I wanted to predict land-use changes using the data that i have from 1991, 2000, 2010 and 2019. how can i predict land-use changes? let's say
in 1991 i have built-up area 55 agriculture is 510 and forest 373
in 2000 built-up area 120 agriculture is 459 and forest 359
in 2010 built-up area 147 agriculture is 401 and forest 390 and
in 2019 built-up area is 219 agriculture is 300 and forest 349
. how can I predict what will be built up agriculture and forest in 2030?
I would not recommend using output variables as regressors. That would not make any sense in this case, because land use changes depend on numerous factors - specific ones for each output variable. I would suggest looking for parameters which influence those you want to predict, such as income, population, climate etc.
A simple way of looking at it in the beginning is to assume that there is a linear relationship. For example, plot Time vs Forest, with time on the x-axis and forest on the y-axis. Then look at the equation of that regression, and put 2030 on the x-axis and see what you get on the y-axis.
Suggested approach by Dr. Frisk in Excel is to plot all three values (columns) with with respect to time. Excel allows linear, curvilinear, logarithmic, and exponential curves to be applied, and will calculate r2, etc. to each column. By plotting the values, time can be extended also to 2030, so the equation best fit line or curve can use that date for each equation. Projecting past and present data and trends into the future can be helpful, but has potential for errors.
I would not recommend using output variables as regressors. That would not make any sense in this case, because land use changes depend on numerous factors - specific ones for each output variable. I would suggest looking for parameters which influence those you want to predict, such as income, population, climate etc.
Prediction of future land use change is highly prone to uncertainty. And so many factors affect land use change such as socio-economic reasons, land use policy, climate variability and etc. thus, the change would seldom be linear. Factors could also be global or regional. Based on historical land use changes, and by considering different factors of land use change (physical and/or socio-economic factors) transition probability can be developed. There are various approaches including but not limited to Markov chains, Cellular Automata, Logistic Regression.
- Badges
- Science topic
- Similar topics
- Agriculture