Estimation de la biomasse de canne par modélisation et télédétection. Application à la Réunion

Abstract : In the context of an increasing demand for sugar, the estimation of sugarcane biomass in smallholding farming countries (of which Reunion Island is an example) is an optimization lever of production and thus of sustainability for the sugar industry facing giants such as Brazil, India of China. The objective of this thesis is to explore the contribution of remote sensing for the estimation of sugarcane yields at field scale on Reunion Island. We organized our work in two main approaches: first, a methodological approach, where we explore the coupling (recalibration and forcing) between remote sensing data and modeling, and second, an operational approach where we compare three methods of yield estimation based on remote sensing : (1) empirical relationships between yield and vegetation indices computed from remote sensing data, (2) the efficiency models, with a low number of parameters and thus easily adaptable to different types of crops and (3) forcing a sugarcane crop growth model with data derived from remote sensing. The MOSICAS sugarcane dedicated crop model, which is adapted to the cropping conditions of Reunion Island, was used. Our tests were made on sixty three fields located on two contrasted in-farm sites, and on seven plots located on an experimental site. Our dataset was composed of remote sensing data (SPOT4 & 5 images and thermal infrared data), yield data, climatic data, soil data and cropping practices data (irrigation schedules and harvest dates). Concerning the methodological approach, obtained results showed that remote sensing data, through a better inclusion of the actual state of development of the crop or an optimized parameterization of the model, results in a significant enhancement of the estimation of the yield by the MOSICAS model. In particular, we showed that forcing the model resulted in a gain of accuracy of 2.6 t ha-1. We also recalibrated the radiation use efficiency parameter for each studied cultivar. Finally, we determined an optimized value of the rooting depth parameter using recalibration and the water stress index CWSI as an adjustment variable. Concerning the application approach, our results also showed that the more complex methods of yield estimation do not provide the best results when considering the precision. We therefore recommend using the simple empirical relationship between yield and vegetation indices for the estimation of the sugarcane biomass on Reunion Island. These results offer several prospects: firstly, a better inclusion of the heterogeneity of cultivars used on Reunion Island by recalibrating the key parameters of the yield computation for each of these cultivars in order to test various scenarios of cultivar implantation as a function of climatic zones of the island. The estimation method selected here should also be exported to other sugarcane smallholder countries, particularly with introduction of the Sentinel-2 system to provide open access and high spatial resolution images.
Document type :
Theses
Complete list of metadatas

Cited literature [111 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01155273
Contributor : Abes Star <>
Submitted on : Tuesday, May 26, 2015 - 1:52:05 PM
Last modification on : Friday, May 17, 2019 - 1:19:45 AM
Long-term archiving on : Tuesday, September 15, 2015 - 6:53:51 AM

File

2014lare0021_jmorel.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01155273, version 1

Citation

Julien Morel. Estimation de la biomasse de canne par modélisation et télédétection. Application à la Réunion. Océanographie. Université de la Réunion, 2014. Français. ⟨NNT : 2014LARE0021⟩. ⟨tel-01155273⟩

Share

Metrics

Record views

2189

Files downloads

2068