ارزیابی روش های هموارسازی برای بازسازی سری زمانیNDVI و برآورد فنولوژی از داده های ماهواره لندست8

 
Extended Abstract
Introduction
Many methods have been developed to identify phenological events based on remote sensing data. Most methods for detecting phenological events involve two basic steps; (1) Generate time series from satellite data (2) Use time series to determine phenological events based on relational sets. The first step involves building the time series of each indicator based on remote sensing data and smoothing the data to reduce noise and produce a smoother time series. But this time series is affected by clouds, humidity, and weather, disrupting the signals received by satellite sensors. Many time series reconstruction methods have been used to reconstruct disturbed satellite signals. Recently, widely used methods such as Savitsky-Goli filter (S-G), least squares for Gaussian asymmetric functions (AG), and Double logistics (DL) functions have been used. In previous studies, many researchers have concluded that the performance of smoothing methods in estimating plant phenology, spatially and temporally, due to bias and random errors due to clouds and plant type and physical conditions of the environment, no single method of performance Does not display superior. The purpose of this study is to evaluate three time series smoothing methods, using statistical criteria and phenological parameters extracted from the NDVI Index obtained from Landsat 8 satellite images.
 
Methodology
In this study, phenological parameters of the start of season (SOS) and end of season (EOS) for rapeseed vegetation growing season in Chaharmahal and Bakhtiari province, Farokhshhar region obtained from observational data and NDVI index of Landsat 8 satellite images in the period 2018-2016 were used. To smooth the data and extract the phenological parameters of the beginning of the season and the end of the season of satellite images, Savitsky-Goli filter, least squares for Gaussian asymmetric functions (AG) and double logistic functions (DL) in TIMESAT software were used. In all methods, the adaptation to upper envelope with the raw NDVI time series was used to reduce bias. In the Savitsky-Goli method, in addition to adapting upper envelope, the window size parameter (r) was also used. Pearson correlation coefficient and root mean square error (RMSE) were used to compare the output of time series of smoothing functions and raw time series of NDVI.
 
Results and Discussion
Statistical evaluation of smoothed time series
Statistical analysis of the output of smoothing functions showed that the time series produced by the S-G model compared to the raw time series of the NDVI index had the lowest root mean square error (RMSE = 0.342) and the highest correlation (r = 0.98) belong to S-G model. The advantage of DL and AG models is that the difference between the mean correlation coefficient for all performances and the correlation coefficient for the best execution is small and it can be inferred that the software parameter settings have little effect on the outputs of these models. After plotting the smoothed time series curves, the results showed that the use of smoothing models effectively eliminated noise and disturbed the raw time series of the NDVI index, and reconstructed smoother and softer time series. The results also showed that time series that have a higher correlation coefficient show more details and changes within the inter-seasonal, such as the recession stage (dormancy).
Evaluation of smoothed time series in estimating phenological parameters
The results showed that in estimating the start and end of the season (SOS / EOS), the output of DL model is more accurate than the output of S-G and AG models. Compared to observational data, the output of all models has a time delay in estimating the EOS. Overall, the DL model performed better in estimating the SOS and EOS phenology parameter with 1 and 9 day difference with observational data respectively.
In this study, we showed to what extent the time series of the three smoothing methods SG, AG and DL in the reconstruction of the raw time series of the NDVI from the Landsat 8 and estimating the phenological parameters of the start and end of the season are accurate. The results of this study showed that the adaptive S-G model is more robust for reconstructing raw time series than AG and DL functions, and this is due to the sensitivity of this model to small changes in the NDVI time series. The AG and DL functions tend to eliminate noise at the peaks and bottoms of the time series. The results also showed that the time series with the highest correlation coefficient (r) are more suitable for reconstructing the raw time series of the NDVI index compared to the time series that produced the smallest RMSE. The DL model performed better in estimating the SOS and EOS phenology parameter. In SOS estimation, the S-G model performs worse than the AG and DL functions. The efficiency of any smoothing method depends on the choice of parameters. For example, the use of adaptation upper envelope generally improves the results. AG and DL fitting function methods are the preferred option for smoothing low quality data (eg high noise and high data loss) due to less sensitivity to regulatory parameters. The AG and DL fitting functions are limited when giving inter-seasonal details of the time series curve. Numerous factors such as vegetation index selection, satellite sensor data and vegetation type are affected in evaluating time series and estimating phenological parameters. However, the results of this study are valid for the data and the location under study, and the results may vary with other data or under other circumstances.
 
Conclusions
This study showed that the statistical criterion of Pearson correlation coefficient (r) is superior to the root mean square error (RMSE) and the S-G model is superior to the AG and DL models for reconstruction of time series and the DL function show the best performance for estimating SOS and EOS phenological parameters.