The use of water from streams depends on their quality in often degraded by excessive nutrient loads such as phosphorus. The study of this nutrient by a method of sequential analysis in sediments aims to provide information on environmental quality, while determining the form of bioavailable phosphorus.
This work aims to put the focus on the different forms phosphorus could be found in the aquatic sediments of the main rivers in the region of Meknes (Morocco).
The determination of total phosphorus level in sediments is made using spectrophotometry applied on the supernatant after mineralization of the sediment with (H2SO4/K2S2O8) according to ammonium molybdate method using ascorbic acid as reagent. The chemical fractioning of phosphorus in the sediment is done following the Golterman fractionation scheme that uses EDTA as specific chelating agent that extract the mineral fraction without disturbing the organic phosphorus by adjusting the pH of extracting solutions to that of the sediment.
This scheme will allow us to determine the proportions of five forms of phosphorus: water-soluble phosphorus (o-P), iron-bound phosphorus (Fe (OOH)-P), calcium-bound phosphorus (CaCO3-P), acid-soluble organic phosphorus (ASOP) and residual organic phosphorus (ROP).
The results of these extractions showed that the sediment phosphorus of studied samples is mainly under its mineral form [Fe(OOH)-P + CaCO3-P]. These forms represent around 77% of total phosphorus with a predominance of the CaCO3-P fraction. The organic forms (ASOP + ROP) represent only 23%.

This work presents the development of an artificial neural network model based on the Multi-layer Perceptron (MLP) and Radial Basis Function (RBF) for predicting the moisture in the zone of Chefchaouen (Morocco). Our objective is to treat a chronological series of measured data for network response evaluation. For this reason, we used a basic learning neural model for the years 2008 and 2013, the latter consists of a number of meteorological parameters such as humidity, temperature of the air, dew point temperature, air pressure, visibility, cloud cover, wind speed and rain. To determine the network architecture to be used, we varied the number of hidden number of neurons in hidden layers, transfer functions and pairs of transfer functions and learning algorithms. Models performances have been evaluated and developed through the study of the Mean Squared Error (MSE) and correlation coefficient (R). We demonstrated in this study for the prediction of moisture, the best performing model is the one used as transfer functions, the Tansig function in the hidden layer and the Purelin function in the output layer, while using a learning algorithm LM, PMC type of configuration.