Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Etats Unis → 2021 → Estimation of Spatial Change in Cropland Area and Evaluation of Irrigation Performance in Imperial Valley Using Remotely Sensed Data

University of Nevada, Las Vegas (2021)

Estimation of Spatial Change in Cropland Area and Evaluation of Irrigation Performance in Imperial Valley Using Remotely Sensed Data

Poudel Usha

Titre : Estimation of Spatial Change in Cropland Area and Evaluation of Irrigation Performance in Imperial Valley Using Remotely Sensed Data

Auteur : Poudel Usha

Université de soutenance : University of Nevada, Las Vegas

Grade : Master of Science in Engineering (MSE) 2021

Résumé partiel
The Imperial Valley (IV) in the US is an extensively irrigated agricultural region, which includes multiple crops changing on an annual and semiannual basis. The valley is facing grave concerns about water management due to its semi-arid environment, water intensive crops, and limited water supply. A simple, inexpensive, and repeatable method to detect changes in cropping patterns may assist irrigation managers to understand crop diversification and associated consumptive use. In addition, a spatial assessment of existing water irrigation system performance and productivity is crucial to benchmark and improve current water management strategies. This thesis estimates the spatial pattern of change in crop distributions from 2018 to 2019 across the IV, using remotely sensed data with high resolution and a machine learning algorithm. Furthermore, it also quantifies the irrigation performance indicators based on the equity, adequacy, and water productivity of water intensive crops utilizing remote sensing, Vegetation indices, and county level crop production statistics.

First, we addressed the spatial analysis of cropland change in an agricultural field of the IV over 2018 and 2019. Optical images from the Sentinel-2 platform were used to develop an annual cropland map using a random forest algorithm in R version 4.0.2. The reflectance from the Sentinel images and Normalized Difference Vegetation Index (NDVI) served as a predictor variable. A cropland data layer was utilized to identify the field’s crop type for ground truthing. We used the dataset provided by the United States Department of Agriculture to access the accuracy of classification. The changes in cropping patterns were quantified by preparing a transition matrix through image the differencing technique in Geographical Information System (GIS). The spatial analysis of change was characterized by generating a map showing the change in cropping proportion for major crop types over the two-year period. We obtained the overall classification accuracy of 85% for each year.

Présentation

Version intégrale (5,5 Mb)

Page publiée le 27 novembre 2022