Relationship between Plant Indices of Sentinel 2A Image with Percentage of Cover and Production in Rangeland Habitats of Mountainous Region Adjacent to Choghakhor Wetland, Chaharmahal and Bakhtiari

This study aimed to compare the spectral resolution of Sentinel-2A imagery at 10 and 60 m and ground quadrat sizes of 1×1, 1×2, 2×2, and 3×3 m for estimating vegetation production and cover in mountainous rangeland habitats surrounding Choghakhor Wetland, Chaharmahal and Bakhtiari Province, Iran. Field sampling was conducted in 30 units measuring 30×30 m along three transects using two sampling methods involving three and six quadrats. Different quadrat sizes were arranged in grids to estimate vegetation cover and production. The density of dominant species was determined by counting bases, vegetation cover was estimated using a gridded frame, and production was measured using the double-sampling method based on its relationship with canopy cover. The distribution patterns of dominant species were determined using statistical tests. After correcting the Sentinel-2A imagery and calculating vegetation indices, the relationships between the indices and vegetation cover and production were evaluated. The results showed that, at the 10 m resolution, NDVI, CTVI, MSAVI2, Ratio, RVI, SAVI, and TVI were significantly related to vegetation cover and production. The 1×1 m quadrat showed insufficient correlation, whereas the six-quadrat sampling method produced stronger relationships and more reliable models than the three-quadrat method for the other quadrat sizes. At the 60 m resolution, only MSAVI2 and RVI showed significant relationships with vegetation cover and production, and the correlations under both sampling methods were weaker than those at 10 m resolution. Overall, the use of 10 m resolution bands, 1×2 m quadrats, and six-quadrat sampling is recommended as a suitable combination for estimating rangeland vegetation cover and production. This combination improves estimation accuracy and is feasible for field application.
Extended Abstract
 
Introduction
Recognition and evaluation of rangeland ecosystems constitute the first and most essential steps in the sustainable management of these valuable resources. Effective management requires comprehensive and accurate knowledge of rangeland conditions, while such information should ideally be obtained with minimum time and cost. However, the vast spatial extent of rangelands, particularly in Iran, poses a major challenge to direct field-based assessment. The considerable area of the country’s rangelands, combined with limited financial and human resources, makes comprehensive field evaluation difficult, even at a relatively coarse scale. Therefore, the development and application of rapid, accurate, and cost-effective methods for rangeland assessment are of considerable importance. Remote sensing and satellite data provide an effective approach for addressing these challenges by enabling the assessment of extensive areas at relatively low cost. The ability to provide a broad and synoptic view of vegetation communities, repeated observations, consistent spatial coverage, and timely information are among the major advantages of satellite-based data for vegetation assessment. Accordingly, numerous studies have employed remote sensing techniques to evaluate vegetation cover and rangeland characteristics, demonstrating their considerable potential for such applications. Against this background, the present study was conducted under the semiarid conditions of Iran to determine appropriate models for different sampling patterns and dimensions, evaluate the effectiveness of Sentinel-2A satellite data at spatial resolutions of 10 and 60 m, and identify suitable vegetation indices for estimating vegetation cover and aboveground production in rangelands dominated by broadleaf plant species.
 
Methodology
The study area is located 40 km southwest of Shahrekord, near Choghakhor Wetland in Chaharmahal and Bakhtiari Province, Iran. The dominant broadleaf plant community was identified through surveys and patrols. Within this community, 30 sampling units measuring 30 × 30 m were established along three 900-m transects, with 60 m between adjacent units. Ground sampling used quadrats of four sizes (1 × 1, 1 × 2, 2 × 2, and 3 × 3 m) to assess vegetation cover and production. Dominant species density was determined by counting individuals within 2 × 2 m quadrats, while vegetation cover was estimated and aboveground production was measured using a double-sampling method based on cover percentage. Statistical tests were applied to determine spatial distribution patterns of dominant species. Sentinel-2A imagery at 10- and 60-m resolutions was preprocessed and corrected. Vegetation indices were calculated, and their relationships with field-measured vegetation cover and production were evaluated. The study compared satellite and ground sampling scales to identify suitable spatial resolutions and sampling dimensions for rangeland vegetation assessment.
 
Results and Discussion
The results indicated that most relationships between the NDVI, CTVI, MSAVI2, Ratio, RVI, SAVI, and TVI vegetation indices and vegetation cover and aboveground production were statistically significant, demonstrating the potential of these indices for assessing broadleaf-dominated rangelands. However, correlation strength and model performance varied according to sampling method, quadrat size, and Sentinel-2A spatial resolution. The six-quadrat sampling method generally produced stronger correlations and more reliable models than the three-quadrat method. At the 10-m spatial resolution, this difference was mainly attributed to better spatial correspondence between field quadrats and the 10 × 10 m satellite pixels. In contrast, the three-quadrat design provided less representative ground information within each pixel, reducing correlation strength and model reliability. Quadrat size also had a substantial influence on the results. The 1 × 1 m quadrat generally yielded weak and mostly nonsignificant relationships with both vegetation cover and production, accompanied by relatively high RMSE values and poor model validity. Therefore, this quadrat size was considered unsuitable for representing the spatial characteristics of the studied plant community. Increasing quadrat dimensions generally improved the relationships between field measurements and vegetation indices, although differences among larger quadrats were not always statistically significant. A similar pattern was observed for aboveground production, with stronger correlations and more reliable regression models obtained using six quadrats. For the 60-m bands, increasing quadrat size generally improved correlation strength and significance; nevertheless, the relationships remained weaker than those obtained from the 10-m bands. This reduction was mainly associated with the larger 60 × 60 m pixel size, which increased spatial heterogeneity and weakened the correspondence between field observations and satellite-derived indices. The effect was particularly evident in the three-quadrat sampling method. Overall, the findings demonstrate that sampling configuration, quadrat size, and satellite spatial resolution are critical factors in rangeland vegetation assessment, with the six-quadrat method and appropriately sized quadrats providing more reliable estimates of vegetation cover and aboveground production.
 
Conclusion
By examining the validity of the models, it can be seen that in most of the indices, the models obtained in the 1×1 m quadrats in both methods and in the 1×2 quadrats in the  3 quadrats sampling method are not sufficiently valid. Because the difference between the grounds harvested factor and the result of the model is significant and their RMSE is significantly high. For this reason, the 1×1 quadrat was not suitable for the 10 m Sentinel bands. It was the same way. In the 1×2 quadrat and 3 square sampling methods, the validity of this image (60 m separation) is reduced and is not statistically valid. Therefore, considering this issue, when using 60 m separation bands, it is not recommended to use 1×2 quadrat  and 2 ×2 quadrat with 6 sampling units can be  used.  The correlation and significance of regression relationships in most indices is low compared to the coverage percentage. So that the models obtained from sampling method 1 in quadrats 1×2 and 2×2 have sufficient validity. In the use of bands with a resolution of 60 m, the relationships obtained in both sampling methods (3 and 6 plates) have a weaker correlation than bands with a resolution of 10 m. In general,  it is suggested that bands with a separation of 10 m and a quadrate of 1×2 with a 6 square sampling method should be used to obtain acceptable results and be feasible in terms of sampling . The procedure of increasing and decreasing the correlation and validity of the models is the same as the coverage percentage. Due to the high cost of sampling in the field and other research operations, providing financial resources will help advance the goals of such studies.