Olive yield forecasting from remote sensing and climate datasets in the Jaen province (Spain)
(2024) In Master Thesis in Geographical Information Science GISM01 20241Dept of Physical Geography and Ecosystem Science
- Abstract
- The severe decline in olive oil production recorded in the last years is having an economic impact on many Mediterranean countries. The aim of this MSc thesis is to predict the olive yield by means of satellite vegetation indices and meteorological data mostly retrieved by satellite remote sensors. Estimating olive yield production using remote sensing data is reasonably new as few scientific papers focused on this topic so far. Collecting and analysing remote sensing data is cheap and could cover large extensions of olive orchards. It is more important that remote sensing data are usually publicly available, and their analyses do not require extraordinarily complex machinery and installation, apart from powerful computers. In this study... (More)
- The severe decline in olive oil production recorded in the last years is having an economic impact on many Mediterranean countries. The aim of this MSc thesis is to predict the olive yield by means of satellite vegetation indices and meteorological data mostly retrieved by satellite remote sensors. Estimating olive yield production using remote sensing data is reasonably new as few scientific papers focused on this topic so far. Collecting and analysing remote sensing data is cheap and could cover large extensions of olive orchards. It is more important that remote sensing data are usually publicly available, and their analyses do not require extraordinarily complex machinery and installation, apart from powerful computers. In this study several optical sensors have used: the MODIS sensor onboard Terra and Aqua satellites, the optical sensor ETM+ onboard Landsat-7, OLI on Landsat 8, and the MSI onboard on Sentinel-2.
Multiple linear regression technique has been used to evaluate and model for relationship of olive yield (dependent variable) to multiple independent variables, such as vegetation indices, precipitation, temperature, and drought index. The olive yield data were collected from the Survey on Areas and Crop Yields (ESYRCE) Survey. The thesis goal is to estimate which independent variable can best explain the variation in the olive yield in several olive orchards i.e. the dependent variable. The time range considered for the statistical analyses in this study is from November 2001 to November 2020.
The best correlation coefficients of olive yield with 936 monthly parameters (78 parameters for 12 months) were used to run the multilinear regression models. These coefficients corresponded to the vegetation indices and meteorological parameters for all months of the statistical series analysed in this MSc thesis. In this MSc thesis, twenty-four regression models have been run. The best multilinear regression models have a significance level of <0.01 and are those where both meteorological data and vegetation indices parameters have been included, based on the lowest p-value criteria.
The regression models based on the lowest p-values independently of the sensor, generated adjusted R2 values above 0.85. All regression models show the same pattern: the fewer observations, the higher R2 values, and lower-resolution climate datasets and MODIS reach lower R2 values than 0.85. The highest R2 values correspond to regression models using parameters retrieved from Sentinel-2 data, whereas climate datasets and MODIS sensor data retrieve R2 values below 0.84.
NDVI and NDMI are the best among 12 tested vegetation indices, suitable for predicting olive yield. The correlation coefficients of NDVI and NDMI from four platforms, Landsat-7, 8, Sentinel-2, and MODIS are similar. For other indices, there is no common vegetation index that can be used for olive yield prediction; ARVI is the best for Sentinel-2, whereas EVI and AVI work well with MODIS, and EVI and GCI gives the highest correlations with Landsat-7.
For climate variables, the water-related ones such as SPEI, topsoil moisture, rainfall, specific humidity, and evapotranspiration clearly show higher correlations than those temperature-dependent. On the other hand, the surface net thermal radiation seems better parameter to assess the impact of thermal stress rather than those temperature-based parameters. In summary, the promising results confirm that vegetation indices and weather data extracted from satellite sensors can be an accurate tool for forecasting olive crop yield on Mediterranean climates. (Less) - Popular Abstract
- The severe decline in olive oil production recorded in the last seven years is having an economic impact on many Mediterranean countries. The prediction of olive yield by means of satellite vegetation indices and meteorological data could be a very important tool to manage production and minimise economic uncertainties. This methodology will be easy to implement as collecting and analysing remote sensing data is cheap and could cover large cropland areas.
The multilinear regression technique has been used to evaluate and model for relationship between olive yield and multiple independent variables, such as vegetation indices from remote sensors, precipitation, temperature, and drought index calculated by monthly averages. The olive... (More) - The severe decline in olive oil production recorded in the last seven years is having an economic impact on many Mediterranean countries. The prediction of olive yield by means of satellite vegetation indices and meteorological data could be a very important tool to manage production and minimise economic uncertainties. This methodology will be easy to implement as collecting and analysing remote sensing data is cheap and could cover large cropland areas.
The multilinear regression technique has been used to evaluate and model for relationship between olive yield and multiple independent variables, such as vegetation indices from remote sensors, precipitation, temperature, and drought index calculated by monthly averages. The olive yield data were provided by the Spanish Ministry of Agriculture and the time range considered for the statistical analyses were from November 2001 to November 2020.
