the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Apr 2025
A 30m resolution annual cropland extent dataset of Africa in recent decades of the 21st century
Abstract. Accurate cropland mapping is essential for understanding agricultural dynamics in Africa, a critical global issue with significant implications for the Sustainable Development Goals (e.g., Zero Hunger). Large-scale cropland mapping encounters several challenges, including the varying landscape characteristics of cropland across different regions, extended cultivation periods, and the limited availability of reference data. The study developed a 30-meter resolution African annual cropland distribution (namely AFCD) dataset for Africa spanning the years 2000 to 2022. To extract this large-scale cropland distribution data, we employed random forest classification and Continuous Change Detection algorithms on the Google Earth Engine platform. Robust training samples were generated, and a locally adaptive model was applied for cropland extraction. The final output consists of annual binary Crop/Non-Crop maps from 2000 to 2022. Independent validation samples from numerous third-party sources confirm that the map’s accuracy is 0.86±0.01. A comparison of the cropland area estimates from AFCD with those of the FAO for Africa yielded an R-squared value of 0.86. According to our estimates, Africa’s cropland expanded from 194.35 Mha in 2000 to 210.92 Mha by 2022, marking a net increase of 8.53 %. Prior to 2005, changes in Africa’s cropland area were gradual, but after 2006, there was a marked acceleration in cropland expansion. Despite this continued growth, Africa also experienced significant cropland abandonment. By 2018, abandoned cropland accounted for 11.52 % of the total active cropland area. AFCD also avoided the misclassification of buildings, roads, and trees surrounding cropland, a common issue in LGRIP products. The study further highlights the unique advantage of AFCD in providing a dynamic annual cropland dataset at 30-meter resolution for Africa. This dataset is a crucial resource for understanding the spatial-temporal dynamics of cropland and can support policies on food security and sustainable land management. The cropland dataset is available at https://doi.org/10.5281/zenodo.14920706 (Lou et al., 2025).
- Preprint
(2213 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 27 May 2025)
-
RC1: 'Comment on essd-2025-133', Anonymous Referee #1, 21 Apr 2025
reply
The research successfully overcame the temporal coverage limit of current African cropland classification products, providing a valuable and timely dataset for tracking interannual changes in African croplands. The overall accuracy compared to the independent validation dataset is good. It used a new sample generation approach, however, the mapping accuracy in some areas remains low, like previous studies. The study is well-designed and well-written. I have some minor comments below.
- Title: Consider this title instead – An Annual Cropland Extent Dataset for Africa at 30m Spatial Resolution from 2000 to 2022
- Abstract, line 25: “The study developed a 30-meter resolution African annual
- Introduction, Line 40: change “croplands play is of critical importance” to “croplands are of critical importance”
- cropland distribution (namely AFCD) dataset for Africa spanning the years 2000 to 2022.”— delete “for Africa”.
- Line 30-31 mentioned results of R-square to compare with other product, if it is a part of evaluation accuracy assessment maybe worth to include it in 3.4 Accuracy assessment as well
- Line 36: what does LGRIP stand for?
- Line 76: what does GCEP stand for? Also is it a mistake in the citation? the cite ‘Xiong et al., 2017b’ refers to dataset Global Food Security-support Analysis Data (GFSAD) not GCEP
- Mention table 1 somewhere in L101-103 could make the context more clear and help to understand all the dataset names in the following paragraph
- Should the Fig 1 (a) -(e) be mentioned somewhere? If the plan is to only mention (f) in L197 maybe just make it an individual figure. Also it’s quite confusing what is the relationship between (a) - (e) and (f), there are two similar color bar for different meaning
- For Fig 2 (b), a subplot showing location (which part of Africa) and size of the area could be helpful for the reader to Also if the plan is to mention (b) before (a), the sequence can be switched.
- Line 136 mentioned ‘samples were randomly selected for further validation by students and experts’: quite confused about what kind of validation has been done here, brief introduction about the validation method and validation result could be better
- Fig 2 in L142 is a typo?
- In section 3.1, the step of reclassification existing LULC into the cropland/ non-cropland under your definition (the ‘remapping’ in L160) should be mentioned; it can better explain how you make the different LULC definition consistent for your use.
