XJ-MC30: 30-m Annual Multi-Crop Maps for Xinjiang, China (2000-2025)

XJ-MC30 is a 30-meter resolution annual crop type dataset covering the Xinjiang region of China from 2000 to 2025. The dataset, created by Lichang Yin, maps seven major crops using the Phenology-Aware Temporal-Consistency framework trained on Landsat imagery and polygon annotations. It comprises 26 years of single-band GeoTIFF files with an overall accuracy of approximately 0.80 against field samples.

Use Cases

• Analyzing long-term crop planting trends and shifts based on annual maps from 2000 to 2025.
• Validating agricultural models and statistics using crop-specific R² values ranging from 0.59 to 0.95.
• Assessing the impact of the PTCNet framework on classification performance, which yielded F1-score gains of 0.14–0.63 for certain crops.
• Studying regional agricultural patterns within the geographic extent of 73°40′–96°18′E, 34°25′–48°10′N.

Strengths

• Provides 26 years of spatially consistent annual data from 2000 to 2025.
• Achieves an overall accuracy of 0.80 against 2018 field samples and 0.79 against 2023 third-party samples.
• Shows strong agreement with agricultural statistics, with a provincial-level R² of 0.93.
• Offers high spatial resolution of 30 meters for detailed crop mapping.

Limitations

• Early years, especially 2012, have higher uncertainty due to limited Landsat-7 observations.
• Some crops, such as maize in Kashgar, are underestimated because of regional phenological deviations.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Lichang Yin

Collection Method

Generated using the Phenology-Aware Temporal-Consistency framework (PTCNet) integrating intra‑annual phenological modeling with inter‑annual contrastive learning, trained on polygon‑based annotations from 2019 and Landsat time‑series imagery.

Time Range

2000 to 2025

Freshness

Last updated 2026-05-09 11:33:04

Geography

Xinjiang Uygur Autonomous Region, China (73°40′–96°18′E, 34°25′–48°10′N)