Sugarcane breeding is resource-intensive and time-consuming, and could benefit substantially from the integration of aerial phenotyping (AP) for rapidly identifying genotypes with superior yield traits. The study aimed to assess the feasibility of using AP to enhance sugarcane breeding by rapidly identifying genotypes with superior yield traits. The specific objectives of the study were to: (1) assess the impacts of canopy cover and stomatal conductance on stalk dry mass yield (SDM); (2) assess the feasibility of estimating these traits with aerially sensed normalized difference vegetation index (NDVI) and canopy temperature (Tc); (3) evaluate the potential for predicting SDM from NDVI and Tc; (4) formulate best AP procedures. The study comprised a replicated field trial near Komatipoort, South Africa, with 54 genotypes grown under well-watered and water deficit conditions. Traits were measured on the ground (canopy cover and stomatal conductance) and remotely sensed from the air with a drone (NDVI and Tc) throughout the plant and first ratoon crops, and SDM was measured at harvest. Measurements were categorized by crop water status and extent of canopy cover, and phenotypic trait correlations were analyzed for these different categories. The study confirmed canopy cover and stomatal conductance as influential traits for determining SDM. Canopy cover could be used as a proxy for identifying high- and low-yielding genotypes early on in water stress-free crops. Findings suggest that high stomatal conductance benefits well-watered crops, while relatively low conductance could be advantageous in dry environments, though further investigation is needed. Canopy cover was predicted well from NDVI at partial canopy for well-watered crops, while the prediction of stomatal conductance from Tc lacked reliability. It was concluded that NDVI and Tc could be used to identify high- and low-yielding genotypes when measured earlier on in the growth cycle for well-watered crops. Results also showed potential for using water treatment differences in Tc and SDM to identify drought tolerant genotypes. Lastly, the study highlighted methodological challenges and insights for future agronomic trait prediction using AP techniques. The findings of this study will be used in further testing in the early stages of the breeding programme along with the breeding populations, ultimately helping to manage breeding strategies for target environments. This has the potential to enhance breeding efficiency and ultimately genetic gains towards productive sugarcane cultivars for the future.

Spatial information on crop productivity and resource use is required to enable efficient sugarcane production with limited resources and under a changing climate. The objective of this study was to estimate biomass, sugar and ethanol yields for high-sucrose (HS) and high-fibre (HF) sugarcane cultivars for current and future climate in water limited South Africa. An upgraded version of the Canegro sugarcane model, calibrated for a HS and HF cultivar, was used to simulate biomass component yields for 1,986 agroclimatic zones. Ethanol yields were calculated from simulated biomass fractions and theoretical conversion efficiencies. Historical daily weather data for 1971-1990 were used to represent the baseline climate, while daily weather data generated from three global circulation models for 1971-1990 and 2046-2065 were used to project future changes in climate. Simulations show that the HF cultivar produced higher (15-35%) biomass and ethanol yields than the HS cultivar, but also used slightly more (similar to 4%) water. Climate change is projected to increase dryland yields for both cultivar types (8-19%) Irrigated yields will not change much in current high potential areas (1-5%), given adequate water supply, while yields could increase substantially in current cool areas (similar to 20%). Water and irrigation requirements are expected to increase (9-15%) under a future climate. New areas could be become suitable for irrigated and dryland production. The information produced in this study can be used to assist decision-making for: (1) optimizing production and processing processes and (2) the development of sustainable greenfield projects in marginal areas of South Africa.

The authors regret that some of the data (the ranking columns ‘N41’, ‘R570’, ‘CP88-1762’ and ‘G rank cv%’) presented in Table 3 are unfortunately incorrect. The corrected data are shown below. In consequence, in Section 4.2.1 it was incorrectly noted that “Strong G variation combined with consistent G rankings (high GRcv) suggest significant G effects, with minimal GxE interaction effects, for P crops”. This should read “Strong G variation is evident, with unresolved GxE interaction effects, for P crops”. The conclusion in Section 5.1.1, “Germination is strongly E- and G-controlled with little GxE interaction; …” is also consequently incorrect, and should rather read “Germination is strongly E- and G-controlled; …”. Authors M.R. Jones and A. Singels would like to apologise for any inconvenience caused.

