Mapping cotton fields using phenology-based metrics derived from a time series of Landsat imagery

Dhahi Al-Shammari1, Thomas F.A. Bishop1, Ignacio Fuentes1, Patrick Filippi1

1 Sydney Institute of Agriculture, School of Life & Environmental Science, The University of Sydney, Central Ave, Eveleigh, Sydney, New South Wales, 2015


A phenology-based crop type classification was carried out to map cotton in New South Wales and Queensland, Australia. The workflow was implemented in Google Earth Engine (GEE) platform as it is time efficient and does not require processing in multiple platforms to complete the classification steps. A time-series of images were generated from Landsat 8 Surface Reflectance (L8SR) data. The Normalised Difference Vegetation Index (NDVI)time series was calculated from satellite specifically Landsat imagery and a Harmonic Model (HM) was fitted to it to produce the Harmonised NDVI (H-NDVI). Phase and amplitude images were generated to visualise active cotton in the targeted fields. These images were used as predictor variables with H-NDVI and other raw bands in the Random Forest (RF) classification model. The results of RF proved that both phase and amplitude increased the accuracy of the classification. Moreover, cotton classification accuracy increased as the season progressed.  

Canopeo, a new mobile device application with potential to measure seed colour change in canola

Rick Graham1*, Rohan Brill2, Rodney Bambach1, Jan Hosking1, Neroli Graham1

1NSW DPI, Tamworth Agricultural Institute, 4 Marsden Park Road, Tamworth 2340, Australia,

2NSW DPI, Wagga Wagga Agricultural Institute, Pine Gully Road, Wagga Wagga 2650, Australia

*Presenting author:


Seed colour change (SCC) is commonly used to determine windrow timing in canola, it is however often regarded as being subjective and time consuming. The aim of this research was to compare SCC measurements using a standardised visual assessment procedure with Canopeo, a new mobile device application commonly used to measure fractional green canopy cover. Windrow timing experiments were conducted at Tamworth in northern NSW, in 2017 and 2018, with SCC measurements undertaken at 2-3 day intervals from the commencement of SCC on the primary stem up until 100% SCC on branches. Pictures of seed at each windrow timing were analysed by Canopeo and then compared to a standardised visual assessment measurement. There was a linear relationship (R2 = 0.93 to 0.95) between SCC measured using the mobile device application Canopeo and the detailed visual assessment method. Canopeo was also between 10 and 20 times faster than the visual assessment procedure. These results indicate that Canopeo can be used to take pictures to measure SCC and may be a viable alternative for determining SCC and hence windrow timing in canola crops. Importantly, Canopeo offers the potential for a rapid standardised objective measurement for determining SCC, thereby alleviating issues around subjectivity commonly associated with visual assessment methods and as such, is considered a useful new technique for the determination of SCC in canola.

What makes a ‘good’ seasonal forecast? Delivering actionable climate outlooks for grains farming

Patrick J. Mitchell1, Jaclyn N. Brown1

1 CSIRO Agriculture and Food, College Rd, Sandy Bay, Tasmania, 7005,


The prediction of climate patterns and weather conditions at the farm scale represents an important innovation for managing within season and year-to-year variability in crop production. Assessing skill and potential value of long-range, seasonal climate forecasts hinges on answering the fundamental questions: “Should I use this forecast when making my decision and how ‘good’ is it?” Here, we use model output from the new seasonal forecasting system, ACCESS-S1 to compare forecast approaches for deriving relevant and credible seasonal climate information for Australia’s cropping regions. This evaluation addresses the role of two important components: categorisation of the model output and anchoring the forecast using antecedent conditions (fallow season rainfall). Overall, the model had relatively low accuracy at predicting correct forecasts across much of the forecast locations and seasons, whereas it had greater skill in the avoidance of false alarms i.e. false negative outcomes. The percentile categories used to derive the expected forecast had a large effect on the skill in terms of the rate of false alarms and the choice of categories can be matched to user requirements of both accuracy and resolution. Anchoring rainfall forecasts on antecedent conditions can reduce false alarms across the growing season and may be a useful guide when presented alongside a forecast based solely on in-season predicted rainfall. The next generation of climate data products and services for agriculture need to consider how a forecast system interacts with both on-farm biophysical drivers of yield and decision-making preferences of the user.

Mapping rare and infrequent crops from space

François Waldner1, Yang Chen2, Roger Lawes2, Zvi Hochman1

1 CSIRO Agriculture and Food, 306 Carmody Rd, St Lucia QLD 4067, Australia;

2 CSIRO Agriculture and Food, Underwood Ave, Floreat WA 6014, Australia;


Most cropping systems around the world are organised around few dominant crops and a larger number of less frequent crops. Data about the location of infrequent crops derived from satellite data are generally inaccurate, largely owing to the class imbalance problem. Class imbalance occurs when only few instances of some classes are available for classifier training and leads to large classification errors of the infrequent classes. Here, we assessed the magnitude of the class imbalance problem in crop classification and evaluated data-level treatments to combat it by creating synthetic minority instances. 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. 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. We found that data-level treatments boosted GM by 0.1-0.35 and that the price for increasing the accuracy of minority classes is a drop in OA. While oversampling methods have clear potential to improve the classification of minority crop types, more control over the loss of overall accuracy needs to be gained before transitioning these methods to operations.

