Methodology
Landsat TM imageries taken from different dates for ten years were used for the study. The list of imageries are presented on Table 1. The imageries were rectified using roads and river vector files as source of ground control points (GCPs). More than 50 GCPs were selected for each image and a root mean squared (RMS) error range of 1.5 to 1.9 was maintained for all imageries. All the imageries were classified and adjoining scenes were mosaicked.
Table 1. List of Landsat TM imageries used for the study and the date they were taken.
| Landsat TM Scences |
Date Taken |
| 116-48 and 116-49 |
January 31, 1988 |
| 116-48 and 116-49 |
April 10, 1990 |
| 116-48 and 116-49 |
March 30, 1992 |
| 116-49 |
April 2, 1993 |
| 116-48 |
May 20, 1993 |
| 116-48 and 116-49 |
August 11, 1994 |
| 116-49 |
March 25, 1996 |
| 116-48 |
July 29, 1996 |
| 116-48 and 116-49 |
October 22, 1997 |
| 116-48 and 116-49 |
January 10, 1998 |
| 116-48 and 116-49 |
March 18, 1999 |
The raw and classified imageries were examined and areas with very evident change were selected as subareas. The subareas were visited, local folks were interviewed regarding the history of the area, pictures and GPS points were taken.
Research on policies were done via interview of local government officials, library research and world wide web search.
Results and Discussion
The Macro-analysis of Land Use/Land Cover Change
The result of the land cover classification is presented on Table 2.
Table 2. Area calculation of land cover classes from different years based on Landsat TM atellite imagery classification.
The result of the classification of the 1999 satellite imagery was not included in the analysis because the image was 28% cloud covered. The clouds lowered the result of the area calculation of high-relief areas, which were usually dipterocarp and mossy in their land cover nature. They were not included in the final analysis because they did not reflect the true area of the dipeterocarp and mossy class. Including them would pull down the result of the trend analysis.
The distribution of land use/land cover of Upper Magat Watershed is shown below on Table 2. Majority of the land cover of the watershed were comprised of grasslands followed by dipterocarps, brushes, mossy forests, agricultural areas and barelands.
Table 2 Distribution of Land Use/Land cover in the Study Site
| 1998 Classes
| Percent of
total watershed
area (%)
|
| Built-up
| 0.4
|
| Agriculture
| 7.0
|
| Dipterocarp
| 16.2
|
| Brush
| 10.6
|
| Mossy
| 7.2
|
| Grassland
| 55.6
|
| Bareland
| 2.5
|
| River
| 0.6
|
| Cloud&Shadows
| 0.0
|
Table 3. Regression Analysis for Land Use Classes
| Class
| Equation*
| Adjusted R Squared
|
| Built-up |
A=-130053 + 65.60 Y |
0.65 |
| Agriculture |
A=690378.1 + -338.37 Y |
0.031 |
| Dipterocarp |
A=-1227550 + 631.29 Y |
-0.084 |
| Brush |
A= - 1463588 + 747.83 Y |
-0.137 |
| Mossy |
A= -495758 + 255.62 Y |
-0.042 |
| Grassland |
A= - 3474645 + -1679.03 |
-0.08 |
| Bareland |
A= -593284 + 299.59 Y |
0.26 |
| River |
A= - 30055.5 + 15.48 Y |
-0.163 |
| Cloud&Shadows |
A= 4390.81 + 1.99 Y |
-0.2 |
A* is the predicted area of a class in hectare
Y is the Year
The linear regression analysis (Table 3) was performed, with year as the independent variable, not as a predictive model for each class but a means of previewing the trend of land use/land cover classes with the progression of time.
Except for the built-up class, all of the classes had very low adjusted R 2 . A definite trend on land use/land cover was not established as expected. It was known for a fact that land use change would have a carry over effect for the coming years. For example grassland could not just change to forest in just a span of one year. This meant that the land use/land cover change analysis on a macro level approach was not applicable for the study site. The result of the macro-analysis approach was too general and represented the average dynamics of the watershed. It failed to pinpoint in what part of the watershed is a specific change active. For example, illegal logging can be very active in areas of the watershed that are not too accessible for environmental officers as compared to accessible areas where policies and programs are strictly implemented.