Crop modelling has the potential to assist plant breeding by identifying favourable genotypic (G) traits for specific environments (Es). Sugarcane crop models have not been rigorously evaluated against a factorial GxE dataset. It is imperative that models are evaluated in this way before they are applied to plant breeding problems. Our objectives were to (1) calibrate, (2) assess, and (3) identify weaknesses and recommend improvements to, three sugarcane models, DSSAT-Canegro, Mosicas and APSIM-Sugar, in relation to their predictions of observed E, G and GxE interaction effects in response to abiotic factors (temperature and solar radiation). Data from an international GxE growth analysis trial were used; these consisted of five irrigated experiments at four sites (Belle Glade, Florida, USA; Chiredzi, Zimbabwe; La Mare, Reunion Island; and Pongola, South Africa), with cultivars N41, R570 and CP88-1762. Observed G and E effects on final above-ground dry mass (ADM) yields were explained in terms of seasonal radiation interception (FIPARa) and seasonal average radiation use efficiency (RUEa). Calibration was undertaken where possible by translating phenotypic parameters derived from observations into model input trait parameter values representing genetic traits. E and G effects on FIPARa were generally simulated satisfactorily, while GxE interaction effects were poorly predicted due to inadequate responses to temperature. E, G and GxE effects on RUEa were poorly predicted by all models, although data shortcomings (arising from uncertainty regarding date of primary shoot emergence and impacts of lodging) prevented us from making strong conclusions in this regard. Models accurately predicted G differences in RUEa during mid-season biomass sampling periods where data confidence was greater. Although the models were able to predict final ADM yield per G and per E reasonably well, none of the models predicted GxE interaction effects well. All models also under-estimated the variation in RUEa and ADM. Recommendations for experimental protocols for exploring RUEa are made. Our key recommendations for future work to improve models for sugarcane breeding applications are to explore G-specific thermal time base temperatures for germination and canopy development processes, and to improve linkages between carbon availability and canopy development.

Sugarcane is a globally important crop used for producing sugar and for generating renewable energy. Timely and accurate forecasts of sugarcane yield and production are needed to optimize supply chain operations. Crop growth models (CGMs) are frequently used for sugarcane yield forecasting and have been shown to benefit from using remotely sensed data to force (calibrate) biophysical state variables, such as the fraction of absorbed photosynthetically active radiation (fAPAR). Little is known about the robustness of multispectral vegetation indices for modelling fAPAR in sugarcane growing regions were environmental conditions and farming practices are diverse. This study investigated how the relationships between multispectral Landsat-8 satellite imagery and in situ sugarcane fAPAR measurements vary over large heterogeneous areas. Specifically, it examined which spectral bands and indices are most appropriate for modelling fAPAR under particular production environments and assessed the robustness of the models for application in areas where sugarcane is grown under varying agro-climatic conditions. It was found that cropping and environmental conditions were the main drivers of sugarcane fAPAR modelling success. Significantly (40%) lower mean root mean squared errors (RMSEs) values were recorded in Pongola, which is attributed to the relatively homogenous conditions under which sugarcane is being grown in this area. Generally, the Sezela models were much weaker and the normalized difference vegetation index (NDVI) and soil-adjusted vegetation index (SAVI) models performed relatively poorly, with the best performing models being dominated by the SWIR bands and/or indices generated from it. The non-linear models dominated and are thus recommended for operational implementation owing to their relative simplicity and robustness. From these results we conclude that the use of remotely sensed data for estimating fAPAR throughout the growing season is highly beneficial, but that the selection of suitable variable (index) is critical, especially when the sugarcane area being considered is diverse in terms of farming practices, terrain and climate.

