Limitations of cloud cover for optical remote sensing of agricultural areas across South America

Abstract

Abstract: Earth Observation data support the large-scale monitoring of croplands across South America (SA). For instance, optical remote sensing (ORS) data are typically used to map and monitor plant development during a crop season. However, the monitoring of agricultural areas is highly affected by cloud cover frequency (CCF), especially in the rainy season, and the implications of cloud cover for ORS of agricultural areas are still poorly understood in SA. In this study we evaluated the monthly CCF and variability focused on implications for agricultural monitoring in SA. Cloud cover was derived from daily MCD19A2 Collection 6 Moderate Resolution Imaging Spectroradiometer (MODIS) product between 2000 and 2015. A monthly average was computed using daily observations for the studied period. To evaluate the effects of clouds in agricultural areas, we used a cropland mask from the Worldwide Croplands project and divided the CCF into quarters. The results show that cloud cover affects the monitoring of croplands depending on geographic location and crop season. In the P1 period (September to November) 68% of South American croplands have CCF between 40% and 60%. In the P2 period (December and February), SA croplands have CCF concentrated in the classes of 40%–50% and 70%–80%. In the P3 period (March to May) 42% of SA croplands have CCF concentration in class 5 (40-50% of cloud cover). In the P4 period (June to August), we observed values from 30% to 60% of cloud cover in 41% of South American croplands. These patterns make summer crop monitoring via ORS data difficult, mainly soybean and maize. In this sense, for the Brazilian states of Mato Grosso, Goias, Bahia, Tocantins, and Parana, the use of ORS is limited in providing an accurate summer crop monitoring. While cloud cover is an intrinsic challenge for agricultural monitoring of some crops, the combination of multi-sensors (e.g. microwave and optical sensors) and CubeSats can improve the earth observation frequency and help to work around this limitation, thus enabling a better time series analysis.