Remotely sensed green area index for winter wheat crop monitoring: 10-Year assessment at regional scale over a fragmented landscape

Despite remarkable technological advances in earth observation systems and radiative transfer modelling, enabling the retrieval of canopy biophysical variables from satellite imagery, the use of remote sensing for operational crop monitoring at a regional or global scale has remained more qualitative than quantitative. One of the main reasons lies in the fact the imagery that can be used operationally and economically over large areas with high temporal frequency have coarse spatial resolution. However, recent research has demonstrated that coherent crop specific biophysical variables such as green area index (GAI) can be retrieved from medium spatial resolution imagery such as MODIS even when the size of the fields is close to the size of the pixels (close to 250 m). Leveraging on these results, the present paper attempts to go beyond by retrieving GAI from a more fragmented landscape, over a much larger geographical area and covering a 10-year period. Results demonstrate the possibility to monitor the dynamic processes of growth and senescence of winter wheat and grasp the inter-annual seasonal variability of growing conditions encountered over a decade. Furthermore, the satellite-derived GAI is not only consistent with ground measurements at regional scale (RMSE = 0.65 m2/m2, RRMSE = 25.7%), but even shows encouraging results at field level (RMSE = 0.82 m2/m2, RRMSE = 37.6%, when pixel/field spatial adequacy is high). By showing the possibility of monitoring crop specific growth quantitatively over a complicated landscape, the major step necessary before implementing such approach in operational situation remains identifying early in the season where the target crop is.

► Crop specific GAI can be retrieved from remote sensing time series at regional scale. ► Considering thermal time allows enhanced smoothing of crop specific GAI time series. ► Satellite GAI grasps the inter-annual variability encountered within 10 years. ► RMSE between satellite GAI and ground measurements is 0.65 m2/m2 at regional level. ► When pixel-target adequacy is high, RMSE can reach 0.82 m2/m2 at field level