Identifying ways and practices to alleviate water scarceness is an important policy issue across sectors, particularly in the agricultural sector in arid countries. The present study examines how water scarcity can be alleviated by decreasing the water footprint of sugarcane production using different soil mulching and irrigation systems in South Africa. The study also quantifies the economic benefits of reducing blue water footprints. The MyCanesim model and water footprint assessment methodologies were employed to estimate blue and green water footprints under the different systems in the Malelane region of South Africa. The findings reveal that blue water consumption for sugarcane grown with a thick mulch cover was substantially lower than for that grown with a light mulch cover. The difference was larger for centre pivot-irrigated sugarcane than for subsurface drip-irrigated sugarcane. The blue and total (blue plus green) water footprint values for crops grown with a thick mulch cover were only marginally lower than for those grown with the light mulch cover. The blue water footprint for subsurface drip-irrigated sugarcane was 8-10 m(3)/t lower than for centre pivot-irrigated sugarcane due to its higher application efficiency. The economic productivity of blue water usage for subsurface drip-irrigated sugarcane was higher than for centre pivot-irrigated sugarcane crops. In addition, the economic water productivity of blue water usage for crops grown with a thick mulch cover was slightly higher (5%) than that of those grown with a light mulch cover under subsurface drip irrigation. The findings support the notion that water-use efficiency in sugarcane production can be improved and the water footprints reduced by implementing more efficient irrigation systems, by covering the soil with a thick mulch cover to limit evaporation, and by implementing effective irrigation scheduling.

A large portion of global sugarcane is produced under irrigation, and this often occurs in areas where water supply is not abundant or reliable. Crop management decisions during limited water supply are complex and require information on the impacts of irrigation strategies on crops and profitability. This paper describes the development of a computerized system to support farm level management of limited irrigation water for sugarcane production. The system comprises a daily crop and water balance model, an irrigation module and a gross margin calculator. The model calculates crop yield and survival for the current (Y1) and the next season (Y2), for multiple fields on a farm, for a given irrigation strategy and water supply/climate scenario. Irrigation strategies that can be explored include: (1) scheduling irrigation using growth phase specific soil water thresholds (SWT), and (2) postponing replanting and/or abandoning low potential fields. Farm gross margin is calculated from simulated yields and production costs at field level and takes into account re-establishment costs when crops fail. The system was applied in a case study for a hypothetical farm of 18 fields near Komatipoort, South Africa.. Four possible restricted water allocation scenarios were investigated, namely, a mildly and severely restricted allocation ("50% and 25% of the full allocation) over a 24 month and 12 month period. Results from the case study show that under most circumstances a SWT of 60% of plant available soil water capacity applied during the germination and stalk growth phases produced the best outcome. Reducing SWT to 30% during the tillering phase makes more water available for use on other fields, resulting in higher crop survival under severe restrictions. Abandoning low potential fields under severe water restriction limited financial loss in Y2, but reduced future productive capacity thereafter. Results suggest that the system produces realistic responses to irrigation applied and drought. It has the potential to aid strategic decision-making for irrigated sugarcane production during drought.

