Sentinel-2-based crop type map of Jos Plateau, Nigeria (small holder sub-Sahara Africa).

Crop type mapping in smallholder settings in sub-Saharan Africa (SSA) is complex, given the irregular field patterns and small field sizes, diverse management practices, highly heterogeneous growing environments, and other mapping difficulties associated especially with smallholder farming systems. Adding to that, frequent cloud cover during the growing season renders crop type mapping, including intercropping, a challenging exercise. Therefore, freely available satellite data of the European Copernicus Program, specifically Sentinel-2A/B (S2), allow for major advances in mapping and monitoring crop types in smallholder farming systems of SSA.

The dataset features crop type map of one of Nigeria’s most complex agricultural settings. We generated a crop type map of the Jos Plateau in Nigeria based on freely available S2 time series to map the main crops—maize and potato—and intercropping systems including these two crops. We analyzed the spectral-temporal behavior of the mixed crop classes to improve our understanding of inter-class spectral mixing. We preprocessed S2 time series and derived spectral-temporal metrics (STM) from S2 spectral bands for the main temporal cropping windows. These STMs were used as input features in a hierarchical random forest classification, first identifying annual croplands and second the crop types therein. Our results provide the first wall-to-wall crop type map for this key agricultural region of Nigeria.

Our cropland identification had an overall accuracy of 84%, while the crop type map achieved an average accuracy of 72% for the five relevant crop type classes. Our crop type map shows distinctive clusters and regional variations in the distribution of crop types. However field-level crop type demarcation is relatively poor due to small field sizes within the smallholder domain in SSA.

Spatial coverage
The map was produced at 10m spatial resolution and covers the heterogonous Jos Plateau region of Nigeria. The region is a mountainous landscape covering about 4,150 km² and located in central Nigeria. Jos Plateau is the dominant producer of potato in Nigeria and the fourth-largest producer of maize.

Temporal extent
Three satellite data archives were used:
- Landsat ETM: this data was used solely for co—registration of the Sentinel 2 data. In principle, the Landsat NIR band was used as a base layer over a 5-year period of acquisitions (in our case, 2015–2020) and co-registered via a hierarchical pyramid layer approach.

- Sentinel-2: all available data from 2019 was used. This was the main data for crop type mapping and phenological analysis.

- SkySat: this data was tasked for the 2020 cropping season, it was acquired at five critical windows and used solely for validation.

- Reference data: Field data were collected during the 2019 and 2020 cropping seasons. We collected 530 samples of potato and maize fields in 2019 and 690 samples in 2020. These samples cover all ecological gradients and different management practices on the Jos Plateau.

Data format
The data is maintained as raster format. The projection is EPSG: 3857. The images are compressed GeoTiff files (*.tif) and can be viewed in most Geographic Information Systems (GIS) packages. The numeric values in the dataset represent the different crop types identified by the random forest classification. Non-croplands within the study area are flagged as "32767" and areas outside of the region of interest are labeled as NA values.  The crop classes are categorized as follows; 1-maize, 2-potato, 3-potato-maize, 4-maize-legumes, 5-other crops.

Further information
For further information, please see the publication or contact Esther Shupel Ibrahim (esther.shupel.ibrahim@hu-berlin.de).

Publication
Ibrahim, E.S., Rufin, P., Nill, L., Kamali, B., Nendel, C., & Hostert, P. (2021). Mapping Crop Types and Cropping Systems in Nigeria with Sentinel-2 Imagery. Remote Sens. 2021, 13(17), 3523; https://doi.org/10.3390/rs13173523

Acknowledgments
The dataset was generated by FORCE v. 3.1 (paper, code), which is freely available software under the terms of the GNU General Public License v. >= 3. Sentinel imagery were obtained from the European Space Agency and the European Commission. The Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ), Jos Plateau State, Nigeria, provided relevant reports and contacts of farmer associations. Likewise, we thank the Plateau Agriculture Development Program, National Root Crops Research Institute-Potato (Kuru-Jos) and Daniel Lenka of the Department of Agriculture, University of Jos, for providing additional supporting documents. The authors appreciate Rifkatu Ibrahim, Joshua Azi Aware, Serah Habu, Alloycious Dere, Shomboro Karau, Samuel Odi, and Atang Itse for their field support in 2019 and independently collecting field data for the 2020 cropping season. We also acknowledge the cooperation of 119 farmers (questionnaire respondents) on the Jos Plateau, they provided relevant information

Funding
This research is funded by the Nigerian-German Postgraduate Training Programme PhD, 2019 (57473408). The National Centre for Remote Sensing, Jos Plateau State, Nigeria, Humboldt University of Berlin, Germany, and Leibniz Centre for Agricultural Landscape Research (ZALF), Brandenburg, Germany offered the necessary support.