Previous PageTable Of ContentsNext Page

Yield Gap Analysis: quantifying the gap between farmers’ wheat yields and water limited yield potential

Zvi Hochman1, David Gobbett2, Dean Holzworth3, Oswald Marinoni1, Tim McClelland4, Harm van Rees5, Javier Navarro Garcia1 and Heidi Horan1

1 CSIRO Ecosystem Sciences/Sustainable Agriculture Flagship, EcoSciences Precinct, 41 Boggo Road, Dutton Park, QLD 4102, Australia

2 CSIRO Ecosystem Sciences/Sustainable Agriculture Flagship, PMB 2, Glen Osmond, SA 5064, Australia

3 CSIRO Ecosystem Sciences/Sustainable Agriculture Flagship, Toowoomba Qld 4350, Australia

4 BCG PO Box 85, Birchip VIC 3483 Australia

5 Cropfacts P/L, 69 Rooney Rd., RSD Strathfieldsaye Victoria 3551 Australia


As agronomists striving to help famers feed a growing world population in the coming decades, we should not underestimate the importance of the gap between yields that are currently achieved by farmers (Ya) and those potentially attainable in rainfed farming systems (Yw). The first step towards reducing this yield gap (Yg) is to obtain realistic estimates of their magnitude and their spatial and temporal variability. Previous attempts to estimate yield gaps at scales ranging from the local to the global have had low agronomic relevance as they failed to adequately represent the weather, soils and crop management at the local scale. In this paper we outline a framework for assessing yield gaps at a range of scales from 1.1 km2 to the whole region by combining data from various sources including ABS statistics, remotely sensed NVDI data, soil maps, soil characterisation data, and climate data to estimate Ya and Yw as applied to wheat yields in the Wimmera, Victoria region. We show that there is a large exploitable yield gap in the Wimmera and that bridging it would increase the region’s average annual wheat production by 43%. Mapping the yield gap indicates what parts of the region should be targeted for further investigation into bridging this gap.

Key Words

Food security; yield potential; wheat; yield map; assessment; simulation.


Exploiting the gap between yields currently achieved on farms and those that can be achieved by using the best adapted crop varieties and best crop and land management practices is a key pathway to overcoming the considerable challenge of feeding more than 9 billion people by 2050. Knowledge of the size of the yield gap and where the greatest exploitable yield gaps exist is a powerful tool for determining research priorities and for challenging farmers to lift their productivity. Potential yield (Yp) is the maximum yield that can be reached by a crop in a given environment, as determined, for example, by simulation models with plausible physiological and agronomic assumptions. Actual farm yield (Ya) is defined as reflecting farmers’ natural resource endowment, their access to technology, and their skill and exposure to real market economics. Given that 70% of the world’s wheat cropping area is rainfed, water-limited yield potential (Yw) is an important concept. Yw is defined as the yield of an adapted crop variety when grown under rainfed, favourable conditions without growth limitations from nutrients, pests or diseases. For rainfed crops the yield gap (Yg) is the difference between Yw and Ya. Alternatively, the concept of relative yield (Y%) is used where Y% = 100 x Ya/Yw. As a farmer’s yield approaches Yw, the law of diminishing returns would suggest that further gains will become more difficult and less economically attractive to achieve. Consequently, average farm yields can be expected to peak at Y% = 75%. We therefore distinguish between the absolute yield gap (Yg) and the more pragmatic exploitable yield gap.

In this paper we focus on a case study of rainfed wheat in the Wimmera region (as defined by ABS) of Victoria, Australia, applying a framework that uses various sources of data (local statistical data, satellite data, spatially distributed but site specific historical weather data, soil characterisation data, soil maps) and calculation methods (simulation, GIS mapping) to derive the components of the yield gap. These components (Ya, Yw, cropped area, yield maps) are integrated to achieve spatially and temporally explicit estimates of Yg. The framework includes ad hoc ground testing of Yg and its component parts based on field level monitoring and farmer records.


Ya are estimated from statistical data gathered at various levels of spatial detail. In Australia, farmers’ crop yields are collected annually by the Australian Bureau of Agricultural and Resource Economics and Sciences (ABARES) through its Australian Agricultural and Grazing Industries Survey (AAGIS) while the more comprehensive Agricultural Census data are collected every five years by the Australian Bureau of Statistics (ABS 2009). The annual data for wheat are aggregated up from individual farms to SLAs to 11 regions in three ago-climatic zones. The 11 regions in the Australian wheat-sheep zone are viewed by ABARES as the smallest unit for which their annual survey is designed to produce reliable wheat crop estimates (ABS 2009). Remotely sensed Normalised Difference Vegetation Index (NDVI) from the MODIS satellite captures the greenness of a pixel and these are combined with statistical data to provide a map of Ya at 1.1 km2 resolution.

Simulation, using a locally validated cropping system model, can be used to determine Yw at any geospatial location provided that a minimal data set is available including: 1. Daily weather data including minimum and maximum temperatures, rainfall, evaporation and solar radiation; 2. Soil characterisation that matches the local soil type, especially with respect to its water holding capacity; 3. Estimates of soil water status at the time of sowing; 4. A specified best practice that can be consistently applied with regard to time of sowing, sowing density, variety and dates and rates for application of nitrogen fertilizer. Yw is calculated from APSIM simulations based on weather data from the Silo website; a map of the plant available water capacity (PAWC) of soils in the region from the ASRIS national soils database; characterised soils from the APSoil database of over 700 APSIM reference soils were selected to represent the range of PAWC values of soils in the Wimmera and matched to mapped areas. On-farm data for ground testing were derived from a number of sources. Here we present results from: 1. Crop contest yield data from Longerenong Cropping Challenge, 2. Yield Prophet subscribers’ yield and water use efficiency (WUE) data. For a full description of the assessment framework and methods used in this study see Hochman et al. (2012).