The best correlation coefficients of olive yield data with the monthly parameters were used to run the multilinear regression models. The regression models independently of the sensor, generated adjusted R2 values above 0.85. All regression models show the same pattern: the fewer observations, the higher R2 values, and lower-resolution climate datasets retrieve lower R2 values than 0.85. The highest R2 values correspond to regression models using parameters retrieved from Sentinel-2 data, whereas climate datasets and MODIS sensor data retrieved R2 values below 0.84.
The vegetation indices NDVI and NDMI are the best among 12 tested for predicting olive yield. For other indices, there is no common vegetation index that can be used for olive yield prediction.
Water-related variables such as SPEI, moisture, rainfall, specific humidity, and evapotranspiration clearly show higher correlations than those temperature-dependent. However, surface net thermal radiation seems to better assess the impact of thermal stress rather than those temperature-based parameters. In summary, vegetation indices and weather data extracted from satellite sensors can be an accurate tool for forecasting olive crop yield on Mediterranean climates. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9172345
- author
- Martinez Duran, Pedro LU
- supervisor
-
- Lars Eklundh LU
- Hongxiao Jin LU
- Lanhui Wang LU
- organization
- course
- GISM01 20241
- year
- 2024
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Keywords: Olive yield, Remote sensing, Meteorological data, Vegetation Indices, Multilinear regression
- publication/series
- Master Thesis in Geographical Information Science
- report number
- 179
- language
- English
- id
- 9172345
- date added to LUP
- 2024-08-26 13:49:44
- date last changed
- 2024-08-26 13:49:44
@misc{9172345,
abstract = {{The severe decline in olive oil production recorded in the last years is having an economic impact on many Mediterranean countries. The aim of this MSc thesis is to predict the olive yield by means of satellite vegetation indices and meteorological data mostly retrieved by satellite remote sensors. Estimating olive yield production using remote sensing data is reasonably new as few scientific papers focused on this topic so far. Collecting and analysing remote sensing data is cheap and could cover large extensions of olive orchards. It is more important that remote sensing data are usually publicly available, and their analyses do not require extraordinarily complex machinery and installation, apart from powerful computers. In this study several optical sensors have used: the MODIS sensor onboard Terra and Aqua satellites, the optical sensor ETM+ onboard Landsat-7, OLI on Landsat 8, and the MSI onboard on Sentinel-2.
Multiple linear regression technique has been used to evaluate and model for relationship of olive yield (dependent variable) to multiple independent variables, such as vegetation indices, precipitation, temperature, and drought index. The olive yield data were collected from the Survey on Areas and Crop Yields (ESYRCE) Survey. The thesis goal is to estimate which independent variable can best explain the variation in the olive yield in several olive orchards i.e. the dependent variable. The time range considered for the statistical analyses in this study is from November 2001 to November 2020.
The best correlation coefficients of olive yield with 936 monthly parameters (78 parameters for 12 months) were used to run the multilinear regression models. These coefficients corresponded to the vegetation indices and meteorological parameters for all months of the statistical series analysed in this MSc thesis. In this MSc thesis, twenty-four regression models have been run. The best multilinear regression models have a significance level of <0.01 and are those where both meteorological data and vegetation indices parameters have been included, based on the lowest p-value criteria.
The regression models based on the lowest p-values independently of the sensor, generated adjusted R2 values above 0.85. All regression models show the same pattern: the fewer observations, the higher R2 values, and lower-resolution climate datasets and MODIS reach lower R2 values than 0.85. The highest R2 values correspond to regression models using parameters retrieved from Sentinel-2 data, whereas climate datasets and MODIS sensor data retrieve R2 values below 0.84.
NDVI and NDMI are the best among 12 tested vegetation indices, suitable for predicting olive yield. The correlation coefficients of NDVI and NDMI from four platforms, Landsat-7, 8, Sentinel-2, and MODIS are similar. For other indices, there is no common vegetation index that can be used for olive yield prediction; ARVI is the best for Sentinel-2, whereas EVI and AVI work well with MODIS, and EVI and GCI gives the highest correlations with Landsat-7.
For climate variables, the water-related ones such as SPEI, topsoil moisture, rainfall, specific humidity, and evapotranspiration clearly show higher correlations than those temperature-dependent. On the other hand, the surface net thermal radiation seems better parameter to assess the impact of thermal stress rather than those temperature-based parameters. In summary, the promising results confirm that vegetation indices and weather data extracted from satellite sensors can be an accurate tool for forecasting olive crop yield on Mediterranean climates.}},
author = {{Martinez Duran, Pedro}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master Thesis in Geographical Information Science}},
title = {{Olive yield forecasting from remote sensing and climate datasets in the Jaen province (Spain)}},
year = {{2024}},
}