 2017 and 85% in winter 2019. Our method is scalable, sensor independent and economical to run. As such, it lays the foundations for reconstructing and monitoring the cropping dynamics in Australia.
Global food security requires that grain yields continue to increase to 2050, yet yields have stalled in many developed countries. This disturbing trend has so far been only partially explained. ...Here, we show that wheat yields in Australia have stalled since 1990 and investigate the extent to which climate trends account for this observation. Based on simulation of 50 sites with quality weather data, that are representative of the agro‐ecological zones and of soil types in the grain zone, we show that water‐limited yield potential declined by 27% over a 26 year period from 1990 to 2015. We attribute this decline to reduced rainfall and to rising temperatures while the positive effect of elevated atmospheric CO2 concentrations prevented a further 4% loss relative to 1990 yields. Closer investigation of three sites revealed the nature of the simulated response of water‐limited yield to water availability, water stress and maximum temperatures. At all three sites, maximum temperature hastened time from sowing to flowering and to maturity and reduced grain number per m2 and average weight per grain. This 27% climate‐driven decline in water‐limited yield is not fully expressed in actual national yields. This is due to an unprecedented rate of technology‐driven gains closing the gap between actual and water‐limited potential yields by 25 kg ha−1 yr−1 enabling relative yields to increase from 39% in 1990 to 55% in 2015. It remains to be seen whether technology can continue to maintain current yields, let alone increase them to those required by 2050.
We investigated why Australia's wheat yields have stalled since 1990 by simulating yields at 50 representative sites in the grain zone. Yield potential declined by 27% over 26 years. We showed that this is due to declining rainfall and to rising temperatures while the positive effect of elevated atmospheric CO2 concentrations prevented a further 4% loss. This climate‐driven decline in yield potential was balanced by an unprecedented rate of adoption of technology‐driven gains closing the gap between actual and potential yields.
Most cropping systems around the world are organised around few dominant crops and a larger number of less frequent crops. While rare and infrequent crops occupy a small share of the cropped area, ...they produce ecological benefits on farmland, contribute to sustainability and help provide food and nutritional security. However, data about their location and extent derived from satellite imagery generally lack accuracy, largely owing to the class imbalance problem. Class imbalance occurs when only few instances of some classes are available for training classifiers, and leads to large error rates of the infrequent classes. In this study, we assessed the magnitude of the class imbalance problem in crop classification and evaluated balancing methods to combat it by creating synthetic minority observations or by removing majority observations. To that aim, we generated 18 unbalanced data sets from Sentinel-2 time series and crop type observations in Victoria, Australia. These data sets covered a wide range of complexity, number of classes, number of samples per class and spectral separability which enabled us to gather evidence about the benefits and drawbacks of balancing methods in various settings. Classification accuracy was assessed with two metrics: the Overall Accuracy (OA), which gives more weight to majority classes, and the G-Mean accuracy (GM), which is more sensitive to minority classes. Results showed that class imbalance explained near 40% of the accuracy variability. We found that balancing methods boosted GM by 0.01–0.54 but no single best solution emerged. The price for increasing the accuracy of minority classes was a drop in OA of a magnitude that was problem- and method-specific. We thus applied an algorithm selection method called the F-race to identify optimal balancing methods in a computationally economic fashion. Optimal balancing methods lead to maximum gain in GM and minimum loss in OA. We demonstrated that this approach either successfully identified optimal balancing methods or ones that were not significantly sub-optimal, while reducing the computational cost by up to 60%. It can readily be incorporated to operational crop classification systems with little disruption to the existing processing chains. This contribution paves the way for achieving a more comprehensive and detailed view of crop distribution and cropping sequences.
•Rare and infrequent crop classes are usually poorly mapped.•Class imbalance is one of the main reason behind their poor accuracy.•We proposed an efficient method to boost the accuracy of minority classes.•The method reduces imbalance by removing or generating synthetic data.•It offers control over the unavoidable loss in accuracy of the main classes.
► We define the concepts relevant for yield gap analysis. ► We review different methods for local and global yield gap analyses. ► Global methods are coarse and local studies use different methods. ► ...A number of methods is compared using data sets from three regions. ► Components of a protocol for global yield gap analysis with local relevance are proposed.
Yields of crops must increase substantially over the coming decades to keep pace with global food demand driven by population and income growth. Ultimately global food production capacity will be limited by the amount of land and water resources available and suitable for crop production, and by biophysical limits on crop growth. Quantifying food production capacity on every hectare of current farmland in a consistent and transparent manner is needed to inform decisions on policy, research, development and investment that aim to affect future crop yield and land use, and to inform on-ground action by local farmers through their knowledge networks. Crop production capacity can be evaluated by estimating potential yield and water-limited yield levels as benchmarks for crop production under, respectively, irrigated and rainfed conditions. The differences between these theoretical yield levels and actual farmers’ yields define the yield gaps, and precise spatially explicit knowledge about these yield gaps is essential to guide sustainable intensification of agriculture. This paper reviews methods to estimate yield gaps, with a focus on the local-to-global relevance of outcomes. Empirical methods estimate yield potential from 90 to 95th percentiles of farmers’ yields, maximum yields from experiment stations, growers’ yield contests or boundary functions; these are compared with crop simulation of potential or water-limited yields. Comparisons utilize detailed data sets from western Kenya, Nebraska (USA) and Victoria (Australia). We then review global studies, often performed by non-agricultural scientists, aimed at yield and sometimes yield gap assessment and compare several studies in terms of outcomes for regions in Nebraska, Kenya and The Netherlands. Based on our review we recommend key components for a yield gap assessment that can be applied at local to global scales. Given lack of data for some regions, the protocol recommends use of a tiered approach with preferred use of crop growth simulation models applied to relatively homogenous climate zones for which measured weather data are available. Within such zones simulations are performed for the dominant soils and cropping systems considering current spatial distribution of crops. Need for accurate agronomic and current yield data together with calibrated and validated crop models and upscaling methods is emphasized. The bottom-up application of this global protocol allows verification of estimated yield gaps with on-farm data and experiments.