© 2014 Elsevier B.V. Extensive stock-breeding systems developed in mountain areas like the Pyrenees are crucial for local farming economies and depend largely on above-ground biomass (AGB) in the form of grass produced on meadows and pastureland. In this study, a multiple linear regression analysis technique based on in-situ biomass collection and vegetation and wetness indices derived from Landsat-5 TM data is successfully applied in a mountainous Pyrenees area to model AGB. Temporal thoroughness of the data is ensured by using a large series of images. Results of on-site AGB collection show the importance for AGB models to capture the high interannual and intraseasonal variability that results from both meteorological conditions and farming practices. AGB models yield best results at midsummer and end of summer before mowing operations by farmers, with a mean R2, RMSE and PE for 2008 and 2009 midsummer of 0.76, 95 g m−2 and 27%, respectively; and with a mean R2, RMSE and PE for 2008 and 2009 end of summer of 0.74, 128 g m−2 and 36%, respectively. Although vegetation indices are a priori more related with biomass production, wetness indices play an important role in modeling AGB, being statistically selected more frequently (more than 50%) than other traditional vegetation indexes (around 27%) such as NDVI. This suggests that middle infrared bands are crucial descriptors of AGB. The methodology applied in this work compares favorably with other works in the literature, yielding better results than those works in mountain areas, owing to the ability of the proposed methodology to capture natural and anthropogenic variations in AGB which are the key to increasing AGB modeling accuracy.
|Journal||International Journal of Applied Earth Observation and Geoinformation|
|Publication status||Published - 1 Jun 2015|
- Aboveground biomass modelling
- Landsat imagery
- Mountain stock-breeding
- Multiple regression techniques
- Vegetation and wetness indices