Publications

Chen, YL; Lu, DS; Luo, LF; Pokhrel, Y; Deb, K; Huang, JF; Ran, YH (2018). Detecting irrigation extent, frequency, and timing in a heterogeneous arid agricultural region using MODIS time series, Landsat imagery, and ancillary data. REMOTE SENSING OF ENVIRONMENT, 204, 197-211.

Abstract
Mapping irrigated area, frequency, timing, and amount is important for sustainable management of water resources in semi-arid and arid regions. Various studies exist on the mapping of irrigation using remote sensing and census statistics, but they mainly focus on the mapping of irrigation extent without taking frequency and timing into account. In this study, we proposed a new approach to extract irrigation attributes including irrigation extent, frequency and timing using multi-source data moderate resolution imaging spectroradiometer (MODIS), Landsat, and ancillary data. A time-series dataset with 30-m spatial resolution was generated by fusing 480-m time-series MODIS and Landsat imagery. We used the greenness index (the ratio of NIR and green spectral bands) to detect irrigation events during the first half of the growing season. Rainfall events were assumed as water supplement events along with irrigation events. The number of water supplement stages were then recorded cumulatively when a water supplement event was detected using a threshold-based model. To estimate the possible dates of each water supplement stage, Gaussian process regression and linear regression models were applied. The new framework was applied to the Hexi Corridor in northwestern China, an intensively irrigated region with a semi-arid climate. Results show that the overall accuracy of water supplement stage using the proposed method is 87%. Validation of the number of water supplement stages and possible dates of water supply by GRP model with a "strict" (or "loose") assessment method shows an overall accuracy of 55% (94%) and 59% (89%), respectively. The good accuracy of the additional independent validations for different years and sites demonstrates the robustness of the proposed method, suggesting the general applicability to other regions. Overall, this research demonstrates that the proposed method is promising in detecting irrigation attributes such as frequency and timing which have not been explored in previous research.

DOI:
10.1016/j.rse.2017.10.030

ISSN:
0034-4257