Assessment of land use changes using multispectral satellite images and artificial neural network

Introduction
Land use change reporting over time is necessary to assess and monitor the state of natural and agricultural resources. Knowing about the change of use is necessary to identify the priorities of public investment in the management of natural resources and to evaluate its effectiveness. The purpose of land change investigation is land use management. Among the management cases, we can mention the evaluation of the effect of economic activities and development on the environment. In such cases, these organized reports are the best sources of decision-making. Land use management can ensure that resources are used efficiently and people's future resources are preserved. This process is the main component of a development plan. Timely and accurate detection of land use changes is the basis for a better understanding of the relationships and interactions between humans and natural phenomena, and as a result, provides better management and more appropriate use of natural resources. Satellite images as a type of remote sensing data are well used in the field of natural sciences for quantitative and qualitative measurement of land cover changes. The construction of the Houzian dam in 2015 and also the expansion of mines and construction in the natural resources of Aligudarz city in Lorestan province and also the lack of accurate statistics on the amount of land cover/use changes in the region make such research necessary. In the present study, land use changes in Aligudarz city were investigated during nine years for the years 2012 and 2014 with the help of multi-spectral satellite images and the artificial neural network. In the structure of the artificial neural network, numerous nodes work together in parallel with the purpose of processing. Each node is a data structure. This data structure is placed in a communication network with each other and the network is taught by humans.
 
Materials and Methods
In this study, several key steps were used to prepare and identify LULC changes in Aligudarz city, which include data pre-processing, image processing and classification implementation as well as validation. The required images were selected among the available images in such a way that they have minimum cloud cover and maximum greenness in the plants and trees in the area, and the date of the images are related to the same month. This study uses land use change detection in the east of Aligudarz county using Landsat 8 OLI and TIRS images. The spatial resolution of these images was improved to 15 m using the fusion technique and panchromatic band. At first, preliminary pre-processing including radiometric, atmospheric, and geometric corrections were done on the raw data. The geometric correction was done with the RMS square root error of 0.22 pixels. Radiometric and atmospheric corrections were done in ENVI 5.3 software using Radiometric Calibration and Quick Atmospheric Correction tools. The artificial neural network method was used to prepare land use maps for 2013 and 2021. The neural network structure used in this research is a three-layer perceptron neural network, which includes seven input neurons (number of satellite image bands), eleven intermediate neurons, and six output neurons (number of land cover map classes). The classification accuracy was evaluated quantitatively by comparing the LULC classes obtained from the training phase with the data obtained from the testing phase. The classification accuracy was evaluated quantitatively by comparing the LULC classes obtained from the training phase with the data obtained from the testing phase. In this study, the points taken from the ground surface and Google Earth Pro 7.3.4.8642, using the error matrix and the Confusion Matrix Using Ground Truth ROIs tool were used. Detection of changes between two classified maps was done with Change Detection Statistics and Workflow Change Image and Spear Change Detection tools.
 
Results and Discussion
The results of this study showed that the artificial neural network has an acceptable performance in investigating land use changes and, The Kappa coefficient for 2013 and 2021 was 0.83 and 0.71%, respectively. Due to the construction of Houzian Dam in 2016, water areas have witnessed an increase of 1.34%. Also, the construction of the dam has led to an increase in the area under irrigated cultivation, so the area under cultivation in 2021 experienced an increase of 5.53% compared to 2013. In addition, the construction of the dam has caused the highlands to decrease by 4.30 %. Because the water of the dam has been used to irrigate the highlands where there was not enough water to irrigate them before the construction of the dam. The area of mines has increased by 0.23% during the studied period. The area of uncovered regions has decreased by 1.74% compared to 2013. Also, the area of habitation regions has decreased by 1.06% to 18.18 square kilometers in 2021.
 
Conclusion
The survey of the land use map of Aligudarz showed that the heights and water areas have the largest and smallest areas, respectively. The results of this study showed that in the years after the construction of Houzian dam compared to before its construction. The total area of water and total vegetation has increased. Since the construction of a dam in an area has short-term and long-term effects, it should be noted that the increase in vegetation and the cultivated area is considered a short-term effect. Therefore, it is necessary to investigate the impact of creating this water structure in the region's ecosystem in the long term by forecasting models. In the investigation of mines, the appearance of water areas indicates an increase in the depth of excavation. Since this city is an important center for stone production, the absence of a specialized regulatory body on the number of harvests and the impact of mining on the environment is felt in this region. Another part of the increase in water areas is due to the existence of errors in the classification of land use in the artificial neural network. In using the results of this research, it is important to mention that these results were obtained for the area of the dam and the increase in vegetation caused by the construction of the dam cannot be generalized to the entire basin.