Home » Methods to explore the relationship between land change prediction and its variables through accuracy assessment. by Hao Chen
Methods to explore the relationship between land change prediction and its variables through accuracy assessment. Hao Chen

Methods to explore the relationship between land change prediction and its variables through accuracy assessment.

Hao Chen

Published
ISBN : 9781109125719
NOOKstudy eTextbook
121 pages
Enter the sum

 About the Book 

This three-article dissertation aims to investigate the complexity among these influential factors and to find the sources of modeling errors and uncertainties. The goal is to achieve a deeper understanding concerning land change dynamics and henceMoreThis three-article dissertation aims to investigate the complexity among these influential factors and to find the sources of modeling errors and uncertainties. The goal is to achieve a deeper understanding concerning land change dynamics and hence to provide useful guidelines for LUCC scientists to evaluate and to improve their modeling practices.-The first article focuses on exploring how prediction successes and errors are associated with the gradient of an explanatory variable. The paper proposes a method to quantify the goodness-of-fit of a land change prediction along the gradient of an explanatory variable, by classifying each pixel as one of four types: Correct due to Observed Persistence Predicted as Persistence (i.e. null sucesses), Error due to Observe Persistence Predicted as Change (i.e. false alarms), Correct due to Observed Change 2 Predicted as Change (i.e. hits), Error due to Observed Change Predicted as Persistence (i.e. misses). This helps to answer the following research questions: (1) Where along the gradient of an explanatory variable are the successes and errors of a land change prediction?, (2) How does the gradient of the explanatory variable relate to the stationarity of the land transition processes?, and (3) How can we use the answers to questions 1 and 2 to help to search for additional explanatory variables? The paper illustrates the method by examining a case study of an application of the model Geomod in Central Massachusetts, USA. Results reveal that the model predicts more than the observed amount of change on flat slopes and less than the observed amount of change on steep slopes. One reason for these types of errors is that the land change process during the calibration interval is different from the process during the validation interval with respect to slope. The method allows modelers to use the validation step as a diagnostic tool to search for potentially influential missing variables and to gain insight into land transition processes.-The second article is methodologically an extension of the first article. This article is intended to address a complicated research issue in LUCC modeling concerning how the results from a land change model can vary based on the spatial resolution at which the analysis is performed and the precision of the independent variable(s). This article introduces an approach to measure the variation of a land change models accuracy triggered by alteration of the pixel resolution and the precision of the independent variable, which is distance to previous built area for our case study. We illustrate the principles with an application of the Geomod land change model contained in the Idrisi GIS applied to 3 simulate the gain of built land in Central Massachusetts, USA. Results reveal four general principles: (1) change in pixel resolution using the Majority-Takes-All rule can influence quantity error- (2) change in bin width of an independent variable does not influence the quantity error- (3) bin widths of an independent variable that are very small or very large are not useful- (4) resolution and bin width interact so that bin width does not have an effect when bin width is smaller than pixel resolution. The method and knowledge developed in the paper can help land change scientists to understand more deeply the behavior of their models at various pixel resolutions and precision levels of the independent variables, and thereby guide them to prepare the data in a more efficient and appropriate manner.-The third article presents a method to measure how the predictive power of a land change model varies when the number of categories becomes less due to the aggregation of land categories. The study examines how the detail in the land cover map, thereby the complexity of land transitions, can affect the models predictive ability. The paper carries out a sequence of land change simulations using the Land Change Modeler (LCM) in the Idrisi GIS. The simulations are performed by hierarchically altering the number of land categories through aggregating two land categories to form a new aggregated set of classes in each of the runs until only two land classes remain. It is a mathematical fact that the amount of overall observed transition at the map level decreases or remains unchanged with land category aggregation, while the results for our study area show that the amount of predicted transition can either increase or decrease. Both overall accuracy and the change accuracy rate increase when the number of categories becomes smaller, for our 4 case study. The method and statistics proposed in the paper are generic, so they can be used for assessment of any land change predictions concerning multi-category transitions. (Abstract shortened by UMI.)