Crop improvement aims to produce high yielding genotypes for target environments. Crop models simulate yield formation as the outcome of a series of low-level processes, driven by environmental (E) variables and regulated by genetic (G) traits. There is potential for crop models to aid sugarcane breeding, by identifying desirable genetic traits for target environments. The objective of this study was to evaluate existing concepts of G and E control of plant processes for explaining crop development, growth and yield, using an international growth analysis dataset. Crop development, growth and yield were monitored in the plant and 1st ratoon crops for seven cultivars (N41, R570, CP88-1762, HoCP96-540, Q183, ZN7 and NCo376) grown under well-watered conditions at La Mare (Reunion Island, France), Pongola (South Africa (RSA), Chiredzi (Zimbabwe), and Belle Glade (Florida, USA). Weather data were collected and environmental conditions characterized for each experiment. Derived process-level phenotypic parameters, based on concepts from four sugarcane growth simulation models (DSSAT-Canegro, Mosicas, APSIM-Sugar and Canesim), were calculated from observations and used to (1) evaluate current understanding of E drivers of sugarcane growth and development processes, and (2) identify and quantify G control at a process level. Final yields showed significant E and GxE variation; dry above-ground biomass and stalk yields were highest in La Mare and lowest in Pongola. Cultivar rankings in stalk dry mass for the common cultivars (N41, R570, CP88-1762) varied significantly between Es. Significant E variation in phenotypic parameters describing germination, tillering and timing of the onset of stalk growth (OSG) revealed shortcomings in the underlying simulation concepts. Significant G variation was found for germination rate, leaf appearance rate and canopy development rate per unit thermal time (TT), and maximum radiation use efficiency, indicating strong G control of the associated underlying processes. Solar radiation was found to influence tillering rate per unit TT, and TT to OSG, challenging the current theory of TT as the sole driver of these processes. By explaining more of the E variation, more stable and accurate G-specific model parameters can be defined and evaluated. This is anticipated to lead to less GxE confounding of modelled processes, and hence crop models that are better-equipped for supporting sugarcane crop improvement.

The objectives of this paper are to characterise South African sugarcane production for the 2017/18 milling season from an agricultural perspective. This is done to provide insight into successes and failures of recent production strategies, and identify priorities for improved efficiency in producing high quality sugarcane in South Africa. The lengthy and severe drought of 2014/15/16 was finally broken in 2017 with good rainfall bringing relief in rainfed production areas. Yields in these areas improved markedly from 2016. Cane quality also improved although it remained below the long term mean because of disruptive rainfall in May and in some parts in October. Irrigation water supplies remained constrained for a large part of the growing season but started improving from January 2017. This brought about improved yields, while excellent cane quality was achieved in these areas. Smut levels increased in the northern parts of the industry, but levels of other diseases were relatively low. Eldana levels also declined after good summer rainfall in coastal areas. A new pest, the longhorn beetle, posed a serious threat, but seems to have for the present been contained. Vigilance, especially in hotspot areas, remains a high priority. Substantial sugar imports combined with increased local production necessitated exports of locally produced sugar at low world prices, leading to low producers price and profitability, and threatening long term sustainability. Effective tariff protection is urgently required to turn this around. Overall, the 2017 production season will be remembered for a remarkable turnaround in sugarcane production from one of the most severe droughts ever experienced. Agronomic recovery, however, did not translate into financial recovery, due to a low product price. Continued efforts are needed to improve efficiencies along the value chain, for the industry to remain competitive.

Crop models have the potential to support plant breeding by predicting genotype response to environmental factors, and identifying desirable genetic traits for improved crop performance. The study tested whether the Canegro sugarcane model can predict genotypic differences in stalk dry mass (SDM) yields observed in field trials using independently derived genetic trait information. Other objectives included the estimation of three trait parameters (TP) for selected genotypes, and assessing their role in determining genotypic differences in SDM yields. Phenotyping was conducted in a well-watered pot trial at Mount Edgecombe, South Africa comprising 14 genotypes. Gross photosynthate produced per unit of intercepted photosynthetically active radiation under ideal conditions (PARCEo) was estimated from leaf level photosynthetic efficiency (A) and stomatal conductance (g(s)). Thermal time from shoot emergence to the start of stalk elongation (CHUPIBASE) was estimated from measurements of leaf number. Maximum fraction of aerial dry biomass growth partitioned to stalks (STKPFMAX) was estimated from the measured stalk fraction of aerial biomass at harvest. Values of PARCEo (A) and PARCEo (g(s)) differed significantly between genotypes with a range of 47% and 67% of the mean, respectively. CHUPIBASE values also differed significantly between genotypes and showed a range of 23% of the mean. STKPFMAX values did not differ significantly between genotypes and showed the least variation with a range of 17% of the mean. The Canegro model predicted SDM yields and rankings well (r = 0.90**) for nine genotypes grown in well-watered field trials at Pongola, South Africa, using these independent estimates of PARCEo (A), CHUPIBASE and STKPFMAX values. The overestimation of the observed genotypic range in SDM yields were corrected by dynamically scaling leaf level photosynthetic efficiency using fractional sunlit leaf area. The reliable prediction of genotype performance was mostly ascribed to the impact of PARCEo. The extent of genetic variation in PARCEo found in the relatively small number of genotypes for well-watered crops, suggest that sugarcane improvement could be enhanced by screening breeding populations for high values of this trait. The study provided proof of concept that realistic sugarcane models could be used for identifying key traits (in this case PARCEo) and their ideal values (in this case as high as possible), and therefore could be used to assist in defining sugarcane breeding targets.

