Skip to main content
ARS Home » Plains Area » Fort Collins, Colorado » Center for Agricultural Resources Research » Water Management and Systems Research » Research » Publications at this Location » Publication #385191
Research Project: Improving the Sustainability of Irrigated Farming Systems in Semi-Arid Regions

Location: Water Management and Systems Research

Title: Estimating fractional vegetation cover of maize under water stress from UAV multispectral imagery using machine learning algorithms

Author
item NIU, YAXIAO - Northwest A&f University
item HAN, WENTING - Northwest A&f University
item Zhang, Huihui
item ZHANG, LIYUAN - Northwest A&f University
item CHEN, HAIPING - Northwest A&f University

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/17/2021
Publication Date: 8/30/2021
Citation: Niu, Y., Han, W., Zhang, H., Zhang, L., Chen, H. 2021. Estimating fractional vegetation cover of maize under water stress from UAV multispectral imagery using machine learning algorithms. Computers and Electronics in Agriculture. 189. Article e106414. https://doi.org/10.1016/j.compag.2021.106414.
DOI: https://doi.org/10.1016/j.compag.2021.106414

Interpretive Summary: To estimate crop fractional vegetation cover (FVC) rapidly and accurately under various water stress, we conducted studies in a maize field located in Inner Mongolia, China with different irrigation levels during the entire 2018 and 2019 growing seasons. Specifically, the applicability of the fixed-threshold method proposed in our recent study was tested for UAV RGB imagery and three FVC estimation models were established based on vegetation indices derived from UAV multispectral imagery by three regression algorithms (RF: random forest, ANN: artificial neural network, and MLR: multivariate linear regression) and tested for different maize growing seasons, growth stages, and water stress. The results show that the initial and adjusted classification threshold of -2.72 and 4.60 of the fixed-threshold method was not influenced by the differences of image sensors on different platforms. In both maize growing seasons, significant correlations were found between vegetation indices and FVC references. However, an RF model established using 2018 data had the best versatility when it was used to estimate maize FVC in 2019. Overall, this study provides a low cost and easy way to estimate maize FVC and its inter-field variability under various water statuses in different maize growing seasons or growth stages.

Technical Abstract: Crop water stress is an inevitable and increasing challenge for agriculture. Management of water stress requires rapid, accurate estimation of crop fractional vegetation cover (FVC) under various water stress. We conducted studies in a maize field located in Inner Mongolia, China with different irrigation levels during 2018 and 2019 growing seasons to test applicability of the fixed-threshold method (proposed in our recent study) for UAV RGB imagery. Three FVC estimation models were established based on vegetation indices derived from UAV multispectral imagery by three regression algorithms (RF: random forest, ANN: artificial neural network, and MLR: multivariate linear regression) and were tested for different maize growing seasons, growth stages, and water stress. Initial and adjusted classification thresholds of -2.72 and -4.60 of the fixed-threshold method were not influenced by the differences of image sensors on different platforms. In both maize growing seasons, significant correlations were found between vegetation indices and FVC references, with the highest r of 0.93 for the ratio vegetation index in 2018 and the highest r of 0.96 for the normalized difference vegetation index in 2019. In 2018, when five-fold cross-validation was adopted, all the three regression algorithms could be used successfully to estimate maize FVC with adjusted coefficient of determination of 0.892, 0.890, and 0.893, and mean root mean square error of 0.066, 0.067, and 0.066, respectively. However, RF model had the best versatility when these models established in 2018 were used to estimate maize FVC in 2019 for different growth stages and water stress. The low estimation accuracy for high FVC levels was the reason why MLR model could not be used in the other maize growing season. Overall, this study provides a low cost and easy way to estimate maize FVC and its inter-field variability under various water status in different maize growing seasons or growth stages.