Estimating dry matter content in forage grasslands using Sentinel-2 satellite data
(2024) In Student thesis series INES NGEM01 20231Dept of Physical Geography and Ecosystem Science
- Abstract
- Leaf-level vegetation dry matter and water content as plant traits have been commonly studied, but few studies have investigated the capabilities of satellite remote sensing in estimating community-level dry matter content as percentage (DMCaP) in forage crops for agricultural applications. It is a crucial quality factor for maximizing nutrient yield and producing palatable forage.
In this thesis, we evaluated the capabilities of primarily Sentinel-2 data, and combined with environmental variables, in estimating DMCaP in forage grasslands in Southern Norway. Two types of models were developed, the interpretable model and the Random Forest Regression (RFR) model. The evaluation of the model performance showed that the RFR model... (More) - Leaf-level vegetation dry matter and water content as plant traits have been commonly studied, but few studies have investigated the capabilities of satellite remote sensing in estimating community-level dry matter content as percentage (DMCaP) in forage crops for agricultural applications. It is a crucial quality factor for maximizing nutrient yield and producing palatable forage.
In this thesis, we evaluated the capabilities of primarily Sentinel-2 data, and combined with environmental variables, in estimating DMCaP in forage grasslands in Southern Norway. Two types of models were developed, the interpretable model and the Random Forest Regression (RFR) model. The evaluation of the model performance showed that the RFR model outperformed the interpretable model, achieving an RMSE of 3.88%, and the use of environmental predictors further improved the estimation accuracy (RMSE = 2.90%). The interpretable model performed better when applied to individual fields respectively, indicating its better suitability for highly local-scaled data, and there was no significant difference in performance when using different combinations of water index and vegetation index. Although the performance evaluation suggested that DMCaP can be estimated with acceptable accuracy, model development was challenged by the inadequate temporal and spatial resolution of Sentinel-2 data, noise and uncertainties in the dataset, and the small data volume.
This thesis underscores the potential of satellite remote sensing for estimating the DMCaP quality factor in agricultural applications, providing valuable insights for the forage-harvesting process and pasture management. It addresses the gap in estimating vegetation dry matter content between the agricultural and academic communities, as well as the limitations observed in previous studies when matching satellite and in-situ data pairs. (Less) - Popular Abstract
- The dry matter and water content of leaves (usually species-specific) have been commonly studied, but few studies have investigated the capabilities of satellite remote sensing in estimating dry matter content as percentage (DMCaP) in mixed types of forage plants for agricultural applications. It is a crucial quality factor for maximizing nutrient yield and producing tasty forage for animals.
In this thesis, we evaluated the capabilities of primarily Sentinel-2 data, and combined with weather, in estimating DMCaP in forage grasslands in Southern Norway. Two types of models were developed, the interpretable model and the machine learning Random Forest Regression (RFR) model. The results showed that the RFR model performed better than... (More) - The dry matter and water content of leaves (usually species-specific) have been commonly studied, but few studies have investigated the capabilities of satellite remote sensing in estimating dry matter content as percentage (DMCaP) in mixed types of forage plants for agricultural applications. It is a crucial quality factor for maximizing nutrient yield and producing tasty forage for animals.
In this thesis, we evaluated the capabilities of primarily Sentinel-2 data, and combined with weather, in estimating DMCaP in forage grasslands in Southern Norway. Two types of models were developed, the interpretable model and the machine learning Random Forest Regression (RFR) model. The results showed that the RFR model performed better than the interpretable model, and the use of weather data further improved the estimation of DMCaP. The interpretable model performed better when built on individual fields respectively, suggesting that it is more suitable for data collected from nearby locations. Vegetation indices are numerical values used to quantify properties of vegetation, calculated based on different spectral bands of a remote sensing system. The use of different combinations of vegetation indices did not affect the performance of the interpretable model. The average quality and small size of the sample dataset made the estimation not very easy.
This thesis shows how satellite remote sensing can be utilized to estimate the DMCaP quality factor in small-scale forage grasslands, and thereby provide valuable insights for the forage-harvesting process and pasture management in agricultural practices. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9167786
- author
- Wang, Fan LU
- supervisor
-
- Zhanzhang Cai LU
- Zheng Duan LU
- organization
- course
- NGEM01 20231
- year
- 2024
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Physical Geography and Ecosystem Analysis, dry matter content, machine learning, remote sensing, forage grasslands, vegetation index
- publication/series
- Student thesis series INES
- report number
- 667
- language
- English
- id
- 9167786
- date added to LUP
- 2024-06-24 16:44:07
- date last changed
- 2024-06-24 16:44:07
@misc{9167786,
abstract = {{Leaf-level vegetation dry matter and water content as plant traits have been commonly studied, but few studies have investigated the capabilities of satellite remote sensing in estimating community-level dry matter content as percentage (DMCaP) in forage crops for agricultural applications. It is a crucial quality factor for maximizing nutrient yield and producing palatable forage.
In this thesis, we evaluated the capabilities of primarily Sentinel-2 data, and combined with environmental variables, in estimating DMCaP in forage grasslands in Southern Norway. Two types of models were developed, the interpretable model and the Random Forest Regression (RFR) model. The evaluation of the model performance showed that the RFR model outperformed the interpretable model, achieving an RMSE of 3.88%, and the use of environmental predictors further improved the estimation accuracy (RMSE = 2.90%). The interpretable model performed better when applied to individual fields respectively, indicating its better suitability for highly local-scaled data, and there was no significant difference in performance when using different combinations of water index and vegetation index. Although the performance evaluation suggested that DMCaP can be estimated with acceptable accuracy, model development was challenged by the inadequate temporal and spatial resolution of Sentinel-2 data, noise and uncertainties in the dataset, and the small data volume.
This thesis underscores the potential of satellite remote sensing for estimating the DMCaP quality factor in agricultural applications, providing valuable insights for the forage-harvesting process and pasture management. It addresses the gap in estimating vegetation dry matter content between the agricultural and academic communities, as well as the limitations observed in previous studies when matching satellite and in-situ data pairs.}},
author = {{Wang, Fan}},
language = {{eng}},
note = {{Student Paper}},
series = {{Student thesis series INES}},
title = {{Estimating dry matter content in forage grasslands using Sentinel-2 satellite data}},
year = {{2024}},
}