Integration of Geostatistical and Sentinal-2AMultispectral Satellite Image Analysis for Predicting Soil Fertility Condition in Drylands

• Published in: ISPRS international journal of geo-information Vol. 11; no. 6; p. 353
• Main Authors: Shokr, Mohamed S., Mazrou, Yasser S. A., Abdellatif, Mostafa A., El Baroudy, Ahmed A., Mahmoud, Esawy K., Saleh, Ahmed M., Belal, Abdelaziz A., Ding, Zheli
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01-06-2022
• Subjects: Agriculture; Arid zones; cLHS; Confidence intervals; Decision making; Discriminant analysis; drylands; Environmental conditions; Environmental monitoring; Fertility; Geostatistics; GIS; Hypercubes; Image analysis; Image processing; Laboratories; Latin hypercube sampling; Mapping; Nitrogen; ordinary kriging; Organic matter; Organic phosphorus; Organic soils; pH effects; Phosphorus; Potassium; Reflectance; Regression analysis; Remote sensing; S2A image; Sampling; Satellite imagery; Satellites; Software; Soil characteristics; Soil conditions; Soil fertility; soil fertility modelling; Soil organic matter; Soil properties; Soil surfaces; Spaceborne remote sensing; Statistical analysis; Statistical methods; Variables; Mediterranean Sea; Egypt
• ISSN: 2220-9964; 2220-9964
• DOI: 10.3390/ijgi11060353

For modelling and predicting soil indicators to be fully operational and facilitate decision-making at any spatial level, there is a requirement for precise spatially referenced soil information to be available as input data. This paper focuses on showing the capacity of Sentinal-2A(S2A) multispectral imaging to predict soil properties and provide geostatistical analysis (ordinary kriging) for mapping dry land soil fertility conditions (SOCs). Conditioned Latin hypercube sampling was used to select the representative sampling sites within the study area. To achieve the objectives of this work, 48 surface soil samples were collected from the western part of Matrouh Governorate, Egypt, and pH, soil organic matter (SOM), available nitrogen (N), phosphorus (P), and potassium (K) levels were analyzed. Multilinear regression (MLR) was used to model the relationship between image reflectance and laboratory analysis (of pH, SOM, N, P, and K in the soil), followed by mapping the predicted outputs using ordinary kriging. Model fitting was achieved by removing variables according to the confidence level (95%).Around 30% of the samples were randomly selected to verify the validity of the results. The randomly selected samples helped express the variety of the soil characteristics from the investigated area. The predicted values of pH, SOM, N, P, and K performed well, with R2 values of 0.6, 0.7, 0.55, 0.6, and 0.92 achieved for pH, SOM, N, P, and K, respectively. The results from the ArcGIS model builder indicated a descending fertility order within the study area of: 70% low fertility, 22% moderate fertility, 3% very low fertility, and 5% reference terms. This work evidence that which can be predicted from S2A images and provides a reference for soil fertility monitoring in drylands. Additionally, this model can be easily applied to environmental conditions similar to those of the studied area.