Modelling Above ground net primary production of Sabalan rangelands using vegetation index and non-linear regression

Background and objectives: Alongside with the progresss in satellite sensors, their role on obtaining information and study of the environmental events and phenomena have become increasingly important. One of these precious data which is the basis for many planning and decision making in rangelands, is the above ground net primary production assements. The traditional method of estimating ANPP is clipping and weighing. The high cost, time compseation and difficulty in cliping are the limitations of the traditional method, which makes it extremely difficult and expensive to estimate the ANPP of large areas. Thus, the aim of this study was to estimate above ground net primary production (ANPP) using vegetation indices. Methodology: Sampling of vegetation was performed in rangelands of Sabalan elevations in Ardabil province in altitude ranges from 1500 to 3300 meters in 2016. Nine sites were selected in study area and in each site three 100 m transects were placed with 50 m interval. Along each transect 5 plots (1 square meter) with 20 m from each other was placed and in each plot, total production and life forms including shrubs, grasses and forbs ANPP were measured. The initial net production samples were placed in an oven at 70 ° C for 24 hours and then weighed to determine their dry weight. Furthermore, veg/etation indices including NDVI, PVI3, RVI and SAVI were calculated using data of Landsat 8 Operational Land Imager (OLI) images for 2015. Due to the fact that the maximum growth of vegetation in the region is in June, the images were also selected at the same time. Then using general additive model in software R, curve of relationship between ANPP and vegetation was analyzed in two way Individual and Combined data. Finally, ANPP was modeled using non-linear regression. Results: The application of generalized additive models for each of the vegetation indices with total ANPP and life forms separately shows that NDVI, PVI3 and RVI have a nonlinear relationship with total ANPP and life forms. However, the SAVI index has a linear relationship with the ANPP of total and grasses. Also, all vegetation indices have a significant relationship with ANPP. The ranking of vegetation indices affecting the ANPP based on the coefficient of determination shows that the most important and least important plant indices are SAVI and PVI3 for shrubs, and PVI3 and SAVI for grasses, forbs and total ANPP. In the present study, contrary to expectations, NDVI vegetation index, which has many applications in the vegetation studies and its increase indicates the presence of more vegetation in the region, in the generalized additive model and the combined study of vegetation index and ANPP show a nonsignificant relationship. Results showed that nonlinear regression significantly increased the accuracy of ANPP estimation using vegetation indices. The coefficient of determination for total ANPP (0.80) is more than shrubs (0.74), grasses (0.75), and forbs (0.56) and among the life forms, forbs have the lowest coefficient of determination. Conclusion: Based on the results, suitable vegetation index for estimate the ANPP of life forms are different. Also, ANPP estimation using vegetative indices at the total level is more accurate than life forms. According to The results, the OLI image and the use of nonlinear regression models were able to adequately estimate the ANPP in the study area. Therefore, if similar results are obtained in other areas of Ardabil, it will be possible to generalize the results and estimate this important ecological indicator with less time and cost. This tool can also be used to provide information on the amount of forage production and thus determine the stocking rate, as well as the degree of pasture degradation.