4.1 LULC Maps for different years (2015–2021)
After the analysis of data and performing all the tasks by using the google earth engine platform and creating the GEE script the Land use/land cover map for consecutive seven years (2015–2021) generated. After that all the generated map exported as a tiff file and displayed in the ArcGIS software and then all the tiff file converted to feature class and the classification with the coverage area obtained as shown in Fig. 4.1 and Table 4.1 respectively. In this specific study all the land covers significantly changed. Forest and agriculture areas abruptly increase when we come to from 2015 to 2021but the grassland and wetland decreases for continues years.
In 2015 the land cover areas for forest, agriculture, grassland and wetland were 8.5 km2, 5.4 km2 32.0 km2 and 5.6 km2 respectively. But in the later years specifically in 2021 the land cover areas for forest, agriculture, grassland and wetland were 18 km2, 20.5 km2, 10.9 km2 and 2.2 km2 respectively. Therefore the forest and agriculture increases by 9.5 km2 and 15.0 km2 indirectly the grassland and the wetland decreases by 21.0 km2 and 3.4 km2. The reasons behind these result were due to population growth that directly increases the demand of agricultural area so that the residents farm the grassland and wetland areas for their survival of their food demand. In other way the farmer’s attitude to day changed to charcoal production for market supply. Some residents said that after the charcoal production they sold and got better money than the normal crops produced with in a year so that in the recent years most farmers change their agriculture areas to forest for charcoal production.
Table 4.1: Land use/land cover area coverage (2015–2021)
Year | 2015 | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 |
Ac (km2) | Forest | 8.5 | 10.7 | 12.3 | 13.7 | 11.8 | 15.6 | 18.0 |
Agriculture | 5.4 | 5.9 | 9.2 | 9.7 | 12.0 | 16.7 | 20.5 |
Grassland | 32.0 | 26.3 | 21.5 | 19.3 | 19.0 | 13.3 | 10.9 |
Wetland | 5.6 | 8.8 | 8.5 | 8.9 | 8.8 | 5.9 | 2.2 |
As shown in the above Fig. 4.2 the grassland area decreases year to year and reaches almost 10 km2 from 32 km2 in the year 2015. These directly implies the grassland areas changed to agriculture areas and the rest to the forest one. In other side the agriculture and the forest areas increase year to year for the production of crop and charcoal that implies population of growth affects the land use/land cover of the study area.
Table 4.2: Land use/Land cover area coverage difference for 2015 and 2021
Year | 2015 (km2) | 2021(km2) | Difference | Remark |
Forest | 8.5 | 18.0 | 9.5 | Increase |
Agriculture | 5.4 | 20.5 | 15.0 | Increase |
Grassland | 32.0 | 10.9 | 21.0 | Decrease |
Wetland | 5.6 | 2.2 | 3.4 | Decrease |
4.2 Accuracy assessment of the Brante watershed Land use/Land cover mapping
Accuracy assessment is an important part of any classification project. It compares the classified image to another data source that is considered to be accurate or ground truth data. Ground truth can be collected in the field; however, this is time consuming and expensive. Ground truth data can also be derived from interpreting high resolution imagery, existed classified imagery or GiS data layers. Therefore for this specific study high resolution imagery with GiS data layers used for accuracy of the image classified. The basic principle for all accuracy assessment is to compare estimates with reality, and to quantify the difference between the two. In the context of remote sensing-based land cover classifications, the ‘estimates’ are the classes mapped for each pixel, and ‘reality’ is the actual land cover in the areas corresponding to each pixel. Once we have produced a land cover (or other) classification from a remote sensing image, an obvious questions is “how accurate is that map?” It is important to answer this question because we want users of the map to have an appropriate amount of confidence in it. If the map is perfect, we want people to know this so they can get the maximum amount of use out of it. And if the map is no more accurate that a random assignment of classes to pixels would have been, we also want people to know that, so they don’t use it for anything. An accuracy assessment of a classified image gives the quality of information that can be obtained from remotely sensed data. Accuracy assessment is performed by comparing a map produced from remotely sensed data with another map obtained from some other source. Landscape often changes rapidly.
In the google earth engine platform it can be used a confusion matrix algorithm to see the exact difference and match of the model and the trusted map obtained from a primary source of data. Once you have created a set of validation data that you trust, you can use their georeferenced to pair them up with the corresponding land cover mapped in the classification. But for this specific study there is no trusted data for comparison of the model result with the corresponding land cover mapped classification. Therefore we tried to perform accuracy assessment with another option by using Arc GIS with higher resolution map of google earth. The kappa statistic is a measure of how closely the instances classified by the machine learning classifier matched the data labeled as ground truth, controlling for the accuracy of a random classifier as measured by the expected accuracy.
Table 4.3: Accuracy assessment for land cover map using google earth and GIS for 2015
Land class | User accuracy | Producers accuracy | Overall accuracy | Kappa coefficient | Range value | Decision |
Forest | 90 | 75 | 90 | 87 | UA > 70 | The model performs well. |
Agriculture | 100 | 83 | PA > 70 |
Grassland | 80 | 100 | OA > 85 |
Wetland | 90 | 100 | K > 75 |
Table 4.4
Accuracy assessment for land cover map using google earth and GIS for 2021
Land class | User accuracy | Producers accuracy | Overall accuracy | Kappa coefficient | Range value | Decision |
Forest | 100 | 91 | 95 | 93 | > 70 | It performs well. |
Agriculture | 100 | 91 | > 70 |
Grassland | 90 | 100 | OA > 85 |
Wetland | 90 | 100 | K > 75 |
Overall accuracy = Total number of correctly classified pixels (Diagonal)/total number of reference pixels * 100
User Accuracy = Number of correctly classified pixels in each category/total number of classified pixels in that category (the Row total)
Producer Accuracy = Number of correctly classified pixels in each category/total number of Reference pixels in that category (the Column total)
Kappa coefficient (T) = (TS*TCS)-\(\sum \left(Columon total*Row total\right)\)/TS2-\(\sum \left(Columon total*Row total\right)\)
After classification of images, producer's accuracy, user's accuracy, overall accuracy and kappa coefficient values have been calculated with the help of confusion / error matrix. For most image classification the accuracy assessment result was good but still requires further validation. For instance for the year 2021, the overall accuracy and Kappa coefficient obtained 90% and 87% respectively and for the year 2021 the same analysis did and the overall accuracy obtained 95% and the Kappa coefficient 83% obtained. So that still it can be adjusted and further analysis can be held and obtained better result by collecting many ground control points for each land use/land cover types as much as possible.