SOIL SALINITY FORCASTING USIG SPECTRUM REFLECTIVITY DATA *

Prof. Assist. Prof. Researcher 1,2 Assis. Prof. Soil Sci. and Water Reso. Dept., Agric. College, Diyala Univ., and Al-Anbar Univ., Iraq respectively 3 Postgradute Student, Soil Sci. and Water Reso. Dept., College of Agric., Al-Anbar., Iraq Corresponding Author: altamimiraad29@gmail.com ABSTRACT This study was conducted to assess soil salinity forecasting using spectral soil reflectivity. Artificial salinization was carried out on silty clay loam soil. Collected soil sample was handy crushed, sieved through a 4 mm sieve and backed in plastic columns. The columns were closed from the bottom with a perforated plastic lids with the presence of sand-gravel filter. Columns placed vertically at plastic basins contain saline ground water and left for salinization by capillary rise. At the desired salinity level, soil reflectivity was measured using spectroradiometer and wave length between 350-2500 nm, and band width 1 nm. Soil salinity and moisture were determined soon after spectral measurements. Data processed and converted to digital data using ViewSpecPro software. MS Excel 2010 was used to calculate reflectivity data for bands equivalent to those used with the sensor OLI used at LandSat-8. SPSS V.23 statistic program was used to formulate mathematics models (Multiple linear, Quadratic and Cubic) that describe the relationship between soil salinity and spectral reflectivity at three soil moisture levels i.e. 8, 18 and 24%. Results confirmed the efficiency of the three models to forecast soil salinity at 19 dS m -1 or higher and at soil moisture of 24%. The quadratic and cubic models also gave good results at soil salinity of 9 dS m -1 or more and at 8% soil moisture level. At soil moisture of 18%, the Quadratic and Cubic models showed behavior similar to their behavior at the lower moisture level, while the linear model was efficient at salinity level of 40 dS m -1 and higher.


INTRODUCTION
Salt-affected soils are widely separated in the world.It is area estimated at 1 billion hectares.This equal to 7 % of the area of Earth land (9).In the Near East, salt affected soils is estimated at 105.6 million hectares, or 5.9% of the area of that region (5).In Iraq, salinization process is concentrated mainly in Mesopotamian plain (8).Land degradation as a result of soil salinization reduce the area of productive land by about 30% (10).Although salinization process are well known in Iraqi soils since many centuries, but it increased significantly in the years 2002-2013 (10).This was due to the repeated dry seasons and poor management of soil and irrigation.Monitoring and follow-up natural resources and processes, including land salinization using suitable developed means commensurate with the size of the problem and its rapid separation, has become an urgent necessity to achieve sustainable development.Many countries found their aims in the technology of remote sensing.This technology is fast in performance and reduces efforts and costs, as well as being a historical record that can be consulted whenever needed (11).Lillesand and Kiefer (14) mentioned that remote sensing is one of the modern technologies which can be used to diagnose and predict many soil characteristics.This was because of the availability of data for large areas in many spectrum at a short time.Also, Al-Heity and Al-Wehishi (1) reported that the data provided by remote sensing technology has an important role in different studies.All parts of electromagnetic spectrum can be used to increase the understanding and interpretation of most phenomena studied by these technology.The development of spectroscopy equipment analysis and accessories, and the means of aviation and computers has opened up a huge sources of data about atmosphere and natural resources.In the past, access to such as these data was carried out by primitive ways accompanied with many palaces, as well as waste of time, effort and money (3).Salts in arid and semiarid regions are more precipitated and crystallized in the surface of the soils (12).Increasing salt concentration increases the spectral reflectivity of the soil surface (12).These findings were also noticed by other workers (13 and 16).They explained that the reflectivity of the soil in the visible and reflected infra-red (IR) electromagnetic spectrum increases with increasing soil salinity.Sadiq and Howari (15) explained that the best part of electromagnetic spectrum to identify saline soils is the band between 660 to 2200 nm.Other workers (8) recorded that salts increase soil reflectivity at middle IR (WL=1300-3000), except water absorption band.Due to the high benefit of using remote sensing in soil studies, and to the high correlation between soil salinity and its spectral reflectivity, so this work was conducted to assess predicting of soil salinity from soil reflectivity data at different soil moisture levels.

MATERIALS AND METHODS
Non-Saline silt clay loam soil classified as Typic Torrifluvent was used in this study.Soil material was sampled from the surface layer (0-30 cm) of a field at the college of Agriculture, in Abu-Ghraib.Collected soil material was handy broken up, air dried, sieved through a 4 mm sieve and then packed in polyvinyl chloride columns 40 cm in height and 7.5 in diameter.The columns were closed from the bottom with a perforated plastic lids.Table 1 explained some physical and chemical properties of the soil sample under field condition.

