Improvement of AMMRR interpolation and application in CSCS classification of Inner Mongolia grassland

Spatial interpolation of meteorological data is critical in meteorological and grassland research at macroscopic scales. To improve precision and enhance spatial differentiation effects of the Analytic Method based on Multiple Regression Residues (AMMRR) interpolation for the classification of grassland types, the AMMRR was upgraded by adding micro-topography factors such as slope and aspect, while the Inverse Distance Weighted (IDW) function was replaced with Ordinary Kriging (OK) function. And then the Improved Analytic Method based on Multiple Regression Residues (I-AMMRR) interpolation was established. The I-AMMRR was applied in the spatial interpolation of annual accumulated temperature of > 0 ℃ (Σθ), annual precipitation (r) and humidity (K) in the Inner Mongolia Autonomous Region. Subsequently, the grassland was classified and the corresponding Comprehensive Sequential Classification System (CSCS) of rangeland map produced for the Inner Mongolia Autonomous Region. The CSCS and I-AMMRR technique were used in conjunction with meteorological data from 197 stations for the period from 1961 to 2004. The research results suggested that: (1) The mean absolute error (MAE), mean relative error (RME) and root-mean-square error (RMSE) cross validation indices of the training and test datasets of Σθ and r were lower for I-AMMRR than AMMRR. The R correlation coefficients for Σθ and r between simulated and observed values were 0.969 (P < 0.01) and 0.933 (P < 0.01), respectively. This suggested that the interpolation accuracy of I-AMMRR was higher than that of AMMRR. (2) The spatial distribution patterns of Σθ, r and K simulated by I-AMMRR for the Inner Mongolia Autonomous Region showed horizontal and vertical zonalities, which were strongly agreed with actual distribution of natural zones and natural landscapes in the region. Specifically, Σθ increased from southeast to northwest while r and K had the reverse pattern. (3) Grassland in the Inner Mongolia Autonomous Region was classified into a total of 17 CSCS classes, ranging from the cool temperate-humid mixed coniferous broad leaved forest (ⅢF38) to the cool temperate-extra-arid temperate zonal desert (ⅢA3). The cool temperate-arid temperate zonal semi-desert (ⅢB10) covered the largest area while the smallest was under warm temperate-humid deciduous broad leaved forest (ⅣE32), with approximate areas of 2.42×105 km² and 0.647 km², respectively. The landscape types from east to west included forest, forest-rangeland, meadow, steppe, semi-desert and desert in that sequence. With simultaneous increase in altitude, the land cover distribution across the region between Ergun River-Hulun Lake and Da Hinggan Mountains was steppe (ⅢC17), meadow (ⅢD24), montane meadow (ⅢE30) and coniferous forest (ⅢF37) in that successive order. Meadow (ⅢD24), forest-rangeland (ⅢE31), and coniferous and broad-leaved mixed forest (ⅣF38) were orderly distributed in the region from southern Nenjiang Plain to Da Hinggan Mountains. Steppe (ⅢC17), meadow (ⅢD24), forest-rangeland (ⅢE31), and coniferous forest (ⅢF37) were distributed across the area linking Xiliao River Plain to Da Hinggan Mountains. Horizontal and vertical zones were distinct in spatial distribution of grassland vegetation, which well fitted the result of vegetation investigation in Inner Mongolia Autonomous Region. These research results provided theoretical evidence and technical support for dynamic monitoring of the variations in grassland in Inner Mongolia Autonomous Region.