Discriminação de culturas agrícolas utilizando análise harmônica de séries temporais de dados EVI-MODIS / Crop discrimination using harmonic analysis of EVI MODIS time-series data

AUTOR(ES)
DATA DE PUBLICAÇÃO

2008

RESUMO

In the present decade, agriculture has increased worldwide in importance due to sensitive matters such as the current food crisis, biofuels boom, and the land use and land cover changes from savanna/forest areas to agriculture expansion. As a consequence, crop monitoring at regional and national scales has become essential at different levels of knowledge and for diverse discipline areas. This study primarily aims to perform crop discrimination using two sources of remote sensing data in the period from July/2006 to June/2007: a) harmonic terms derived from harmonic analysis of time-series data of MODIS-EVI 16-day composite imagery, in addition with b) medium spatial-resolution data (TM-Landsat and CCD-CBERS). Four municipalities from Mato Grosso State were selected as study area. This region is a large agriculture producer, especially of soybean, cotton and maize, and it is also known as one of the largest agriculture frontiers in the world. These annual crops have a short cycle, which makes crop monitoring hard to achieve only by using medium spatial resolution imagery because there is a coincidence with a period of high cloud cover, particularly during the summer season. We found that crop calendar and management practices, such as succession planting, had a strong relation with the capability of associating crop types to harmonic terms. Annual crop cycles were modeled by the first 3-order harmonic terms while other land cover units exhibited lower values in same harmonic terms. The time-range selection was a sensitive parameter that allowed a better discrimination among crops and their planting patterns. Classifications of MODIS harmonic components (amplitude and phase terms) reached accuracies around of 90%, which outperformed classifications of MODIS reflectance bands (single-date image) during the highest crop development. Statistical comparisons of individual harmonic terms allowed checking the separability between major annual crops due to succession planting: single versus double cropping. A simple GIS integration was used to mix final MODIS and TM data products to generate the final map. Estimates of crop area were obtained and compared with official IBGE (subjective) data. Both data were quite similar in proportion terms but in absolute values our results showed that soybean estimates were lower and cotton were higher than the official data. Finally, we propose a crop-masking procedure using the sum of high-orders amplitude terms derived from EVI-MODIS to setup gross crop area estimates.

ASSUNTO(S)

management practices imagens de resolução espacial média evi modis culturas anuais calendário agrícola análise harmônica evi modis crop calendar annual crops harmonic analysis práticas de manejo medium spatial resolution imagery

ACESSO AO ARTIGO

Documentos Relacionados