Crop models can be used for predicting climate change impacts and exploring adaptation strategies, but their suitability for such tasks needs to be assessed. Although the DSSAT-Canegro model has been used widely for climate impact studies, some shortcomings have been revealed. The objectives were to improve and evaluate the capability of DSSAT-Canegro to predict crop responses to climate change. Model changes included improved simulation of elevated temperature and atmospheric CO2 concentration ([CO2]) impacts, and revised algorithms for tillering, respiration and crop water relations. After calibration, the refined model was tested against an independent set of experimental data, demonstrating acceptable simulation accuracy for aerial dry mass, stalk dry mass and stalk sucrose mass (RMSE = 8.4, 5.2 and 3.3 t/ha respectively). A multiple-site sensitivity analysis revealed that simulated responses by the refined model, of canopy formation, crop water use, crop water status and stalk dry mass to changes in rainfall, temperature and [CO2], were more realistic than those of the old model. Highest average simulated stalk mass was achieved at a temperature regime that was 3 degrees C warmer than current climate, with yield increases ranging from 0.7% (irrigated Ligne Paradis, Reunion Island) to 7% (rainfed Piracicaba, Brazil). Elevated [CO2] increased yields for rainfed production only (7% for La Mercy, South Africa and 6% for Piracicaba, [CO2] = 750 ppm), through reduced transpiration and improved crop water status. The study highlighted the need for improvements in simulating reduced growth of older crops, and [CO2] effects on transpiration. This study has delivered an improved Canegro model that represents plant processes and their interactions with climatic drivers more realistically, and can predict crop growth, water use and yields, for a wide range of climates, reasonably accurately. We propose that this revised Canegro model is included in a forthcoming release of the DSSAT Cropping System Model, for use in climate change impact studies.

The objective of this study was to assess the accuracy, spatial variation and potential value of remote sensing (RS) estimates of evapotranspiration (ET) and biomass production for irrigated sugarcane in Mpumalanga, South Africa. Weekly ET and biomass production were estimated from RS data from 2011 to 2013 using the Surface Energy Balance Algorithm for Land (SEBAL). Ground estimates of canopy interception of photosynthetically active radiation (FPAR) and aerial biomass were compared to RS estimates. ET was estimated with a surface renewal (SR) system in one field. Evaporation coefficient (Kc) values were calculated from ET and reference grass evaporation. Remote sensing FPAR and biomass estimates compared well with field measurements (R2 = 0.89 and 0.78). SEBAL ET estimates exceeded SR estimates by 5 mm/week, while full canopy Kc values for SEBAL compared better with literature values than with SR Kc values. SEBAL estimates of ET and biomass were regarded as reliable. Considerable spatial variation was observed in seasonal RS ET (1 034 ± 223 mm), biomass (45 ± 17 t/ha) and biomass water use efficiency (WUEBIO, defined as dry biomass produced per unit of ET) (4.1 ± 1.0 kg/m3). About 32% of sugarcane fields had values below economic thresholds, indicating an opportunity to increase productivity. Actual yields correlated well with WUEBIO values, suggesting that this may be used for monitoring crop performance and identifying areas that require remedial treatment.

