Improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS
Li, Xueying LU
(2025)
- Abstract
- Meeting the food demand of a growing global population with limited agricultural resources is one of the greatest challenges of the 21st century. Crop yield is highly sensitive to weather variability and climate extremes, which are becoming more frequent due to climate change. Accurate regional yield prediction is essential for helping farmers adapt, ensuring food security, and strengthening the resilience of agriculture. This PhD project focused on improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator), resulting in four research papers. Paper I introduced the first application of a triple collocation (TC)-based merging framework to... (More)
- Meeting the food demand of a growing global population with limited agricultural resources is one of the greatest challenges of the 21st century. Crop yield is highly sensitive to weather variability and climate extremes, which are becoming more frequent due to climate change. Accurate regional yield prediction is essential for helping farmers adapt, ensuring food security, and strengthening the resilience of agriculture. This PhD project focused on improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator), resulting in four research papers. Paper I introduced the first application of a triple collocation (TC)-based merging framework to evaluate the error structure of existing global evapotranspiration (ET) products and merge new ET datasets over the Nordic Region. The satellite-derived ET product Penman-Monteith-Leuning Version 2 (PML-V2) demonstrated the best overall performance among the selected ET products. Validated against Integrated Carbon Observation System (ICOS) in-situ measurements, the merged ET datasets outperformed individual parent products in terms of multiple evaluation metrics. This study provided reliable ET estimates to support the subsequent crop yield prediction. Paper II developed a novel framework for estimating spring barley yields at the district level in southern Sweden using meteorological data and multi-source satellite datasets with the random forest (RF) approach. The combination of vegetation indices (VIs) and solar-induced fluorescence (SIF) achieved high accuracy in crop yield estimation in April and May, suggesting that barley yield can be reliably forecasted two months prior to harvesting. Adding the monthly ET in June had slight contributions to the modelling performance. Paper III enhanced the performance of LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator) in simulating district-level crop yields in southern Sweden by calibrating the model with observed yield and satellite-based ET data. Calibration with observed yield significantly improved accuracy for spring barley and winter wheat, while adding the satellite-based ET product PML-V2 led to only moderate gains. The calibrated model also effectively assessed the drought impacts of 2018, accurately estimating yield losses for both crops. Paper IV assessed the impacts of future climate change on crop yields in southern Sweden using the calibrated LPJ-GUESS model, driven by 3-km high-resolution climate projections. The results showed significant yield increases for spring barley and winter wheat by the end of the century under both Representative Concentration Pathways (RCP)4.5 and RCP8.5 scenarios. Rising carbon dioxide levels and warmer growing-season temperatures drove yield improvements, while reduced precipitation (i.e., drought) was expected to sharply decrease yields. This PhD project developed a robust framework for crop yield prediction using two approaches, following a state-of-the-art TC-based accuracy assessment of existing ET datasets. Our results provided a solid foundation for improving agricultural management in Sweden and supporting global efforts to enhance food security under climate change. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/bd2b0db3-2c4e-4679-97d6-acfcae26d630
- author
- Li, Xueying
LU
- supervisor
-
- Zheng Duan LU
- Wenxin Zhang LU
- opponent
-
- Professor Guan, Kaiyu, University of Illinois Urbana-Champaign, USA.
- organization
- publishing date
- 2025-03-24
- type
- Thesis
- publication status
- published
- subject
- keywords
- Evapotranspiration, Triple collocation, Crop yield prediction, Satellite remote sensing, Machine learning, Process-based crop models, Climate change
- pages
- 69 pages
- publisher
- Lunds universitet
- defense location
- Pangea (GC2:229), Geocentrum 2
- defense date
- 2025-04-25 09:00:00
- ISBN
- 978-91-89187-55-9
- 978-91-89187-56-6
- project
- Improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS
- language
- English
- LU publication?
- yes
- id
- bd2b0db3-2c4e-4679-97d6-acfcae26d630
- date added to LUP
- 2025-03-24 13:52:20
- date last changed
- 2025-04-04 15:09:12
@phdthesis{bd2b0db3-2c4e-4679-97d6-acfcae26d630,
abstract = {{Meeting the food demand of a growing global population with limited agricultural resources is one of the greatest challenges of the 21st century. Crop yield is highly sensitive to weather variability and climate extremes, which are becoming more frequent due to climate change. Accurate regional yield prediction is essential for helping farmers adapt, ensuring food security, and strengthening the resilience of agriculture. This PhD project focused on improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator), resulting in four research papers. Paper I introduced the first application of a triple collocation (TC)-based merging framework to evaluate the error structure of existing global evapotranspiration (ET) products and merge new ET datasets over the Nordic Region. The satellite-derived ET product Penman-Monteith-Leuning Version 2 (PML-V2) demonstrated the best overall performance among the selected ET products. Validated against Integrated Carbon Observation System (ICOS) in-situ measurements, the merged ET datasets outperformed individual parent products in terms of multiple evaluation metrics. This study provided reliable ET estimates to support the subsequent crop yield prediction. Paper II developed a novel framework for estimating spring barley yields at the district level in southern Sweden using meteorological data and multi-source satellite datasets with the random forest (RF) approach. The combination of vegetation indices (VIs) and solar-induced fluorescence (SIF) achieved high accuracy in crop yield estimation in April and May, suggesting that barley yield can be reliably forecasted two months prior to harvesting. Adding the monthly ET in June had slight contributions to the modelling performance. Paper III enhanced the performance of LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator) in simulating district-level crop yields in southern Sweden by calibrating the model with observed yield and satellite-based ET data. Calibration with observed yield significantly improved accuracy for spring barley and winter wheat, while adding the satellite-based ET product PML-V2 led to only moderate gains. The calibrated model also effectively assessed the drought impacts of 2018, accurately estimating yield losses for both crops. Paper IV assessed the impacts of future climate change on crop yields in southern Sweden using the calibrated LPJ-GUESS model, driven by 3-km high-resolution climate projections. The results showed significant yield increases for spring barley and winter wheat by the end of the century under both Representative Concentration Pathways (RCP)4.5 and RCP8.5 scenarios. Rising carbon dioxide levels and warmer growing-season temperatures drove yield improvements, while reduced precipitation (i.e., drought) was expected to sharply decrease yields. This PhD project developed a robust framework for crop yield prediction using two approaches, following a state-of-the-art TC-based accuracy assessment of existing ET datasets. Our results provided a solid foundation for improving agricultural management in Sweden and supporting global efforts to enhance food security under climate change.}},
author = {{Li, Xueying}},
isbn = {{978-91-89187-55-9}},
keywords = {{Evapotranspiration; Triple collocation; Crop yield prediction; Satellite remote sensing; Machine learning; Process-based crop models; Climate change}},
language = {{eng}},
month = {{03}},
publisher = {{Lunds universitet}},
school = {{Lund University}},
title = {{Improving crop yield prediction in Sweden using satellite remote sensing and the ecosystem model LPJ-GUESS}},
url = {{https://lup.lub.lu.se/search/files/212191601/Kappa_Xueying.pdf}},
year = {{2025}},
}