- In Line 195, ‘Previous studies have shown that the classification accuracy of this algorithm is not significantly affected by the specific parameter values.’ maybe worth citing which studies
- In Line 208, ‘the algorithm first decomposes the time-series data into trend and periodic components using the Fourier transform’ and then? feel some paragraph is missing here for how you capture changes
- What is m and what is N in Equation (2)? - I assume m is category number and N is sample number but you need to specify here; eg. if your binary category is 0 and 1, the m should be [0,1] but in that case the k should start 0
- You mentioned ‘similarity matrices’ in L245 but it is hard to understand with simply a cite. Similarity matrix is the confusion matrix in the paper you cited, which is the same thing you used for OA F1 calculation, not something that paper created. If you intended to express you used it like Phalke et al., 2020 did, briefly mention how it can be better with just a cite.
- In Table 2 (L280), why not all the highest values are highlighted? Eg. Accuracy and Precision for Mali
- What do ‘control samples’ and ‘expert samples’ refers to in L285 and L286
- Some details for plot explanation could be added in caption of Figure 6, eg. what is the red/blue point, what is the purple line etc
- The discussion of advantages and limitations is objective and fair.
Citation: https://doi.org/10.5194/essd-2025-133-RC1 -
RC2: 'Comment on essd-2025-133', Anonymous Referee #2, 22 Apr 2025
reply
This paper developed an annual cropland extent for Africa between 2000 and 2022 at a 30-meter resolution by generating training samples from multiple existing land cover products and refining the performance of the derived cropland extent product using the CCDC algorithm. This work is labour-intensive, and the evaluation demonstrates soundness with clear logic. Before recommending it for publication, I would like to raise several concerns that may be helpful in improving this paper.
Major concerns:
- The training data incorporates multiple existing products, yet inconsistencies exist in their cropland definitions. How did the authors address the noise introduced by such definitional discrepancies?
- The Discussion section mentions the utilization of AFCD data for spatial mapping of abandoned cropland in Africa, which represents a highly meaningful endeavour. However, it should be noted that the authors did not specify how abandoned land was defined in this study. We recommend that the authors incorporate relevant descriptions regarding the operational definition of abandoned cropland.
- The Geo-Wiki sample (Laso Bayas et al., 2017) is based on 300m PROBA-V imagery, but AFCD is a 30m product. Does this scale difference lead to validation bias?
- Some of the abbreviations are not explained in detail when they first appear (e.g. LGRIP, CCDC) and it is suggested that the full names be added.
- The terms "cropland" and "farmland" are used interchangeably in the text, and it is suggested that they be standardised as "cropland".
Minor concerns:
- Title: "recent decades of the 21st century" Vague (suggest clarification of year range)
- The abbreviation "SDG" should be defined at its first occurrence in line 43, rather than being introduced later in line 54.
- In Line 25, redundant "for Africa" (appears twice).
- In Line 41, "croplands play is of critical importance" → "croplands plays a critical role"
- In Line 55, "one in five people undernourished" → "one in five people was undernourished"
- In Line 76, "GCEP" is undefined and potentially mis-cited (Xiong et al., 2017b refers to GFSAD, not GCEP).
- In Line 97, duplicate use numeric (2).
- In Line 136, "samples were randomly selected for further validation by students and experts" is vague.
- In Line 142, "combines" should be "combined".
- In Line 177, the third-level heading "3.3.1" should be corrected to "3.2.1"; revise Line 199 to "3.2.2" and Line 223 to "3.3” for consistency.
- In Line 208, delete "first".
- In Line 210 (Equation 1), there appears to be an extraneous "&" symbol that may be a typesetting error. Please verify the mathematical expression's integrity and ensure the formula complies with standard notation conventions.
- In Line 252-254, "The result of this study is developing a new annual cropland dynamic map of Africa at 30-m (Fig. 5). The visual evaluation of the current cropland product shows that cultivated areas are accurately represented across diverse agricultural landscapes throughout Africa (Fig. 5). As shown in the figure above, …" The redundant repeated descriptions of Figure 5 in the text may disrupt the flow; consider consolidating them to strengthen the paragraph's logical coherence.
- 14. In Line 304 (Figure 6), The plot elements of the red and blue dots and the purple line should be explained.
- In Line 415, incorrect serial number.
Citation: https://doi.org/10.5194/essd-2025-133-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 263 | 63 | 10 | 336 | 7 | 5 |
- HTML: 263
- PDF: 63
- XML: 10
- Total: 336
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1