Scholarlink

<p>Soil is the most valuable natural resource for agriculture, containing macro and micro nutrients for healthy vegetation. Soil nutrient estimation is necessary for healthy crop yield. The objective of this study was to estimate soil available Nitrogen (AN), available Phosphorous (AP) and available Potassium (AK) using Landsat 8 OLI satellite data and Support Vector Regression (SVR). Around 40 surface soil samples were collected from the study area and the soil AN, AP, and AK values were obtained through laboratory analysis. The spectral values were collected using an assembled handheld Spectroradiometer (410-900nm). NIR band of Landsat 8 OLI was used to obtain spectral reflectance of soil unsampled pixels. SVR model was used to estimate the soil AN, AP, and AK values for the unsampled pixels of the study area, another unsampled agricultural land which is located nearby the study area. SVR model gives better performance for soil AN (coefficient of determination (R<sup>2</sup>) = 0.856, root mean squared error (RMSE) = 0.706), followed by AP (R<sup>2</sup>=0.735, RMSE =3.014) and AK (RMSE = 5.355, R<sup>2</sup> =0.705). Estimation results of SVR model for another unsampled agricultural land shows (AN: R<sup>2</sup>=0.652, RMSE = 3.635); (AP: R<sup>2</sup>=0.683, RMSE =0.922); (AK: R<sup>2</sup>=0.715, RMSE of 5.355).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Pixel Based Soil Available NPK Estimation Using Landsat 8 OLI Satellite Data and Support Vector Regression

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • J Mowshika,
  • S Jayalakshmi,
  • M Balamurugan

摘要

Soil is the most valuable natural resource for agriculture, containing macro and micro nutrients for healthy vegetation. Soil nutrient estimation is necessary for healthy crop yield. The objective of this study was to estimate soil available Nitrogen (AN), available Phosphorous (AP) and available Potassium (AK) using Landsat 8 OLI satellite data and Support Vector Regression (SVR). Around 40 surface soil samples were collected from the study area and the soil AN, AP, and AK values were obtained through laboratory analysis. The spectral values were collected using an assembled handheld Spectroradiometer (410-900nm). NIR band of Landsat 8 OLI was used to obtain spectral reflectance of soil unsampled pixels. SVR model was used to estimate the soil AN, AP, and AK values for the unsampled pixels of the study area, another unsampled agricultural land which is located nearby the study area. SVR model gives better performance for soil AN (coefficient of determination (R2) = 0.856, root mean squared error (RMSE) = 0.706), followed by AP (R2=0.735, RMSE =3.014) and AK (RMSE = 5.355, R2 =0.705). Estimation results of SVR model for another unsampled agricultural land shows (AN: R2=0.652, RMSE = 3.635); (AP: R2=0.683, RMSE =0.922); (AK: R2=0.715, RMSE of 5.355).