The GSTAR and GSTARX models normally can only be formed from observed locations. The problem that sometimes occurs is that not all locations that want to be modeled have complete data as well as other locations. This study uses GSTAR and GSTARX modeling using SUR approach and combines them with the kriging interpolation technique for forecasting coffee berry borer attack in Probolinggo Regency. This modeling is called GSTAR-SUR Kriging and GSTARX-SUR Kriging. This study aims to determine the best model between GSTAR-SUR Kriging and GSTARX-SUR Kriging for forecasting coffee borer attack in an unobserved location. The result of this study shows that GSTAR-SUR Kriging and GSTARX-SUR Kriging models can be used for forecasting coffee berry borer attack in unobserved locations with high forecast accuracy shown by MAPE values <10%. In this study the GSTARX-SUR Kriging model (1,[1,12])(10,0,0) is the best model for forecasting boffee berry borer attacks in unobserved locations.

The research aimed to use Generalized Space Time Autoregressive (GSTAR) and GSTARX modeling with the Seemingly Unrelated Regression (SUR) approach and combine them with the Kriging interpolation technique in an unobserved location. The case study was coffee borer beetle forecasting in Probolinggo Regency, East Java, Indonesia, with Watupanjang Village as the unobserved location. The results show that GSTAR-SUR Kriging and GSTARX-SUR Kriging models can predict coffee borer beetle attacks in unobserved areas with high accuracy. It is indicated by the Mean Absolute Percentage Error (MAPE) values of less than 10%. The addition of exogenous variables (rainfall) into the model is proven to improve the accuracy of the model. The Root-Mean-Square Error (RMSE) value of the GSTARX-SUR Kriging model is smaller than the GSTAR-SUR Kriging model. The structure of the model produced from the research, GSTARX-SUR (1,[1,12])(10,0,0), can be used as a reference in modeling coffee borer beetle attacks in other regencies. Map of forecasting coffee borer beetle attack shows that the spread of coffee borer beetle attack is spatial clustering with the attack center located in the eastern region of Probolinggo Regency.

Generalized Space Time Autoregressive (GSTAR) is one of multivariate time series modeling that considers aspects of location with heterogeneous location characteristics. The GSTAR model normally can only be used in forecasting an event in the future at the observed locations. The problem that often occurs in some cases is that there are locations to be modeled that do not have sufficient or incomplete data as data in other locations. For this reason, several alternatives can be done, and one of them is by combining the GSTAR model with the kriging interpolation technique. This modeling is known as GSTAR Kriging modeling. In this research, GSTAR Kriging modeling is applied in predicting and mapping coffee berry borer attacks in Probolinggo District. The model parameters are estimated using the GLS method in the SUR equation system. Forecasting results indicate that the GSTAR Kriging model has a high forecasting accuracy and is not much different from the GSTAR model. Meanwhile, based on the forecasting map, it can be seen that the peak of coffee berry borer attack is predicted to occur in July 2019 with the attack center located in Tiris Sub-district.

Generalized Space Time Autoregressive (GSTAR) is one of the multivariate time series models considering heterogenic location. One of the GSTAR model developments is GSTARX model with additional exogenous variables. Parameters of GSTAR model could be estimated using Seemingly Unrelated Regression (SUR) approach to cope with the residual model between locations generally related to each other. The model is commonly called GSTARX-SUR. It is applicable in various fields such as agricultural sector. In this research, GSTARX-SUR model was applied to predict cocoa black pod attack in Trenggalek Regency. Rainfall was used as the exogenous variable. One of the characteristics of GSTARX model is the spatial weights. The correct spatial weights in GSTARX model is expected to improve the accuracy result. The research aimed to obtain the best GSTARX model to predict cocoa black pod attack in Trenggalek Regency. The research findings showed that GSTARX-SUR model (1,[1,12])(0,0,0) using inverse distance weighted matrix was the best model. The prediction result was highly accurate, indicated by a small MAPE value less than 15%.