‘Digital agriculture’ helping farmers reduce impacts of cropping on the Great Barrier Reef

Peter J. Thorburn1, Peter Fitch2, YiFan Zhang1, Yuri Shendryk1, Tony Webster3, Jody Biggs1, Martijn Mooij4, Catherine Ticehurst2, Maria Vilas1 and Simon Fielke5

1CSIRO Agriculture and Food, Brisbane, 306 Carmody Rd, St Lucia, Qld, 4067,,
2CSIRO Land and Water, GPO Box 1700, Canberra, ACT, 2601,
3CSIRO Agriculture and Food, PO Box 12139, Earlville BC, Qld, 4870,
4CSIRO Data61, PO Box 10522, Brisbane Adelaide St, Qld, 4006,
5CSIRO Land and Water, 41 Boggo Rd, Dutton Park, Qld, 4102


Nitrogen (N) losses from sugarcane production need to be reduced to help protect the health of the Great Barrier Reef. This challenge comes at a time when digital technologies are becoming more accessible and thus can be harnessed to improve N fertiliser management. We are developing ‘apps’ and advanced analytics to provide farmers with high quality information on: (1) water quality in their local creeks and rivers; (2) the magnitude of risk to production posed by lower N fertiliser rates; and (3) the abatement of N loss associated with those lower N rates, to help farmers potentially access payments from environmental schemes. We are also developing new ways of remotely sensing sugarcane crops so farmers can evaluate better the impacts of changed management on crop performance. This information will facilitate improved agronomic management leading to reduced impacts on the Great Barrier Reef.

Predicting summer rainfall in coastal NE Australia for improved farming practices in sugar cane

Kavina Dayal1, Jaclyn Brown1, Peter Thorburn2

1CSIRO, 15 College Road, Sandy Bay, TAS, 7005,,,
2CSIRO, 306 Carmody Road, St Lucia, QLD, 4067.


Reliable summer rainfall forecasts in coastal northeast Australia can empower sugar cane growers to better manage their farm, including tailoring nitrogen applications for the coming climate, which will reduce damage to the Great Barrier Reef. We evaluated the ability of a seasonal climate forecasting model (ACCESS-S) to predict summer rainfall for the wet tropics region of northeastern Australia. We found that ACCESS-S was unable to capture the rainfall variability, primarily due to not being able to capture the extreme events.

Predicting soil water holding capacity from climate and crop yield

Enli Wang 1, Di He 1, and Zhongkui Luo1

1 CSIRO Agriculture and Food, GPO Box 1700, Canberra ACT 2601, ACT, Australia,


Soil water holding capacity is a key soil property affecting dryland crop yield, and is therefore important for crop management in semi-arid climates like Australia. This paper explores two approaches: one developed using process-based modelling to inversely predict plant available water capacity (PAWC) of soils from crop yield, another one built with machine learning to predict soil available water capacity (AWC) spatially based on bio-climate variables. Our results indicate that soil PAWC can be skilfully predicted with water-limited crop yield (R2 of 0.84~0.98 and RMSE of 14.5mm~30.2mm across 10 sites) and that the bio-climate variables together with a machine learning approach could explain up to 50% of the variance in soil AWC across sites. These results demonstrate the potential to use climate and crop yield data to predict soil water holding capacity.

Identifying within-season cotton crop nitrogen status using multispectral imagery

 Jordan Romeo1, Patrick Filippi1, Jon Baird2, Anastasia Volkova3, Daniel K.Y. Tan1

1 The University of Sydney, Sydney Institute of Agriculture, School of Life and Environmental Sciences, Faculty of Science, Sydney NSW 2006,
2 NSW DPI, Australian Cotton Research Institute, Narrabri, NSW 2390,
3 FluroSat, 2/4 Cornwallis St, Eveleigh NSW 2015