Table 1. Some properties of the soil used in the study
Filter of 5 cm in height, consisted from two layers of gravels (2 cm thickness for each), and one layer of sand and filter paper was placed at the end of each column.Gravels diameters of the lower layer and the layer above it were 9-4 and 4-2 mm, respectively.The diameters of sand were 2-1 mm.Soil packed in each columns for 33 cm in height, to achieve bulk density as it is in the field.Columns placed in plastic containers which connected to each other by plastic pipes to Water level in containers maintained constant by using a raft placed in the first container.Total number of columns were 200, each 10 columns were placed in one container.When soil salinity reached the desired level, as set of columns taken, while the rest stay at the container to achieve progress salinization.Desired salinity levels have been checked using additional soil columns.Soil Spectral Reflectivity Soil spectral reflectivity were measured at 8 salinity levels and 4 moisture levels, measuring was done at 5 replicatios for each treatment.Table 3 summarize these treatments.Each set of columns, representing 1 soil salinity level × 4 soil moisture × 5 replications) was divided into 4 groups randomly, each group represent one moisture level.Soil reflectivity was measured by spectroradiationmeter using narrow bands (1 nm), have a length between 350 to 2500 nm.After that, soil samples were collected from the upper 5 cm of soil in each column to determine EC and soluble ions.Each group of columns were left to the next day or the next to reach the required less moisture level.Then, its reflectivity was measured at the required moisture.Reflectivity measuring and soil sampling was repeated with each group of columns.

Laboratory Work
Mechanical analysis, bulk density and soil moisture were determined using pipette method, cylinder method and gravimetrically respectively, as was described by Black et al. (6).Electrical conductivity for soil sample collected from the field and those collected from columns was carried out at 1:1 soil: water extracts.Results then converted to soil paste extract using conversion factor between EC e and EC 1:1 for the studied soil which was 2.1.Organic matter was determined using modified method proposed by Walkley-Black.All chemical analysis was carried out as was described by Al-Tamimi (4).

RESULT AND DISCUSSION
Digital Data for spectral Bands Spectral reflectivity data, for bands equivalent to those used with the sensor OLI on the satellite LandSat-8, at different levels of soil salinity and three soil moisture levels illustrated in tables 4. Generally low reflectivity values were recorded at higher moisture.Data in table 3 shows that spectral reflectivity values of the used bands at moisture level 24 %, did not have a given curve with salinity levels.It decreased and increased randomly.Table 4 shows spectral reflectivity values at different salinity levels and at three soil moisture levels, i.e. 8, 18 and 24 %.Results indicate that at 8 and 8% soil moisture levels and high salinity levels (60 and 78 dS m -1 ), the spectral reflectivity values in the first four bands (i.e.B1, B2, B3, and B4) were nearly equal to each other.Also, highest differences in reflectivity values were recorded with B6 band, followed by the band B9 and then B7 at all moisture levels.Whereas the differences decreased with increasing soil moisture level.
In the first five bands (B1 to B5), and at all soil salinity levels, the differences between soil reflectivity increased with increasing band wave length at soil moisture levels 8 and 18%.While at 24% soil moisture level, the differences between the reflectivity of these five bands differed randomly and did not correlate with their length (Table 3).This may be due to the effect of high moisture level in this soil.

Forecasting Soil Salinity
Results indicated that all used mathematical models were suitable and can be used to predict soil salinity from reflectivity data, for equivalent spectrum bands which used in OLI sensor.Three of these models are shown in table 5.These models were multiple linear, Quadratic and Cubic.The rest two models (quartic and exponential) were neglected because their results were similar to those obtained by the linear, quadratic and cubic models which are simpler for application.The multiple linear model explained that soil salinity positively correlated with the band B9 at soil moisture level of 8%.The determination factor (R 2 ) and S.E values were 0.94 and 7.1, respectively.At 18% soil moisture level, soil salinity positively correlated with the recorded reflectivity in the band B6 and negatively with the recorded reflectivity in the band B9.The values of R 2 and S.E were 0.99 and 3.1, respectively.Using Quadratic model at 8 and 24% soil moisture level, a negative correlation was noticed between recorded reflectivity and soil salinity in the band B9, while positive correlation was noticed between squares of reflectivity value and soil salinity in this band.The values of R 2 and S.E were 0.99 and 2.6, respectively at 8% soil moisture level, and 0.98 and 3.3, respectively at 24% soil moisture level (Table 5).At moisture levels of 18%, Quadratic model showed that soil salinity negatively correlated with reflectivity and positively correlated with the squares of reflectivity value in B6 band.The values of R 2 and S.E were 0.99 and 1.3, respectively.Cubic model confirmed that soil salinity negatively correlated with square value of reflectivity at B6 band and positively correlated with reflectivity cubic value at the same band, and at both soil moisture level, 8 and 18%.The values of R 2 and S.E were 0.99 and 2.1, respectively at soil moisture of 8%.Whereas, these values were 0.99 and 1.2 at soil moisture of 18%.At soil moisture of 24%, Cubic model explained that soil salinity had a negative correlation with reflectivity data and a significant positive correlation with cubic value of reflectivity recorded at B9 band.The values of R 2 and S.E were 0.98 and 3.3 respectively (Table 5).To examine suitability and successfulness of the mathematics models in forecasting soil salinity levels, the three used models were tested using new soil samples have different salinity levels and did not participate in creation these models.Absolute relative error percentage was used as a criteria to assess viability and goodness of fit of each model (Table 6).Model was accepted when A.R.E percentage equal or less than 10% (2).Results in table 4 pointed out that the three used models were effective in forecasting soil salinity at the levels equal to 19 dSm -1 or more at moisture level of 24%.At soil moisture of 8%, linear model did not gave a clear results to forecast soil salinity.The quadratic and cubic models both show similar acceptable results.Apart from soil salinity of 13 and 27, the two models can be used to forecast soil salinity of 9 dSm -1 and more (A.R.E <10).At soil moisture of 18%, the linear model was effective in forecasting the last three studied salinity levels i.e. 40, 60 and 80 dS m -1 , while the results with the quadratic and cubic models were similar to those recorded with these two models at 8% soil moisture (Table 6).]Table 6