Abstract: The study of the spatial distribution pattern of plant fine roots is the basis to develop the models of plant roots. At present, the main method includes direct measurement and proper gird measurement. These methods have a lot of limitation, due to human factors, having much error and low efficiency. This paper aims to use remote sensing image processing technology for estimating the root length density of Pinus tabuliformis, with convenient, quick, batch process, and high precision. The Anjiagou catchment in the Longzhong Loess Plateau was selected. The spatial distribution pattern of Pinus tabuliformis fine roots was studied. Fine root samples were obtained in soil profiles. Picking up the fine roots from the soil samples, cleaning up and dry them. Scanning the roots from each sample, and then importing the images into ArcGIS. First, we should observe the grid values by identification, the values of the roots are 120. Then setting the roots 1, others 0 by raster calculator. When measuring the length by ArcGIS, we should use vector data, so needing to change the grid data to vector data. After that, the jagged line will appear, leading to the result bigger than real length. In order to improve the accuracy, we should smooth the vector data. At last, building model by the tools of the above in the ArcGIS, and carrying on batch process to each sample. We can get the result fast and accurately. The root length density of Pinus tabuliformis fine roots was estimated based on remote sensing image processing technology. Absolute error is from 2.2 mm to 4.6 mm, accuracy is 98%. So this shows that calculating root length by geographic information system technology is with high accuracy. The density distribution of fine roots of Pinus tabulaeformis was analyzed and simulated. Both 1D function (including in vertical and horizontal directions) and 2D function of the root length density were established. In vertical direction, the root length density of Pinus tabulaeformis mainly distribute in the 0-100 cm soil layers with logarithm distribution, which makes up 91.9% of whole roots; In the horizontal direction, the root length density augments with the increase of distance to tree from 0 to 240 cm, and decreases with increase of distance to tree from 240 to 500 cm. The distribution of fine roots is mostly crowded in the soil layers from 150 to 300 cm, making up 46.5% of the total roots. Most of the functions of root length density were simulated with function of eexponent, and R2 are 0.94 and 0.804, respectively. The result is helpful to theoretical research in SPAC (Soil-Plant-Atmosphere Continuum) system of Pinus tabulaeformis, at the same time this method can be applied in the extensive research.

Key words: Pinus tabuliformis, image processing technology, roots length density, Anjiagou