Remote sensing of agricultural salinity.

Abstract

Salinity represents the major environmental threat to arable land in Western Australia and many other parts of the world. This study was designed to establish criteria for a practical remote sensing system using the visible, reflected and shortwave infrared for the early detection and mapping of salinity. The results are principally from a group of study sites on the CSIROs Yalanbee Experiment Station, and from other significant sites during the agricultural cycles of 1985-7.Analysis of imagery from the Geoscan Multispectral Airborne Scanner showed that best discrimination between study sites affected by salinity, and those not affected, was provided by bands 3 (650-700 nm), 4 (830-870nm) and band 6 (1980-2080nm). The maximum discrimination occurred in a September 1986 flight (spring-flush). Although excellent discrimination was also evident in August and November in 1985, this could not be reproduced in November 1986. The visible and reflected infrared bands 3 and 4 featured prominently, but the significance of the short wave infrared bands was evident especially when vegetative ground cover became a less dominant factor.Field spectra collected over the same period with the Geoscan Portable Field Spectroradiometer (PFS) supported the aircraft data to a certain extent. Detailed analysis of the fine non-correlated structure of narrow constructed bands, from PFS data, indicated that improved discrimination between sites could be provided over a wider time window extending into the summer and autumn. This is when weather-related conditions, i.e. cloud, soil moisture and sun angle, are more conducive to extensive surveys.The importance of at least one narrow band centred near 1985 nm was determined. Laboratory spectra of bare soil from sites measured on an Hitachi Spectrophotometer also provided the importance of the shortwave region adjacent to the 1900 nm water absorption.The study evaluated the spatial and spectral characteristics of existing satellite systems such as Thematic Mapper and the Multispectral Scanner on the Landsat series and determined that a spatial resolution of about 20-30 metres was most appropriate for detection of salinity at a scale whereby management could be implemented.Ground electromagnetic techniques were evaluated during the study and the EM-38 Ground Conductivity Unit proved valuable for characterizing salinity status of the sites. The Lowtran Computer Code was used to model atmospheric attenuation and results indicated that the positioning of a narrow shortwave infrared waveband, centred at 1985 nm, is possible.