Pringle, M. & Department of the Environment, T. (2022). Winter and Summer Crop Mapping – Landsat, Sentinel-2 and MODIS, Queensland, 1990 - Ongoing. Version 0.5. Terrestrial Ecosystem Research Network. Dataset. https://portal.tern.org.au/metadata/bae77dd5-ba0a-41b4-973b-a800236b8476
Winter and Summer Crop Mapping – Landsat, Sentinel-2 and MODIS, Queensland, 1990 - Ongoing
Ver: 0.5
Status of Data: onGoing
Update Frequency: quarterly
Security Classification: unclassified
Record Last Modified: 2026-04-12
Viewed 895 times
Accessed 199 times
Dataset Created: 2024-12-16
Dataset Published: 2022-07-22
Data can be accessed from the following links:
How to cite this collection:
This dataset shows the broad groups of crops grown in the main cropping areas of Queensland, for the winter and summer growing seasons from 1990 to the current year. The winter growing-season is defined as June to October, and the summer growing-season is November to May. The predicted group is stored in the attribute table (field 'CLASS'), along with the probability of the prediction (field 'P_CLASS', the larger this value, the more certain is 'CLASS'). Each season has 2 maps: an end-of-season prediction and a mid-season prediction. The mid-season prediction is labelled "_vInterim" to indicate that it is based on a relatively short time series and should be used with caution.
Credit
We at TERN acknowledge the Traditional Owners and Custodians throughout Australia, New Zealand and all nations. We honour their profound connections to land, water, biodiversity and culture and pay our respects to their Elders past, present and emerging. Landsat imagery was obtained from the US Geological Survey. Modified-Copernicus-Sentinel-2 imagery was obtained from the European Space Agency. MODIS MOD13Q1 imagery was obtained from the LP DAAC Data Pool.
Purpose
The classification algorithm predicts these classes of crops in summer: “Banana”, "Cotton", "Sugarcane", and "OtherCrop" (predominantly sorghum, but also includes, e.g., maize, mungbean, peanut). In winter, the classification algorithm predicts “Cereal” (including Wheat and Barley) and “Chickpea”. Note that the extent of the mapping changes by season: in winter the maps are restricted to what we define as the 'western' cropping zone only; in summer, predictions extend further, into the potential sugarcane-growing areas of the 'coastal' zone (which includes northern NSW). Any other crops grown in the coastal zone, apart from sugarcane, are not considered.
Lineage
All the data described here has been generated from the analysis of satellite imagery at a spatial resolution of approximately 30 m. A grid of Landsat TM, ETM+ and OLI data were supplemented by Sentinel-2 (after 2016) and MODIS (after 2000) imagery when large temporal data gaps occurred.
An algorithm then interpolates pixel-wise data to weekly averages and determines the best match to one of the seasonal classes. The algorithm was trained with >10,000 field observations and validated against >4,000 independent observations. The predicted group is stored in the attribute table (field 'CLASS'), along with the probability of the prediction (field 'P_CLASS'; the larger this value, the more certain is 'CLASS').
These datasets are GDA2020 compliant.
Method DocumentationData not provided.
Procedure Steps
1.
Attributes:
The predicted class is stored in the attribute table (field 'CLASS'), along with the probability of the prediction (field 'P_CLASS'; the larger this value, the more certain is 'CLASS').
The predicted class is stored in the attribute table (field 'CLASS'), along with the probability of the prediction (field 'P_CLASS'; the larger this value, the more certain is 'CLASS').
Spatial Description
Note that a larger extent is covered by the summer crop mapping compared with the winter.
Temporal Coverage
From 1990-01-01 to on going
Spatial Resolution
Vertical Extent
Data not provided.
Data Quality Assessment Scope
Certain features, associated with a negligible chance of cropping, have been masked from the maps, e.g. 'Conservation and Natural Environments', 'Intensive Uses', 'Water' or any type of forestry or horticulture, as defined by the Australia Land Use and Management Classification.
In winter, ‘Pulse' is often misclassified as ‘Cereal’. In summer, the “Other crop” class - which groups coarse-grain and pulse crops such as sorghum, maize, and mungbean - is the least accurately mapped, particularly in the western area of the QLD–NSW border. Pastures may be incorrectly classified as cropping, particularly in wet years or in seasons when rapid green-up occurs at a similar time to actively growing crops. Formerly-cropped land that is not currently under cultivation may still appear as cropped if the vegetation greens-up during the growing period, e.g. weeds or redundant or abandoned crops and crop residue are dominating the ground cover. A failed crop can only be detected if it reached a certain level of greenness, which may vary between region, seasons and years. Vegetated areas around watercourses and dams may be classified as crop due to the strong greening-up phase after rainfall events. Some areas of shallow water or water with emergent or floating vegetation may be incorrectly classified as cropped, due to seasonal patterns in greenness that may be similar to crops. Areas with steep slopes (i.e. a slope of greater than 10%) are excluded. Inaccuracies in the DEM used for identifying these areas may result in some areas being included. From 2025 onward, manual editing of the crop map was carried out through image interpretation to correct the most critical errors, using Sentinel-2 and Planet imagery, as well as Google Earth, as ancillary data.
