Cruzado Campos, Enric
Loading...
Organizational Units
Organizational Unit
Job Title
ORCID
Panorama page
person.page.panorama
Name
Email Address
1 results
Search Results
Now showing 1 - 1 of 1
- PublicationModelización del posprocesado de los mapas de rendimiento de las cosechadoras en el cultivo de trigo y cebada en España(Universitat Politècnica de València, 2024-09-03) Cruzado Campos, Enric; Castiñeira Ibáñez, Sergio; Arizo GarcÃa, Patricia; Departamento de Producción Vegetal; Departamento de FÃsica Aplicada; Escuela Técnica Superior de IngenierÃa de Telecomunicación; Centro Valenciano de Estudios sobre el Riego; Centro de TecnologÃas FÃsicas: Acústica, Materiales y AstrofÃsica; Escuela Técnica Superior de IngenierÃa Agronómica y del Medio Natural[EN] In view of the forecast of an exponential increase in population in the coming decades, the application of new technologies to optimize agronomic management has become vitally important. In particular, the increase in the production of cereals such as wheat and barley, being the first and fourth most produced cereals worldwide, has become key. In this context, the concept of precision agriculture emerges, one of the outstanding disciplines is remote sensing for modeling the productive response of crops. However, this tool depends on the availability of accurate data acquired by combine harvesters, so it is necessary to study the veracity of these data for subsequent crop modeling. In this work, the suitability of different yield data processing systems has been evaluated in terms of their standard deviation, based on the yield data recorded at intra-plot level, with the aim of obtaining an automatic filtering of wheat and barley yield maps. The area studied, spread over the 2021-2021 and 2021-2022 seasons, amounts to 296 ha (100 fields) for wheat, and 289 ha (98 fields) for barley, all in the province of Burgos. Different types of post-processing have been applied, using the following formula; m±n-SD, where m is the mean and SD is the standard deviation, both in a global adjustment (field level) and in a local adjustment (40x40 m² level). Three values of n (1, 1.5 and 2.5) have been tested, resulting in thirteen different post-processing types. Knowing that such data are intended to be used in performance modeling, Sentinel-2 data (10 and 20 m) were used to study how the coefficient of determination (r2) between the final performance after filtering and the reflectance bands was altered and to discern whether such variations would make a large difference in the reliability of the models. Finally, a mean filter was applied to the proposed postprocessing. G1L1.5F1 was the best post-processed, as only 8% of the 10x10 m pixels were removed and the standard deviation was considerably reduced from 1.438.21 to 892.17 kg-ha-1. Likewise, the geostatistical parameters analyzed (coefficient of variation, range and semivariance) obtained optimal results. Also, the difference between the proposed post-processing (G1L1.5F1) and step 1 (s1) was calculated, it was obtained that 78% of the barley fields and 72% of the wheat fields present a difference below ±100 kg-ha-1. Finally, it was decided to explore the relationship between the coefficient of variation and yield. By breaking down the yield by 500 kg-ha-1 intervals and calculating the coefficient of variation according to post-processing. It was appreciated that all data below 500 kg-ha-1 could be disregarded, as they represent the highest variability with very low percentage of data. Also, it was observed that from the range 2.000 – 2.500 kg-ha-1 the coefficient of variation is not affected in both crops, meaning that post-processing would not be necessary, achieving a great saving in computational power.