Nitrogen (N) plays a key role in the growth and development of a cotton (Gossypium hirsutum) plant and the timing of N fertiliser application has a critical effect on yield. Under-application of N fertilizer reduces yield, while over-application encourages vegetative growth at the expense of reproductive growth, resulting in a higher production cost. In the Australian cotton industry, N fertiliser is typically applied at a uniform rate across a field at a high enough rate to ensure N is not limiting at any area of the field and this practice will not be cost effective with the rising cost of N. This study investigates the use of vegetation indices (VIs) derived from multispectral imagery from planes and satellites (Sentinel 2) to estimate plant nitrogen status to facilitate variable rate fertiliser application and also for predicting lint yield at time of harvest. Sicot 748B3F cotton was sown with six pre-season nitrogen rates 0, 37.5, 75, 112.5, 150 and 187.5 kg N/ha in three replicates in a randomised complete block design layout. Aerial multispectral imagery was collected on three dates across the growing season, as well as from satellites at 17 dates throughout the season. Imagery gathered were processed into VIs including Normalised Difference Vegetation Index (NDVI), Normalised Difference Red Edge Index, Canopy Chlorophyll Content Index (CCCI) and Modified Soil-Adjusted Vegetation Index (MSAVI). Plant tissue samples were collected for nitrogen concentration at 75 and 139 days after sowing, with coinciding aerial imagery. At plant maturity (143 DAS), the coefficient of determination values for N% were r2 = 0.34 for CCCI, r2 = 0.34 for NDVI and r2 = 0.27 for NDRE indicating a relatively low/moderate correlation between VIs and leaf N%. The best time to predict/forecast cotton lint yield using CCCI and NDRE was at 178 DAS, based on satellite imagery. This study suggests aerial and satellite imagery can provide potential for variable rate fertiliser application, and the possibility to replace the current method of N tissue testing.


Development of a pest identification mobile phone application for mungbean in Northwest Cambodia

Isabel C. Hinchcliffe1, Rosanne Quinnell1, Robert Martin1, Van Touch1 and Daniel K.Y. Tan1

1 The University of Sydney, Sydney Institute of Agriculture, School of Life and Environmental Sciences, Faculty of Science, Sydney, NSW 2006, Australia, Email:


In response to the need for crop diversity, Cambodian farmers have begun incorporating mungbean into rice cropping systems. However, these smallholder mungbean producers are facing significant yield loss due to direct impacts of insect and disease pests. Improper pest management has worsened the issue, causing economic losses to farmers and environmental disruption. Improper and over-use of broad-spectrum pesticides as a solution to all observed pests is common place in mungbean fields of lowland Cambodia and these practices are linked to poor sources of agricultural information. This project aimed to discover the pest and beneficial species most common in mungbean fields of lowland Cambodia, and to use this information to develop an informative image-rich mobile phone application to aid Cambodian farmers and field agronomists with insect and disease identification, and so provide specific management recommendations, aligned with the principles of integrated pest management, applicable to the Cambodian context. This study evaluated the feasibility of the proposed app through a survey with potential users. These survey responses were incorporated into the development of the Pest ID app prototype, which was trialled with farmers and subsequently refined by adding audio content in Khmer. The majority of farmers in this study were unable to distinguish between beneficial and pest insect species. The Pest ID app has been well received by farmers with users seeing its potential to support crop management decisions. This app holds potential as an important agricultural education tool for mungbean farmers in the greater Mekong region.

Monitoring Grain Production across the Australian Continent

Roger A. Lawes1, Joanne Chai2, Yang Chen1, Randall J. Donohue3, Gonzalo Mata1, Zvi Hochman4, Franz Waldner4, Chris Sharman5, Roger Butler6

1 CSIRO Agriculture and Food, Underwood Ave, Floreat WA 6014, Australia;,
2 CSIRO Data61, Underwood Ave, Floreat WA 6014, Australia,
3 CSIRO Land and Water, GPO Box 1700, Canberra, ACT 2061, Australia;
4 CSIRO Agriculture and Food, 306 Carmody Rd, St Lucia QLD 4067, Australia;
5 CSIRO Data61, College Rd, Sandy Bay TAS 7005, Australia.
6 CSIRO Data61, 1 Technology Court, Pullenvale, Qld 4069, Australia.


To successfully assess crop yield across Australia there is a need to monitor what has been sown and its progress as the season evolves. Crop type and species need to be identified at the paddock scale to calculate areas. Crop models then need to be applied to each individual paddock to generate a yield estimate.  Finally, the information needs to be packaged at a resolution of interest. To generate state and national scale crop monitoring capability, a co-ordinated, multifaceted data gathering, data training, image capture, data acquisition and crop modelling operations were developed. Crop yield forecasting required new modelling techniques, as existing approaches were overwhelmed by the volume of data.  We describe the detailed process of how we monitor and forecast crop production across the Australian landscape. Near real time crop monitoring products are now available across the Australian continent. This paper describes the overarching workflow of crop monitoring, forecasting and data dissemination to assist agribusiness to respond to the prevailing climate.



The Australian Society of Agronomy is the professional body for agronomists in Australia. It has approximately 500 active members drawn from government, universities, research organisations and the private sector.

Photo Credits

David Marland Photography Graham Centre for Agricultural Innovation, Charles Sturt University

Conference Managers

Please contact the team at Conference Design with any questions regarding the conference.

© 2015 - 2019 Conference Design Pty Ltd