Presentation title Presenter Recent advances in genetic trait modelling in DSSAT Gerrit Hoogenboom New features of the APSIM Sugar model for simulating traits for yield improvement of sugarcane in water limited environments Geoff Inman-Bamber et al. Combining calibration techniques improves the quality and usefulness of sugarcane model predictions Fabio Marin Sugarcane trait modelling at SASRI Abraham Singels Trait parameter estimation and ideotyping with Canegro Natalie Hoffman Selecting sugarcane with higher transpiration efficiency Phil Jackson The ICSM genotype growth analysis dataset Abraham Singels et al. Other genotype growth analysis datasets Sanesh Ramburan Simulating genotype performance with Canegro Matthew Jones Simulating genotype performance with Mosicas Mathias Christina et al. Simulating genotype performance with APSIM-sugar Fabio Marin Process level comparison of approaches to simulating environmental and genetic effects M. Jones et al.

There are indications that high-fibre sugarcane genotypes may produce more biomass and use resources more efficiently than conventional sugarcane cultivars. The objective of this research was to gather quantitative information on resource use for selected conventional and high-fibre sugarcane genotypes and benchmark it against other bioethanol crops. Although conventional sugarcane initially grew slower than sorghum and Napier grass, it produced very high biomass (about 70 t ha−1) and theoretical ethanol (first- and second- generations) yields (about 27 kL ha−1) at 12 months, and used water relatively efficiently (about 5 kg m−3 and 2 kL m−3), out-performing all other crops except sorghum. The contribution of cellulosic ethanol to total ethanol yield varied hugely, from 89% for the high-fibre sugarcane hybrid to about 48% for conventional sugarcane, to as low as 14% for sugar beet. The high-fibre sugarcane hybrid grew faster initially and produced more biomass at eight months (56 t ha−1 vs 45 t ha−1) than the conventional types, but then flowered, reducing its growth rates markedly thereafter. It was also less sensitive to mild drought conditions. The results suggest that cellulosic ethanol production may be a feasible option that could be incorporated into conventional or biomass sugarcane production systems.

Reliable predictions of climate change impacts on water use, irrigation requirements and yields of irrigated sugarcane in South Africa (a water-scarce country) are necessary to plan adaptation strategies. Although previous work has been done in this regard, methodologies and results vary considerably. The objectives were (1) to estimate likely impacts of climate change on sugarcane yields, water use and irrigation demand at three irrigated sugarcane production sites in South Africa (Malelane, Pongola and La Mercy) for current (1980-2010) and future (2070-2100) climate scenarios, using an approach based on the Agricultural Model Intercomparison and Improvement Project (AgMIP) protocols; and (2) to assess the suitability of this methodology for investigating climate change impacts on sugarcane production.Future climate datasets were generated using the Delta downscaling method and three Global Circulation Models (GCMs) assuming atmospheric CO2 concentration [CO2] of 734 ppm (A2 emissions scenario). Yield and water use were simulated using the DSSAT-Canegro v4.5 model.Irrigated cane yields are expected to increase at all three sites (between 11 and 14%), primarily due to increased interception of radiation as a result of accelerated canopy development. Evapotranspiration and irrigation requirements increased by 11% due to increased canopy cover and evaporative demand. Sucrose yields are expected to decline because of increased consumption of photo-assimilate for structural growth and maintenance respiration. Crop responses in canopy development and yield formation differed markedly between the crop cycles investigated.Possible agronomic implications of these results include reduced weed control costs due to shortened periods of partial canopy, a need for improved efficiency of irrigation to counter increased demands, and adjustments to ripening and harvest practices to counter decreased cane quality and optimise productivity. Although the Delta climate data downscaling method is considered robust, accurate and easily-understood, it does not change the future number of rain-days per month. The impacts of this and other climate data simplifications ought to be explored in future work. Shortcomings of the DSSAT-Canegro model include the simulated responses of phenological development, photosynthesis and respiration processes to high temperatures, and the disconnect between simulated biomass accumulation and expansive growth. Proposed methodology refinements should improve the reliability of predicted climate change impacts on sugarcane yield. (C) 2015 Elsevier Ltd. All rights reserved.

