Evaluating the Impact of Smoothing Methods on Phenology Calculation in Zagros Forests using MODIS Satellite Data (Case Study: Marivan)

Climate change refers to long-term alterations in the climatic factors of a region, significantly impacting its ecological processes. The investigation of plant phenological changes is crucial for comprehending the behavioral responses of Earth ecosystems to climate change. Remote sensing methods facilitate the examination of phenological changes over extensive areas. This research aims to explore such changes using time series data obtained from satellite images in a part of the Zagros forests.

The most common method in remote sensing for calculating plant phenology parameters involves using time series of vegetation indices, which are usually disrupted by cloud cover and affected by noise. These issues can be somewhat mitigated by employing smoothing filters although the characteristics of the filters need to be carefully chosen. To extract phenological parameters, time series data of the normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) from MODIS data, over a ten-year period (From 2009 to 2018), were used. For the extraction of phenological parameters, including the start of the growing season and end of the growing season, the time series of NDVIratio and EVIratio were smoothed using two smoothing functions: Savitzky–Golay and Moving Average filters. For the validation of the results, in-situ data were collected in the area through direct observation of tree status (via random sampling).

The use of the Moving Average function with a window size of six and the Savitzky–Golay function with a window size of 11 and a degree of two has led to the highest increase in accuracy for both vegetation indices. After employing the optimal smoothing function, the parameter for the start of the season, obtained from EVIratio, provided a more accurate estimate (R2 = 0.92) compared to the result from NDVIratio. Over the ten year period, the start of the season has been reduced by an average of 0.5 days per year over the ten-year period, and end of the season has been increased by an average of 1.3 days per year.

 Overall, in most areas, the start and end of the growing season has shown a decreasing and an increasing trend, respectively. In other words, the trend of earlier start and later end of the season has led to an increase in the length of the growing season in most of the study areas.