Remotely sensed soil moisture to estimate savannah NDVI.

Niklas Boke-Olén,Jonas Ardö,Thomas Holst,Lars Eklundh,Veiko Lehsten,Ben Bond-Lamberty

doi:10.1371/journal.pone.0200328

Abstract

Satellite derived normalized difference vegetation index (NDVI) is a common data source for monitoring regional and global ecosystem properties. In dry lands it has contributed to estimation of inter-annual and seasonal vegetation dynamics and phenology. However, due to the spectral properties of NDVI it can be affected by clouds which can introduce missing data in the time series. Remotely sensed soil moisture has in contrast to NDVI the benefit of being unaffected by clouds due to the measurements being made in the microwave domain. There is therefore a potential in combining the remotely sensed NDVI with remotely sensed soil moisture to enhance the quality and estimate the missing data. We present a step towards the usage of remotely sensed soil moisture for estimation of savannah NDVI. This was done by evaluating the European Space Agency (ESA) Climate Change Initiative (CCI) soil moisture and three of its individual products with respect to their relative performance. The individual products are from the advance scatterometer (ASCAT), Soil Moisture and Ocean Salinity (SMOS), and the Land Parameter Retrieval Model-Advanced Microwave Scanning Radiometer-Earth Observing System (LPRM-AMSR-E). Each dataset was used to simulate NDVI, which was subsequently compared to remotely sensed NDVI from MODIS. Differences in their ability to estimate NDVI indicated that, on average, CCI soil moisture differs from its individual products by showing a higher average correlation with measured NDVI. Overall NDVI modelled from CCI soil moisture gave an average correlation of 0.81 to remotely sensed NDVI which indicates its potential to be used to estimate seasonal variations in savannah NDVI. Our result shows promise for further development in using CCI soil moisture to estimate NDVI. The modelled NDVI can potentially be used together with other remotely sensed vegetation datasets to enhance the phenological information that can be acquired, thereby, improving the estimates of savannah vegetation phenology.

Highlights

MethodsThe main soil moisture dataset included in this study was the multi-decadal merged soil moisture product being part of the Climate Change Initiative (CCI) provided by the European Space Agency
normalized difference vegetation index (NDVI) modelled with remotely sensed soil moisture products were evaluated against remotely sensed NDVI from the Moderate Resolution Imaging Spectroradiometer (MODIS, [28])
MODerate Resolution Imaging Spectrometer (MODIS)-NDVI data flagged as marginal, snow or ice, or cloudy was filtered out, and the dataset was spatially averaged to the common 0.25 degree resolution used throughout the study and filtered with a Savitsky-Golay filter using the same approach as Boke-Olen et al [21] used

Summary

Methods

The main soil moisture dataset included in this study was the multi-decadal merged soil moisture product being part of the Climate Change Initiative (CCI) provided by the European Space Agency. It combines data from four active and two passive sensors to create a merged dataset [11], on referred to as CCI (Table 1). The soil moisture datasets were temporally gap filled with a linear interpolation to create daily data. They were further smoothed with an 8-day median filter and matched to the 8-day time steps of MODIS-NDVI

Results

Discussion

Conclusion