Examining deforestation and agropasture dynamics along the Brazilian TransAmazon Highway using multitemporal Landsat imagery

Figures & data

Figure 1. Study area: Altamira TransAmazon Highway region, Pará State, Brazil, and Xingu River, highlighting Belo Monte Dam.

Table 1. Datasets used in research.

Dataset	Sources and dates
Landsat images	(1) Landsat 5 TM on 27 July 2011 (2) Landsat 8 OLI image on 20 August 2014 (3) Landsat 8 OLI images on June 16 and 11 July 2017 Spectral bands of blue, green, red, near infrared, and two short wavelength infrared from Landsat TM and OLI were used to develop land-cover maps
Land-cover images	Land-cover classification results were developed from Landsat TM images in 1991, 2000 and 2008 and were directly used in this research. The detailed descriptions of these classification results are provided in Lu et al. (2013).
SPOT 6	Satellite Pour l’Observation de la Terre (SPOT 6) imagery on 19 August 2015, was used for collection of land-cover samples.
Fieldwork	Different land-cover data were collected in January–May 2015
Roads	The roads were digitized from Google Earth images
Population GDP Herds of cattle	Data downloaded from Instituto Brasileiro de Geografia e Estatística for different years

Note: GDP, gross domestic product; TM, Thematic Mapper; OLI, Operational Land Imager

Figure 2. The modified strategy for mapping land-cover distribution from Landsat imagery using the hierarchical-based approach (Note: TM, Thematic Mapper; NDVI, normalized difference vegetation index; MNDWI, modification of normalized difference water index; SF, secondary forest; AP, agropasture).

Table 2. Accuracy assessment result of the 2014 land-cover classification imagery.

Type	Reference data					RT	CT	UA	PA	Accuracy
	PF	SF	AP	UR	WA
PF	48	10				58	54	82.8	88.9	Overall accuracy: 83.4%; Kappa coefficient: 0.78
SF	4	54	11			69	76	78.3	71.1
AP		9	70			79	86	88.6	81.4
UR		1	5	24		30	24	80.0	100.0
WA	2	2			26	30	26	86.7	100.0

Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water; RT, row total; CT, column total; UA, user’s accuracy; PA, producer’s accuracy.

Figure 3. Land-cover distributions at different years between 1991 and 2017.

Figure 4. Proportion of land-cover types over time: a comparison of land-cover trends between 1991 and 2017 at nine-year intervals (solid lines) and between 2008 and 2017 at three-year intervals (dashed lines) (Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water; percent of a land cover type = (A_i/A)*100, where A_i is the total area of the ith land-cover type and A is the total study area).

Table 3. Summary of unchanged land-cover types between 1991 and 2017 and the land-cover types that changed at least once based on the 1991 land-cover data.

Type	Area in 1991 (ha)	Unchanged land covers between 1991 and 2017		Land covers that changed at least once based on the 1991 data
		Area (ha)	Percent	Area (ha)	Percent
PF	470,925	241,368	51.3	229,557	48.7
SF	94,162	30,888	32.8	63,274	67.2
AP	82,491	33,766	40.9	48,725	59.1
UR	1,788	1,773	99.2	15	0.8
WA	1,280	416	32.5	864	67.5
Total	650,645	308,211	47.4	342,434	52.6

Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water

Figure 5. Annual land-cover change (ha/yr) at different change detection periods (nine-year and three-year intervals) (Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water; Annual land-cover change (ha/yr) = [A_i(t2) – Ai(t1)]/(t2-t1), where A_i is the total area of the ith land cover type, A is the total area, and t1 and t2 are prior and posterior years).

Figure 6. Land-cover change trajectories at different detection periods (nine-year and three-year intervals) (Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban city; WA, water).

Figure 7. Comparison of land-cover changes at different detection periods (a: primary forest loss; b: urban gain; c: secondary forest loss and gain; d: agropasture loss and gain) (Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water).

Figure 8. Individual land-cover changes over time (a: nine-year detection interval; b: three-year detection interval) (Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water).

Table 4. Land-cover change trajectories over the study region at different detection periods.

Land-cover change trajectory		Changed area (ha) at nine-year intervals			Changed area (ha) at three–year intervals			Total (2008–2017)
		1991–2000	2000–2008	2008–2017	2008–2011	2011–2014	2014–2017
PF loss	PF–SF	45,513	36,096	35,271	14,208	5,940	10,638	30,786
	PF–AP	50,508	42,296	17,694	9,087	6,312	6,975	22,374
	PF–UR/WA	405	1,480	747	237	111	243	591
	Total	96,426	79,880	53,649	23,535	12,360	17,853	53,748
SF loss	SF–AP	32,463	61,408	37,458	34,686	17,988	3,684	56,358
	SF–UR/WA	279	944	738	141	264	183	588
	Total loss	32,742	62,352	38,196	34,827	18,252	3,867	56,946
SF gain	PF–SF	45,513	36,096	35,217	14,208	5,940	10,638	30,786
	AP–SF	24,057	28,408	70,893	22,851	11,796	59,409	94,056
	WA–SF	63	16	9	0	15	3	18
	Total gain	69,633	64,528	106,119	37,059	17,748	70,047	124,854
AP loss	AP–SF	24,057	28,408	70,893	22,851	11,796	59,409	94,056
	AP–UR/WA	1134	3192	4077	1,374	1638	1,560	4,572
	Total loss	25,191	31,608	74,961	24,225	13,434	60,969	98,628
AP gain	PF–AP	50,508	42,296	17,694	9,087	6,312	6,975	22,374
	SF–AP	32,463	61,408	37,458	34,686	17,988	3,684	56,358
	WA–AP	135	104	36	9	87	3	99
	Total gain	83,106	103,808	55,179	43,782	24,390	10,665	78,837
UR gain	PF–UR	207	528	423	114	63	159	336
	SF–UR	252	776	648	141	255	51	447
	AP–UR	972	2,768	3,447	1,371	1,638	900	3,909
	WA–UR	9	8	306	36	132	9	177
	Total gain	1,431	4,072	4,824	1,662	2,088	1,119	4,869
UR loss	UR–WA	0	0	36	0	0	75	75
Others		540	96	18	0	114	15	129

Note: PF, primary forest; SF, secondary forest; AP, agropasture; UR, urban; WA, water

Figure 9. Intermediate land-cover dynamic changes detected within three-year intervals between 2008 and 2017 (Note: PF, primary forest; SF, secondary forest; AP, agropasture).

Figure 10. Impacts of population, economic growth, and cattle on major land-cover changes (a: impacts of urbanization on deforestation and agropasture expansion; b: impact of cattle herds on agropasture area; c: impacts of population on deforestation and agropasture expansion; d: impacts of gross domestic product on deforestation and agropasture expansion) (Note: PF, primary forest; AP, agropasture; UR, urban; GDP, gross domestic product).

Lu, D., G. Li, E. Moran, and S. Hetrick. 2013. “Spatiotemporal Analysis of Land-Use and Land-Cover Change in the Brazilian Amazon.” International Journal of Remote Sensing 34 (16): 5953–5978. doi:10.1080/01431161.2013.802825.

 PubMed Web of Science ®Google Scholar