Estimating and mapping farmers’ yields (Ya)

The Wimmera region and its location on an inserted map of Australia; the area cropped to winter cereals; the location of weather stations and a map of the estimated soil PAWC values in the Wimmera region are indicated in Figures 1a, 1b and 1c. Based on AgSurf data for the 20 years from 1990 to 2009 the average yield per hectare (2.21 t/ha) varied considerably from year to year with a standard deviation (Sd) of 0.84 t/ha illustrating the impact of climate variability on actual yields. An estimate of the spatial distribution of wheat grain yields was made for 2005, when the latest available Agricultural Census data provided crop estimates at the SLA level. Mean SLA yields ranged between 2.1 t/ha in Yarriambiack – North and 3.02 t/ha in N. Grampians – St Arnaud. Further disaggregating the 2005 yield values by using the remotely sensed NDVI data resulted in a detailed spatial map of Ya in the Wimmera region (Figure 1d).

Figure 1. Maps of the Wimmera region with (a) Statistical Local Areas (SLAs) and towns, (b) location of cereal cropping areas and weather stations, (c) Soil plant available water characteristics, and (d) spatial distribution of farm yields in 2005.

Estimating and mapping the yield gap (Yg)

Simulation of 56 weather stations by 5 soil types over 26 years provided the data for annual Yw maps. These maps captured the spatial and temporal variation in Yw in the Wimmera and were produced using Inverse Distance Weighting (IDW) interpolation of the crop simulation results. The annual Yw maps for the whole region were compared it to the region’s statistical Ya for each year from 1990 to 2009 to calculate Yg and Y%. Annually estimated yield gaps ranged from 0.66 t/ha in 2006 to 4.12 t/ha in 1992 with an average Yg of 2.00 t/ha (Std = 0.98). Y% ranged from 26.3% in 2002 to 77.9% in 2009 with an average Y% of 52.7% (Std = 12.7%). Taking a relative yield of 75% as the exploitable yield, we note that 2009 was the only year in which we did not observe an exploitable yield gap for the region as a whole. Given that Ya data for the 2005 season was available at the finer SLA resolution we were able to estimate Yg and Y% at the SLA level. Yg was largest for N. Grampians – Stawell (4.41 t/ha) and least for Yarriambiack – North (0.60 t/ha). Similarly, Y% was highest (77.7%) at Yarriambiack – North and lowest (35.5%) at N. Grampians – Stawell. By combining the data from Figures 1(d) and the Yw map for 2005, Yg can be calculated for each 1.1 km cell for the year 2005 (Figure 2) to show in greater detail where the largest gaps are likely to exist.

Ground testing water limited yield (Yw) and Yield gap (Yg) results

The Yw value for the grid cell covering the ‘Longerenong Challenge’ site in 2009 was 5.23 t/ha. The 14 yield outcomes in the 2009 ‘Longerenong Challenge’ competition ranged from 0.95 to 4.93 t/ha. The winning yield of 4.93 t/ha (followed closely by a 4.82 t/ha yield) confirms Yw for this grid cell in 2009. For 30 fields monitored for WUE in 2007, available water averaged at 234 mm (Sd = 52) and grain yields averaged 1.98 t/ha (Sd = 0.77). The average Yw for these fields, based on the WUE boundary formula was 3.5 t/ha (Sd = 1.44).

Figure 2. Spatial distribution of yield gaps (Yg) in the Wimmera in the 2005 season.

The average gap between actual yields and potential yields was 1.51 t/ha (Sd = 0.79). This on-ground yield gap assessment based on monitoring and WUE boundary calculations was close to the calculated Yg value (1.49 t/ha) based on simulation and ABARES data for the whole region in 2007. As such the WUE frontier method of estimating Yg provided strong on-ground support for the simulation based calculation method, at least in a particular year.


Spatial analysis of the yield gap in the Wimmera region indicated that while Yarriambiack – North and Hindmarsh SLAs are close to achieving the exploitable yield, N. Grampians – Stawell and Horsham have the greatest potential for yield improvements. Surprisingly, the areas with highest Yw show the highest yield gap in both absolute and relative terms. This may reflect a need for farmers in the more marginal areas to invest the necessary inputs to maximise production in a good season, while farmers in the higher Yw areas are profitable even at lower than optimal input levels and are therefore more risk averse due to their concern with downside risk in case of extreme events. Overall, farmers in this region can increase the average annual wheat produced by 43% (from 1.09 M tonnes to 1.55 M tonnes). This potential increase in dryland grain production supports claims that bridging the attainable yield gap is an important pathway to future global food security. We propose to repeat this study in the remaining 10 cropping regions in Australia as part of an international effort to develop a worldwide yield gap atlas (van Ittersum et al. 2012).


Hochman Z, Gobbett D, Holzworth D, McClelland T, van Rees H, Marinoni O, Navarro Garcia J, Horan H, 2012. Quantifying yield gaps in rainfed cropping systems: a case study of wheat in Australia. Field Crops Research (In press).

van Ittersum MK, Cassman KG, Grassini P, Wolf J, Tittonell P, Hochman Z, 2012. Yield gap analysis with local to global relevance – a review. Field Crops Research (In press).

Previous PageTop Of PageNext Page