Author(s): Chih Chao Ho; Ming Xing Li; Jian Cheng Liao; Shih Wei Chiang; Tsu Chiang Lei

Linked Author(s):

Keywords: Satellite; Uncertainty Rice Crop Interpretation; Irrigation Water Demand Estimation

Abstract: Taiwan is actively implementing smart irrigation technology to manage water scarcity and reduce its impact on society and the environment. By using automated water gates and monitoring hydro-meteorological conditions, soil moisture, and water storage in field, irrigation schedules and quantities are adjusted based on climate variations. This precise, "on-demand supply" approach aims to increase water use efficiency and enhance the stability of regional water supplies. Despite advancements, challenges persist, such as the high costs of field sensor deployment and labor-intensive surveys needed for assessing irrigation zones. These hurdles limit the ability to gather real-time data on irrigation areas and water demand, which is critical for optimizing smart irrigation practices. To address these challenges, the study utilizes satellite imagery for its broad coverage, frequent updates, and multispectral data to accurately identify current rice cultivation areas. This satellite data is integrated with real-time hydro-meteorological information, soil types, water conveyance efficiency, crop growth stages, and crop coefficients to calculate irrigation needs in agricultural regions, providing crucial real-time data for effective smart irrigation management. The research introduces innovative models for analyzing rice cultivation and estimating real-time irrigation water usage. The overall architecture is shown as Figure 1, and the approach is described as follows. 1. Rice Crop Interpretation Model The rice crop interpretation model employs Sentinel-2 multispectral imagery, SCL products, maps of cultivated fields, and rice cultivation surveys data to accurately determine rice-growing areas. To enhance the precision of crop identification, an uncertainty decision analysis approach is employed, strengthening the outcomes of decision trees, neural networks, and logistic regression. 2. Irrigation Water Demand Estimation Model The TaiCropWat model estimates irrigation water demand by combining real-time hydro-meteorological data, crop data such as planting location and field size, and environmental factors like soil texture and canal water loss. This model calculates crop evapotranspiration and field water requirements for each growth stage. The study used satellite imagery from 2021 to classify first crop rice and estimate irrigation water needs in the Hsinchu area. The imagery was collected on February 4, March 16, March 26, April 5, May 15, and June 14. Classification accuracies ranged from 92.15% to 98.78%, with Kappa coefficients from 0.82 to 0.97. The lowest classification accuracy was on February 4 due to the coarse spatial resolution and smaller size of rice seedlings during early transplanting. Using data on rice cultivation areas and growth periods for the first crop rice -- covering the nursery, field preparation, and main crop stages from February 6 to July 10 -- the study incorporated hydro-meteorological, crop, and environmental data into the TaiCropWat model. This simulation provided crop evapotranspiration and field water requirements for each stage of crop growth. Results showed field leakage of 1,507 mm, effective rainfall of 430.13 mm, crop evapotranspiration of 362 mm, and field water requirements of 1,608.85 mm. The study demonstrates the feasibility of replacing traditional manual surveys with satellite remote sensing. The findings indicate that, except for the initial stage of crop growth (February), accuracy consistently exceeds 90%, enabling efficient and rapid acquisition of the actual rice cultivation area. Furthermore, utilizing the TaiCropWat model allows us to account for variations in field water storage due to rainfall and estimate the water required for crops. This approach provides crucial insights for smart irrigation and effectively addresses past challenges, such as the substantial costs associated with sensor installation and subsequent maintenance issues. In conclusion, this study, integrating satellite remote sensing and the TaiCropWat model, paves the way for precision agriculture in Taiwan. Through the application of these technologies, stakeholders can more accurately assess irrigation needs, optimize water resource management strategies, enhance agricultural productivity, and simultaneously reduce resource consumption and environmental impact.

DOI: https://doi.org/10.3850/iahr-hic2483430201-361

Year: 2024