1
00:00:00,020 --> 00:00:04,030
 

2
00:00:04,050 --> 00:00:08,090
We've produced a 30-meter global product of forest loss and

3
00:00:08,110 --> 00:00:12,130
gain on a backdrop of tree cover density.

4
00:00:12,150 --> 00:00:16,170


5
00:00:16,190 --> 00:00:20,210
The basic product looks like this

6
00:00:20,230 --> 00:00:24,240
and it is a percent tree cover layer from 2000,

7
00:00:24,260 --> 00:00:28,060
and then on top of that, forest cover loss and gain.

8
00:00:28,080 --> 00:00:32,290
We have trees as green scale color, so density

9
00:00:32,310 --> 00:00:36,330
is saturated. That's 100% tree cover, you go down to Chaco

10
00:00:36,350 --> 00:00:43,000
and you see darker shades of green, that's 505 tree cover. So this is a percent tree cover layer for 2000.

11
00:00:43,020 --> 00:00:44,470


12
00:00:44,490 --> 00:00:48,520
Probably the most intensively used forest landscape is found in the southeast United States.

13
00:00:48,540 --> 00:00:52,570
And in this product you can see all of the

14
00:00:52,590 --> 00:00:56,610
reds, blues, and magentas that are indicative of forest disturbance

15
00:00:56,630 --> 00:01:00,650
and recovery. And you see some really intense

16
00:01:00,670 --> 00:01:04,680
intense land uses.

17
00:01:04,700 --> 00:01:08,710
Out of this ecozone,

18
00:01:08,730 --> 00:01:12,730
in the southeast US, 30% of forest land

19
00:01:12,750 --> 00:01:16,780
either was regrown or lost during this period, which is 12 years, it's incredible.

20
00:01:16,800 --> 00:01:20,840
Really, trees are as crops here, you might want to re-think a definition of forest

21
00:01:20,860 --> 00:01:24,900
it's a different thing, it's not really natural.

22
00:01:24,920 --> 00:01:28,950
In the picture here we have greens, meaning the forest didn't change in the last 12 years

23
00:01:28,970 --> 00:01:32,990
you can see there's something to do with the watershed protection around a reservoir here.

24
00:01:33,010 --> 00:01:37,050
Everywhere else, the greens are stable, and the

25
00:01:37,070 --> 00:01:41,090
blacks are non-forest and then the dynamic is

26
00:01:41,110 --> 00:01:45,120
red being loss, blue being gain, and these magentas being both, during the 12 year period.

27
00:01:45,140 --> 00:01:49,150
Brazil, in the last decade, has cut their deforestation rate

28
00:01:49,170 --> 00:01:53,160
in half. Despite that decrease in Brazil's deforestation rate

29
00:01:53,180 --> 00:01:57,240
the tropics has a whole have a statistically significant increase

30
00:01:57,260 --> 00:02:02,990
and that is due to increasing rates of loss in Malaysia, Indonesia, Angola, Peru, Paraguay,

31
00:02:03,010 --> 00:02:07,990
all the other countries in our study are making up for the loss in Brazil.

32
00:02:08,010 --> 00:02:13,400
There's three things that changed in the recent past that allowed us to do a global scale Landsat,

33
00:02:13,420 --> 00:02:17,430
which is 30-meter, characterization of the land surface.

34
00:02:17,450 --> 00:02:21,460
First is, the last Landsat sensor, ETM+ on the Landsat 7 satellite,

35
00:02:21,480 --> 00:02:25,490
had a global acquisition strategy.

36
00:02:25,510 --> 00:02:29,510
So we had observations everywhere. But it had a cost model associated with it,

37
00:02:29,530 --> 00:02:33,570
so you had to buy data. We always said that we would use the data we could afford,

38
00:02:33,590 --> 00:02:37,630
not what we really needed. And you were stuck,

39
00:02:37,650 --> 00:02:41,680
you couldn't do large area, large depth time series with Landsat.

40
00:02:41,700 --> 00:02:45,720
So what happened in 2008, they opened up the archive for free access.

41
00:02:45,740 --> 00:02:49,760
 So we didn't even have to ask what we needed, we could use it all.

42
00:02:49,780 --> 00:02:53,790
We started thinking, let's try and mine the archive systematically.

43
00:02:53,810 --> 00:02:57,830
If we did this project on one CPU, it would have taken 15 years.

44
00:02:57,850 --> 00:03:01,850
but if we do it in the cloud, it's a matter of days.

45
00:03:01,870 --> 00:03:05,870
That's the three things: the global acquisition strategy, free data, and cloud computing 

46
00:03:05,890 --> 00:03:09,940
equals the ability to do this.

47
00:03:09,960 --> 00:03:14,000
And what we like about it is if we're working at 30-meters globally, our history has been to work at global scale,

48
00:03:14,020 --> 00:03:18,060
and you get a globally consistent product and you can say what's happened

49
00:03:18,080 --> 00:03:23,860
to the earth in its entirety. But with 30-meter data we can cut out any particular place, and it should be locally relevant.

50
00:03:23,880 --> 00:03:31,210
So we have a globally consistent and locally relevant product.

51
00:03:31,230 --> 00:03:34,240
[music]

52
00:03:34,260 --> 00:03:38,270
[music]

53
00:03:38,290 --> 00:03:44,004
[beep]