عنوان مقاله [English]
In this research, the continuity of land landscape was analyzed from the perspective of land landscape metrics (10 indicators) and graph theory indicators (12 indicators). First the correlation between correlation indices and water quality parameters was performed using Spearman and Pearson correlation coefficient and then with the help of stepwise linear regression method and through power, exponential and logarithmic regression models and using Akaike coefficient, the best model was chosen. The results showed that there was a significant negative correlation with water quality parameters between the land quality measures of CONIG and FRAC and the length of the corridor and most of the continuity indices of graph theory. Also, the highest coefficient of explanation or R2 belonged to CO3 models with a coefficient of explanation of 0.818, water flow with a coefficient of explanation of 0.733, Ca with a coefficient of explanation of 0.772 and TDS with a coefficient of explanation of 0.704. Almost all selected models are nonlinear models. Significant correlation of land feature continuity indicators, especially graph theory continuity indicators and corridor length with water quality indicators, shows the effect of structural features of land landscape on water quality in watersheds.
Keywords: Graph theory, Landscape metrics, Corridors, Forests, Modeling, Water quality.
The forests of the Caspian basin are one of the most important forest resources in Iran country, but what is seen as a large gap in research in this area is the need to pay attention to the structural dimensions, especially to maintain the structural connectivity of these forests. Various studies have attempted to determine the relationship between land cover structure and water quality, but most these, rely on landscape measures. It is also important not only to understand the total area of the resource and storage, but also their spatial arrangement with respect to the routes. Therefore, the evaluation of this relationship remains an important gap in research in this field. Therefore, the present study was conducted with the aim of modeling the effect of landscape connectivity on water quality indicators in forest patches of the Caspian Basin. In order to analyze the connectivity, a combination of graph theory and land use measurement approaches was used.
Each of these methods has limitations and advantages. Graph theory is an image tool based on which a network can be represented with its strengths and weaknesses. Numerous studies have acknowledged that graph theory is an effective way to model the landscape and its interactions and relationships. Graphs are generally recognized as a rapid method for analyzing ecological studies.
The first application of graph theory in ecology was made by Kantwell and Foreman in 1993 by simulating a heterogeneous land image. Then Bann et al. (2000) used graph theory to measure the coherence of the landscape and since then this theory has been increasingly used in the study of coherence. As a result, graph theory is often used in studies to measure and measure connectivity. Whereas the structural patterns that make up the landscape depend on the size, shape, number, origin and, most importantly, the arrangement of the patcehs in the landscape. Therefore, land use metrics are a good tool to express the structure of land appearance. Landscape metrics focus on the spatial characteristics and distribution of patcehs in the landscape. Although individual patcehs have few spatial characteristics, total patcehs can have multiple collective characteristics, which may be a type of patceh or a set of patcehs.
Materials and methods:
The study area is located in the southern catchment area of the Caspian Sea, between 36 degrees and 33 minutes to 38 degrees and 8 minutes north latitude and 48 degrees and 32 minutes to 56 degrees and 19 minutes east longitude of the origin meridian. The Caspian Sea catchment is one of the six main basins of the country and has an area of 58,167 square kilometers. Figure 1 shows the location of the study area.
Figure 1- Location of the study area
First, the forest layer was extracted from the land cover map and 25 large watersheds were selected. Landscape connectivity was analyzed and calculated from the perspective of land landscape metrics (10 indicators) and graph theory indicators (12 indicators). In the next step, first the correlation between correlation indices and water quality parameters was performed using Spearman and Pearson correlation coefficient and then with the help of stepwise linear regression method and through power, exponential and logarithmic regression models and modeling using Akaike coefficient. The best model was chosen.
Discussion and Results:
The results of correlation analysis also showed that there was a significant negative correlation with water quality parameters between the land quality measures of CONIG and FRAC and the length of the corridor and most of the connectivity indices of graph theory. Also, the highest coefficient of determination or R2 belonged to CO3 models with a coefficient of explanation of 0.818, water flow with a coefficient of explanation of 0.733, Ca with a coefficient of explanation of 0.772 and TDS with a coefficient of explanation of 0.704.
Table 1- Results of selected models
Number Model Type Relation R2 RMSE
1 Power LogTDS= -1.793-5.052 (logcontig)+66.877(logfrac) 0.704 0.265
2 exponential LogEC= 2.535-7.138e(corridor) 0.408 0.0352
3 Logarithmic Ph= 15.214-2.608(logdLCP)-1.118(logTE) 0.558 0.400
4 Logarithmic CO3=0.011-0.040(logdA) 0.818 0.025
5 exponential logHCO3=0.564-0.124(dPCintra) 0.426 0.089
6 Power LogCl=0.488-0.168(logIICintra) 0.341 0.082
7 Power LogSo4=0.182-1.337(logdLCP) 0.455 0.218
8 Power LogCa= 0.230-16.450(logFRAC)-0.112(logdPCConnector) 0.772 0.095
9 exponential Logmg= -1.688-0.135(dCPflux)+1.136(PAFRAC) 0.458 0.0739
10 Logarithmic Na= 1.150-2.302(logdLCP) 0.507 0.381
11 Logarithmic K= -0.338-0.464(logcontig)+5.818(logFRAC) 0.695 0.029
12 Linear SAR= 0.315-0.041(dIICintra)-0.548(dIICconnector) 0.479 0.026
13 Power Logdebi= 0.242+2.906(logdIIC)+0.229(logdIIcconnector) 0.733 0.508
Considering the importance of maintaining the cohesion of forest patcehs, and its role in preserving the biodiversity and functions of Caspian forests, the results of this study will be used in forest management and rehabilitation programs; So that in areas with low connectivity, extensive programs are needed to rehabilitate and prevent further rupture of these patcehs. Also, the modeling results showed that almost all selected models with non-linear models are nonlinear models and this shows the nonlinear relationships between connectivity indices and water quality parameters. The results showed that graph theory indicators are a powerful, simple and at the same time efficient tool for examining the connectivity of patcehs in a landscape. Also, the significant correlation of land feature connectivity indicators, especially graph theory connectivity indicators and corridor length with water quality indicators, shows the effect of structural features of land landscape on water quality in watersheds; That is, fragmentation of the landscape and staining greatly affects water quality.
Therefore, this study was the first study to investigate the effect of connectivity indicators on water quality of watersheds, showed that land connectivity in addition to the effects that have been studied in various studies on wildlife biodiversity in an area, Also has a great impact on other functions of an ecosystem, such as water supply and quality, which shows the need to protect and prevent the rupture and staining of the landscape. Due to the ease of use of graph theory and the resulting connectivity indicators, and at the same time the accuracy of the results in examining the connectivity of land features, the use of this approach is recommended as an efficient method. With the identification of valuable patcehes, management planning should be directed towards their protection in order to obtain the maximum profit with the least cost in order to protect important patches.