the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Mapping sugarcane globally at 10 m resolution using GEDI and Sentinel-2
Abstract. Sugarcane is an important source of food, biofuel, and farmer income in many countries. At the same time, sugarcane is implicated in many social and environmental challenges, including water scarcity and nutrient pollution. Currently, few of the top sugar-producing countries generate reliable maps of where sugarcane is cultivated. To fill this gap, we introduce a dataset of detailed sugarcane maps for the top 13 producing countries in the world, comprising nearly 90 % of global production. Maps were generated for the 2019–2022 period by combining data from the Global Ecosystem Dynamics Investigation (GEDI) and Sentinel-2 (S2). GEDI data were used to provide training data on where tall and short crops were growing each month, while S2 features were used to map tall crops for all cropland pixels each month. Sugarcane was then identified by leveraging the fact that sugar is typically the only tall crop growing for a substantial fraction of time during the study period. Comparisons with field data, pre-existing maps, and official government statistics all indicated high precision and recall of our maps. Agreement with field data at the pixel level exceeded 80 % in most countries, and sub-national sugarcane areas from our maps were consistent with government statistics. Exceptions appeared mainly due to problems in underlying cropland masks, or to under-reporting of sugarcane area by governments. The final maps should be useful in studying the various impacts of sugarcane cultivation and producing maps of related outcomes such as sugarcane yields.
- Preprint
(10523 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2024-121', Anonymous Referee #1, 16 May 2024
The manuscript proposes a sugarcane mapping dataset that combines the GEDI and Sentinel-2 datasets to map sugarcane for the 13 largest sugarcane-producing countries based on sugarcane phenology and height for the period 2019 to 2022. Overall, the article is well organized and nicely written and falls within the scope of this journal. However, I have some questions about the data production process, mainly as follows:
Major Comments:
1) The criteria used to classify sugarcane from other crops might not distinguish effectively between bamboo and sugarcane, both of which are tall, perennial members of the grass family. Given the similarities in their growth habits and physical characteristics, there is a risk of misclassification in regions where bamboo is prevalent. It would be helpful to include a sensitivity analysis addressing this potential issue. Adding discriminative remote sensing indices or additional ground truth data to differentiate between bamboo and sugarcane could significantly enhance the classification accuracy.
2) The manuscript used a uniform threshold ("Tall month") across different countries for sugarcane classification, which might not account for regional variations in sugarcane phenology influenced by local climatic conditions and sugarcane varieties. In Figure 3 we see that the shape of the curves of the kappa coefficients in response to the threshold varies considerably from country to country. It may be helpful to explore the diversity of sugarcane cultivated species within different countries and region-specific thresholds to improve classification accuracy and to account for uncertainty in some regions.
3) The validation results show a significant discrepancy between the F1 (0.64) and R2 (0.97 with a slope of 1) in Guangxi. A more detailed investigation into these discrepancies is warranted. Otherwise, it is difficult to distinguish whether the match between the sugarcane planting area obtained and the government report is a coincidence or not.
4) Have the epidemic and climate change had a large impact on sugarcane planting? What specific year's sugarcane extent does map reflect? Or is it the combined acreage for the three years from 2019 to 2022?
Minor Comments:
1) The heading "Area" in Table 1 might change into "Country" to align with Table 2 & 3.
2) Line 52, the abbreviation GEDI should appear after the first occurrence of the full name.
3) Line 82, what is the threshold of the cloud probability you used for masking?
Citation: https://doi.org/10.5194/essd-2024-121-RC1 -
RC2: 'Comment on essd-2024-121', Anonymous Referee #2, 17 May 2024
Dear authors,
How to map sugarcane is vital especially at global scale, this study makes full use of GEDI and Sentinel-2 imagery to generate the sugarcane maps for top 13 producing countries, and achieving >80% agreement.
However, there are several issues in the current manuscript as:
- The novelty of the method is weak, it has been published in their previous works in 2023. The scope of ESSD aim to be innovative not only in terms of results, but also in terms of methodology.
- The method cannot convince me in some key steps.
- How to generate accurate training samples? Authors mentioned that the GEDI can provide the canopy heights, however, the error of GEDI cannot be directly ignored. I think that the quality control in the GEDI data on GEE cannot solve the vertical error. Meanwhile, we also think only GEDI dataset cannot be used to derive high-confidence training samples, for example, the height of maize also reached tall height, so how to distinguish maize and sugarcane. How training samples for other land classes are obtained? How many training samples were used? The quality and size of training samples greatly affected the accuracy of mapping.
- Section 3.4, you reduce spatial artifacts during the mosaicking of adjacent cells by creating predictions for pixels in a 0.2o, it doesn't convince me either. Actually, you trained the classification models in each 2o×2o tile, so the spatial artifacts were caused by the difference in trained classification models.
- How to use the crop mask in ESA, ESRI and GLAD data is also unclear.
Results
- The classification maps are generated in each 2o×2o tile, and the relationships between tall months and kappa score in Figure 3 are analyzed at national scale. So how to determine the thresholds for tiles that span multiple countries.
- More descriptions about the Section 4.3.2 should be greatly strengthen, for example, why China achieved the lower F1 score of 0.47?
Citation: https://doi.org/10.5194/essd-2024-121-RC2
Data sets
Mapping sugarcane globally at 10 m resolution using GEDI and Sentinel-2 Stefania Di Tommaso, Sherrie Wang, Rob Strey, and David B. Lobell https://zenodo.org/records/10871164
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
370 | 182 | 29 | 581 | 25 | 25 |
- HTML: 370
- PDF: 182
- XML: 29
- Total: 581
- BibTeX: 25
- EndNote: 25
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1