Generating High Spatio-Temporal Resolution Fractional Vegetation Cover by Fusing GF-1 WFV and MODIS Data

Tao Tao,Yao Yao,Zhang Zhang,Wang Wang,Zhao Zhao,Jia Jia,Wei Wei,Xie Xie

doi:10.3390/rs11192324

Abstract

As an important indicator to characterize the surface vegetation, fractional vegetation cover (FVC) with high spatio-temporal resolution is essential for earth surface process simulation. However, due to technical limitations and the influence of weather, it is difficult to generate temporally continuous FVC with high spatio-temporal resolution based on a single remote-sensing data source. Therefore, the objective of this study is to explore the feasibility of generating high spatio-temporal resolution FVC based on the fusion of GaoFen-1 Wide Field View (GF-1 WFV) data and Moderate-resolution Imaging Spectroradiometer (MODIS) data. Two fusion strategies were employed to identify a suitable fusion method: (i) fusing reflectance data from GF-1 WFV and MODIS firstly and then estimating FVC from the reflectance fusion result (strategy FC, Fusion_then_FVC). (ii) fusing the FVC estimated from GF-1 WFV and MODIS reflectance data directly (strategy CF, FVC_then_Fusion). The FVC generated using strategies FC and CF were evaluated based on FVC estimated from the real GF-1 WFV data and the field survey FVC, respectively. The results indicated that strategy CF achieved higher accuracies with less computational cost than those of strategy FC both in the comparisons with FVC estimated from the real GF-1 WFV (CF:R2 = 0.9580, RMSE = 0.0576; FC: R2 = 0.9345, RMSE = 0.0719) and the field survey FVC data (CF: R2 = 0.8138, RMSE = 0.0985; FC: R2 = 0.7173, RMSE = 0.1214). Strategy CF preserved spatial details more accurately than strategy FC and had a lower probability of generating abnormal values. It could be concluded that fusing GF-1 WFV and MODIS data for generating high spatio-temporal resolution FVC with good quality was feasible, and strategy CF was more suitable for generating FVC given its advantages in estimation accuracy and computational efficiency.

Highlights

Vegetation is a major component of terrestrial ecosystems, playing a key role in the exchanges of carbon, water and energy [1]
Five digital images were acquired using a digital camera from the nadir, and the average of Fractional vegetation cover (FVC) extracted from these five digital images was considered as the field survey FVC at this sample point
From the validation based on the simulated data (Figure 5), the GF-1 WFV FVC estimation models for the fourFrWomFVthseenvsaolirdsastihoonwbassaetdisfoancttohrey saicmcuulraatecdy (dRa2ta≥(F0i.g9u66re, R5)M, tShEe ≤GF0-.105W)

Summary

Introduction

Vegetation is a major component of terrestrial ecosystems, playing a key role in the exchanges of carbon, water and energy [1]. FVC, especially with high spatio-temporal resolution, is widely used in land surface process models such as land desertification evaluation, agriculture monitoring, soil erosion monitoring and hydrological simulation [4,5,6,7,8,9]. FVC estimation methods based on remote sensing data can be divided into three categories: empirical methods [11,12,13], pixel un-mixing methods [4,14] and physical methods [11,15,16]. Physical methods simulate the physical relationship between vegetation canopy reflectance and FVC based on the radiative transfer models. Physical methods have clearer physical significance and are more suitable for FVC estimation in large-scale regions with various vegetation conditions. The inversion process is typically simplified by a lookup table or machine learning methods

Objectives

Methods

Results

Conclusion