نوع مقاله : مقاله پژوهشی
نویسندگان
1 گروه باغبانی و گیاهپزشکی، دانشکده کشاورزی، دانشگاه صنعتی شاهرود، شاهرود، ایران
2 گروه مهندسی آب، دانشکده کشاورزی، دانشگاه صنعتی شاهرود، شاهرود، ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
The Moroccan locust (Dociostaurus maroccanus), as one of the native pests of Iran, is recognized for its migratory capability, population outbreaks, and severe damage to crops and rangelands. In this study, meteorological data (temperature, precipitation, and relative humidity), field-measured data (soil moisture and locust abundance), and indices derived from Landsat 8 satellite imagery (LST, TCI, OSAVI, and BSI) were used as input variables for a multilayer perceptron (MLP) neural network model to estimate locust abundance at the pixel scale. To this end, the dataset was randomly divided into training (70%) and testing (30%) subsets. Model performance was evaluated using the correlation coefficient (R) and the Nash–Sutcliffe efficiency (NSE). The results indicated that the input variables of the simulation model were appropriately selected and that the model structure was capable of simulating locust abundance with very high accuracy, such that NSE values of approximately 0.98 and 0.99 were obtained for the training and testing phases, respectively. Furthermore, the results showed that soil moisture exhibited the lowest correlation with measured locust abundance (R=0.9), while the TCI index showed the highest correlation (R=0.98). Overall, the findings of this study demonstrate that the combined use of field data and satellite-derived indices within machine learning frameworks can be highly effective for simulating the habitat of the Moroccan locust. The proposed approach enables large-scale and multi-temporal monitoring of Moroccan locust populations, given the availability of satellite imagery, thereby providing valuable support for experts and decision-makers in pest management and control planning.
کلیدواژهها [English]
)