Data Quality Report
Data not provided.
Data Quality Assessment Outcome
In summer, notional user's accuracies are: “Banana” = 99%, "Cotton" = 87%, "Sugarcane" = 98%, and "OtherCrop" = 98%. In winter the user's accuracies are: "Cereal" = 89%, and “Chickpea” = 73%. User’s accuracies cannot be reported as anything more than notional approximations, due to sampling constraints.
GCMD Sciences
Horizontal Resolution
Parameters
Temporal Resolution
Supplemental Information
The paper published in Remote Sensing of Environment [Pringle, M.J., Schmidt, M., and Tindall, D.R. (2018): Multi-decade, multi-sensor time-series modelling---based on geostatistical concepts---to predict broad groups of crops. Remote Sensing of Environment, 216, 183--200] describes the initial mapping approach used. Major modifications to this approach have since taken place, as outlined below.
* Growing seasons are now fixed and extended to 52 weeks for the time series analysis, whereas in the previous paper they varied in length. The 2018 RSE paper recommended including early-season values and related phenological metrics to improve the mapping of “planting areas.” This is done now by fixing 52-week periods from the cut-off dates in the time series model * The current mapping workflow uses a different interpolation method to estimate phenology predictions from the time series model. These interpolated values are then used as inputs to the predictive model, resulting in more accurate biannual crop mapping overall. * Classifications are now performed at the Landsat pixel scale, whereas previously they were based on field boundaries defined through Object Based Image Analyses (OBIA) image segmentation (see explanation in the 2018 RSE paper). The segmentation approach introduced an undesirable smoothing effect in the crop maps hence the Landsat pixel scale. * Models are now developed using both Long-Short-Term-Memory (LSTM) neural networks and random forests (RF), whereas previously only RF was used.
* Growing seasons are now fixed and extended to 52 weeks for the time series analysis, whereas in the previous paper they varied in length. The 2018 RSE paper recommended including early-season values and related phenological metrics to improve the mapping of “planting areas.” This is done now by fixing 52-week periods from the cut-off dates in the time series model * The current mapping workflow uses a different interpolation method to estimate phenology predictions from the time series model. These interpolated values are then used as inputs to the predictive model, resulting in more accurate biannual crop mapping overall. * Classifications are now performed at the Landsat pixel scale, whereas previously they were based on field boundaries defined through Object Based Image Analyses (OBIA) image segmentation (see explanation in the 2018 RSE paper). The segmentation approach introduced an undesirable smoothing effect in the crop maps hence the Landsat pixel scale. * Models are now developed using both Long-Short-Term-Memory (LSTM) neural networks and random forests (RF), whereas previously only RF was used.
Resource Specific Usage
Filenames follow a simple convention: cropmap_.gpkg
Example: cropmap_winter2020.gpkg
Example: cropmap_winter2020.gpkg
Environment Description
Data not provided.
Terrestrial Ecosystem Research Network
80 Meiers Road, Indooroopilly, Queensland, 4068, Australia.
Contact Us
TERN services are provided on an "as-is" and "as available" basis. Users use any TERN services at their discretion and risk. They will be solely responsible for any damage or loss whatsoever that results from such use including use of any data obtained through TERN and any analysis performed using the TERN infrastructure.
Web links to and from external, third party websites should not be construed as implying any relationships with and/or endorsement of the external site or its content by TERN.
Please advise any work or publications that use this data via the online form at https://www.tern.org.au/research-publications/#reporting
Web links to and from external, third party websites should not be construed as implying any relationships with and/or endorsement of the external site or its content by TERN.
Please advise any work or publications that use this data via the online form at https://www.tern.org.au/research-publications/#reporting
Please cite this dataset as {Author} ({PublicationYear}). {Title}. {Version, as appropriate}. Terrestrial Ecosystem Research Network. Dataset. {Identifier}.
The Department of Environment, Tourism, Science and Innovation requests attribution in the following manner. State of Queensland (Department of Environment, Tourism, Science and Innovation) 2025.Update data available at http://qldspatial.information.qld.gov.au/catalogue/