Lodging lowers the productivity of sugarcane through a reduction in radiation use efficiency and stalk damage. However, there are few reports of experiments specifically designed to quantify effects of lodging in sugarcane. Efforts to model onset and progression of lodging, and the impact on crop productivity, have not been attempted. The objectives of this paper were to quantify effects of lodging on sugarcane and to develop modeling capability in terms of predicting lodging onset, progression and impact. Field experiments with irrigated ratoon crops were conducted at Pongola, South Africa. In one treatment the cane in each plot was allowed to grow through bamboo frames that prevented lodging. In the other treatment, the cane was not supported and could lodge at any stage. The degree of lodging was captured weekly by a rating that ranged from 1 to 9, where 1 = fully erect cane and 9 = completely lodged cane. At harvest estimated recoverable crystal percent (ERC %) of stalks and yield (cane and ERC) was measured for each plot. Lodging resulted in decreased ERC yields of up to 20.6%. An algorithm for simulating lodging when aboveground biomass (including rainfall and irrigation water retained on it) exceeds a variety-specific threshold, and which also considers wind speed and soil water content, was evaluated for predicting the extent and impact of lodging in the Pongola experiments, as well as for four deficit irrigation treatments of a field experiment conducted in Komatipoort, South Africa. The study showed that the onset of lodging was simulated reasonably well for various soil/crop/atmospheric conditions, while the extent of lodging at harvest was simulated very accurately for all crops. Simulated lodging was primarily driven by crop size and lodging events were triggered by rainfall that added weight to the aerial mass of the crop, and reduced the anchoring ability of the soil through saturation of the top soil. More accurate simulation of lodging, and its impacts on yield, will improve the accuracy of yield predictions by crop models, increasing their value in applications such as crop forecasting, climate change studies and exploring crop improvement and management options. (C) 2015 Elsevier B.V. All rights reserved.

A perception exists in the South African (SA) sugar industry that sugarcane yields are declining. The objective of this study was to quantify yield decline in the SA sugar industry to inform future research and extension efforts to decrease yield gaps. Regional trends in cane yield were calculated from mill-level cane delivery and harvest area data. Benchmark simulated yields for each region were estimated by the Canesim Crop Forecasting System, using inferred harvest age and observed weather data. Actual yields were annualised to remove harvest age effects and then expressed as fractions of corresponding simulated yields, in order to remove effects of inter-seasonal variations in weather. Trends in this yield ratio (YR) were calculated for several regions and grower categories. Yield decline was defined as a decreasing trend in YR over time (1981-2010).Applying this methodology to historic production data for the South African sugar industry revealed that large scale grower (LSG) yield ratios have declined significantly over the period 1986 to 2010, for the South Coast (1.12%/year) and North Coast (1.24%/year). LSG yield ratios increased in Zululand and the Northern irrigated regions, suggesting that growers were able to cope with deteriorating weather and/or exploit technology improvements. Small-scale grower (SSG) yield ratios declined from 1993 to 2010 by 1.73%/year in the Northern Irrigated region. Since 2001 industry average LSG yield ratios declined by 1.97%/year compared to a decline of 237%/year for SSG yields (both significant). On a regional basis SSG yield ratios declined in the Midlands (1.43%/year) and Zululand (3.33%/year) regions.Yield decline was perceived by stakeholders to be caused mainly by soil degradation, decreased investment as a result of unresolved land claims, and increasing pest and weed pressures, and suggested increased top/sub-soil liming, green manuring, in-field traffic control and soil and leaf testing as possible means of addressing yield decline. Closing the gap between current and maximum economically-attainable yields has the potential to increase annual industry production by approximately 32% (63 million t), and annual grower revenue by ZAR 2.6 billion (based on 2006-2010 industry production). It is recommended that research and extension efforts should be focussed on further understanding and addressing yield decline, particularly for LSGs and SSGs in the Midlands region, LSGs in coastal regions, and SSGs in the Northern Irrigated and Zululand regions. (C) 2015 Elsevier Ltd. All rights reserved.