كليدواژه :
HANTS , دامنه , سري زماني , فنولوژي , NDVI
چكيده فارسي :
بررسي تغييرات پوشش هايگياهي مي تواند اطلاعات ارزشمندي را در مورد گرمايش جهاني،چرخه كربن، چرخه آب و تبادل انرژي به همراه داشته باشد. استفاده از سري هاي زماني تصاوير ماهواره اي و روش هاي سنجش از دور اطلاعات زيادي را در مورد تغييرات و پويايي هاي پوشش هاي گياهي به ما عرضه مي دارند. هدف از پژوهش حاضر، تعيين تغييرات هر كدام از مولفه هاي سري هاي فوريه پوشش هاي گياهي ايران در طول سه دهه گذشته مي باشد. بدين منظور در اين مطالعه ازمحصول NDVI روزانه سنجنده AVHRRبا قدرت تفكيك مكاني 0/05 در 0/05 درجه با نام AVH13C1 استفاده شد. سپس با استفاده از الگوريتم HANTS اجزاي هارمونيك چهار سري زماني يك ساله در زمان گذشته (1982، 1983، 1984 و 1985) و چهار سري زماني يك ساله در سال هاي اخير (2015، 2016، 2017 و 2018) توليد شد. در نهايت تغييرات اجزاي هارمونيك يا همان تصاوير دامنه و فاز در سال هاي اخير نسبت به سال هاي گذشته تعيين شد و اختلاف ميانگين ارزش هاي اجزاي هارمونيك بين چهار سري زماني يك ساله در گذشته وحال با تجزيه واريانس يك طرفه بررسي شد و نقشه هاي معني داري اختلاف بين ميانگين ها بدست آمد. با توجه به نتايج،در مناطق مركزي، شرق و شمال شرق ايران دامنه صفر (ميانگين پوشش گياهي) در سطح احتمال 95 درصد (F-value< 0/05) كاهش يافته و در مناطق شمال و شمال غرب و غرب به ويژه ارتفاعات البرز و زاگرس دامنه صفر به طور معني دار (F-value< 0/05) افزايش يافته است. اختلاف ميانگين ارزش فازها در چهار سري زماني در گذشته و سال هاي اخير در مناطق غرب و شمال غرب و همچنين شرق و شمال شرق ايران در سطح احتمال 95 درصد (F-value< 0/05) معني دار مي باشد. فازهاي سالانه اين مناطق به ميزان 14 درجه كاهش يافته است كه اين موضوع نشان دهنده شروع زودتر فرآيندهاي رشد و فنولوژي گياهان اين مناطق نسبت به سه دهه گذشته مي باشد.
چكيده لاتين :
Investigation of vegetation changes can provide valuable information on global warming, the carbon cycle,water cycle and energy exchange. Satellite imagery timeseriesandremote sensing techniques offers a great deal of information on variations and dynamics of vegetation. Harmonic ANalysis of Time Series (HANTS) has been effectively used to eliminate missing and outliers in time series of vegetation indices and land surface temperature (LST). However, the algorithm has been less frequently used to detect changes in vegetation and phenology. HANTSalgorithm decomposes periodic phenomena into their components(different sines and cosineswith different amplitudes and phases). The value of phases and amplitudes contains valuable information that can be used to investigate variations and identify different characteristics of vegetation such as growth and phenology. The present study aims to determine changes in each componentof vegetation time series in Iranin the past (1982, 1983, 1984 and 1985) and in recent years (2015, 2016, 2017 and 2018).
Materials & Methods
A daily NDVI product of AVHRR sensor, with a resolution of 0.05 at 0.05 ° (i.e. AVH13C1) was used in the present study. To obtain reliable harmonic components (amplitude and phase images), a reliable curve has to be fitted on the primary time series data. To do so, first,parameters of HANTS algorithm were determined and then Root Mean Square Error (RMSE) of the curves fitted on data related to four one-year time series in the past year’s category (1982, 1983, 1984 and 1985) and four one-year time series in recent year’s category (2015, 2016, 2017 and 2018) was estimated. This classification (i.e. four one-year time series in the past and recent years) was used for two reasons. First, extraction and comparison of harmonic components in a single time series in the past and recentyears’ categories cannot reflect real changes, as these changes may occur under the influence ofimpermanent dynamics of vegetation, such as dryor wet periods. Second, with four one-year time series in the past category (1982, 1983, 1984 and 1985), and four one-year time series (2015, 2016, 2017 and 2018) in recent years, statistical comparison of the harmonic components through one-way analysis of variance becomes possible. Following the production of reliable harmonic components, variations of the harmonic components in recent years were compared with their variations in the past using difference method, and mean difference of the harmonic components’value in four one-year time seriesin the past and present categories wasdetermined using one-way analysis of variance. Finally, some maps were produced to exhibitthe significance of differenceinmeans.
Results & Discussion
According to the findings of the present study, mean RMSE of the fitted curves in the four one-year periods ofpresent and past time series were always less than 0.1 unit of NDVI. Moreover, mean RMSEof total area of Iranin the past and present time series were 0.037 and 0.039, respectively. This demonstrates high efficiency of the HANTS algorithm in elimination of missing and outlier data in the daily-NDVI time series ofNOAA-AVHRR. Results indicate thatrange of zero amplitude (the mean value of NDVI or the average vegetation coverage) decreasesin the central, eastern and northeastern regions of Iran atthe 95% probability level (F-value <0.05), whileit increases significantly (F-value <0.05)in the north, northwestern and western regions (especially, the Alborz and Zagros mountains). The meandifferenceof phases value in the four-time series of the past and recent years’categories wassignificant at the 95% probability level (F-value <0.05). Compared to the past time series, first harmonic phase average of total area of Iran in the new time series has decreased by almost 14 degrees. This decrease in the value of the annual and 6-month phases indicates a quicker growth phase and phenological processes of plants compared to past times.
Conclusion
Results indicated that HANTS algorithm can effectively eliminateand reconstruct outliers in the NDVI time series. Zero harmonic (mean value) represents the overall level of vegetation cover and the firstharmonic phase in a one-year time series determines the starting time of growth in seasonal plants or thosewith agrowth period of6-month or less. Annual Phase indicates the angular starting position of the annual cycles and the 6-month phase inherently indicates the fluctuation and angular position of a half-year or 6-month curve. However, interpreting 6-month amplitude and phases are difficult. As most changes are controlled by the first harmonic phase, the first harmonic phase in a one-year time series contains important information about the beginning of growth and the phenological processes of plants. Therefore, harmonic components of a periodic time series canbeusedto identify and determine changes in vegetation coverage and phenological processes.
