Abstract
Pasture yields often fluctuate with differences in management, because pasture is usually cultivated for a number of years under the pressure of grazing or cutting several times a year. However, pasture has wide areas of non-uniform growth distributions in contrast to row crops such as rice or wheat. Therefore a grasp of pasture productivity is difficult to ascertain at present which complicates the inevitable need to build plans for the renovation of pasture and/or the purchasing of supplemental forage on a farm. A study was undertaken for estimating pasture yields using Landssat multispectral scanner (MSS) data. In the previous paper, six land-use groups including a "Grassland" group are identified using Landsat MSS data acquired over Northern Tochigi Prefecture. In this study, multiple regression models were created using first cutting yield data collected from 26 pasture plots in Nishinasuno Town (Table 1) and combining this information with Landsat spectral reflectance data (CCT count) of the six channels appearing in Table 2. Flow of analysing process is shown in Fig.1. Results obtained are as follows ; 1) Band 6 (infrared) had the highest correlation with yield (r=0.823) when it used single band. 2) Multiple regression coefficients improved in accordance with the increment of number of channels entered in the model, however, this increase was retarded when four channels were exceeded (Table 3). The equation used in the best four channels (R=0.923) is expressed as, Y_<est>(kg/ha)=-177.7X_1+132.5X_3+99.9X_4-216.216+10642.5 Here, estimated yield (Y_<est>) is expressed as a function of Xs, and X_1 to X_6 in the equation corresponds to the CCT or computer compatible tape counts of Ch 1 to Ch 6, respectively. 3) Yield estimation by a multiple regression model using the best four channels was applied to our study area, and the magnified Shiobara-Nishinasuno area appeared in a yield map of Fig.4. In the map, four yield classes for the first cutting were indicated with numbering, 1 for 0-1.9ton, 2 for 1.9-4.1ton, 3 for 4.1-5.7ton, and 4 for over 5.7ton per hectare. 4) Because of a lack of reference information, accuracy had to be confirmed by field checks, however the general classification patterns were found to be acceptale.
