Estimation of crop yield in regions with mixed crops using different cropland masks and time-series MODIS data

Citation

Liu, J., Huffman, T., Shang, J., Qian, B., Dong, T., Zhang, Y., Jing, Q. 2016. Estimation of crop yield in regions with mixed crops using different cropland masks and time-series MODIS data. IGARSS 2016, July 10-15, Beijing, China. doi:10.1109/IGARSS.2016.7730868.

Plain language summary

We investigated the potential of using time-series vegetation index (e.g., NDVI) for mapping spatial variability of cropland productivity in south-western Ontario, Canada. Major crop types in the Mixedwood Plains Ecozones were identified using crop phenological indicators derived from time-series NDVI in order to establish crop-specific mask. Results show that county level crop yield is better correlated with vegetation index at peak growth time derived using crop specific masks than using a general cropland mask for the three major crops corn, soybean and winter wheat.

Abstract

Cropland productivity, characterized by crop yields, is determined by soil and meteorological conditions as well as management practices, e.g., crop types and their associated phenological cycles. As canopy spectral reflectance is governed by vegetation photosynthetic activities and is indicative of primary productivity, we investigated the potential of using time-series NDVI for mapping spatial variability of cropland productivity in south-western Ontario, Canada. NDVI was derived from the 8-day composite 250-m MODIS surface reflectance data, using a general cropland mask and crop specific masks, respectively. It was observed that for the three major annual crops (corn, soybean and winter wheat), using a general cropland mask, the strongest positive linear correlation between county level crop yield and NDVI was reached between the end of July and early August; whereas using crop specific masks the time of strongest linear correlation for wheat was shifted to between mid-May and early June. Large differences in phenological patterns and interleaved spatial distribution of these different crops led to difficulties for yield estimation using low resolution remote sensing data
in this region.

Publication date

2016-11-03