Popular in Engineering

Electronic Engineering

Control of a Wheelchair-Mounted Robotic Arm System

University of South Florida Redwan Alqasemi

Engineering

Spatiotemporal evaporating droplet dynamics on fomites enhances long term bacterial pathogenesis

Indian Institute of Science Saptarshi Basu

Civil Engineering

Suitability of excavated London clay as a supplementary cementitious material

ARUP Fragkoulis Kanavaris

Chemistry

Polyfluorene-Wrapped Carbon Nanotubes Decorated with Photocleavable Side-Chains

McMaster University Alex Adronov

Electronic Engineering

A Trusted and Privacy-Preserving Internet of Mobile Energy

Queensland University of Technology

Engineering

Water Security: Why We Need Global Solutions

Cardiff University

Engineering

Decision-Making Principles for Better Software Design Decisions

Swinburne University of Technology

Prediction of Impending Drought Scenarios Based on Surface and Subsurface Parameters in a Selected Region of Tropical Queensland, Australia

View more insights in Professional - Engineering - Civil Engineering

View more figures from James Cook University

View the page of Bithin Datta

Read the Study

Next in Civil Engineering

Least cost pathways to a low carbon electricity system for Australia

Roger Dargaville | Monash University

Towards Sustainable Construction

Xiancun Hu | University of Canberra

Droughts occur in all climatic regions around the world costing a large expense to global economies. Reasonably accurate prediction of drought events helps water managers proper planning for utilization of limited water resources and distribution of available waters to different sectors and avoid catastrophic consequences.

Image courtesy of interviewee. May 27, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video video discusses the impact of a severe drought in Northern Queensland during 2013-2014, particularly its effects on water supply in Townsville and surrounding agricultural areas known for sugar cane farming. It highlights challenges in predicting drought conditions and explores drought indices such as the Palmer Drought Index and Standardized Precipitation Index, emphasizing their limitations in local-scale applications.

The speaker, a civil engineer with an interest in signal processing, proposes using artificial neural networks (ANN) for data analysis to identify relevant parameters that influence drought conditions based on past meteorological data and groundwater levels. The goal is to predict droughts three to six months in advance to allow for better resource management in urban and agricultural settings.

The study’s findings indicate that certain parameters, particularly soil moisture levels, significantly affect drought predictions. Results show improvements in prediction accuracy for groundwater levels and dam levels, although operational policies may influence dam level outputs. The speaker is satisfied with the initial findings but suggests that further research could enhance the model and expand its applications in various geographic and climatic conditions, including agriculture and mining sectors. Overall, the concept of using ANN for better drought prediction is presented as a promising avenue for future research.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.015 --> 00:02.685
There was a severe drought in Northern Queensland,

00:03.235 --> 00:06.765
sometimes in 2013, 2014.

00:07.385 --> 00:10.765
And there was a severe effect on the ban water supply,

00:11.265 --> 00:14.445
the water availability for, the Townsville city.

00:14.915 --> 00:19.325
Then also this a very large agricultural area nearby Baran

00:19.325 --> 00:21.765
area where, you know, a lot of sugar cane farming.

00:21.785 --> 00:23.245
It is international known for that.

00:23.745 --> 00:25.045
And that requires a lot of water.

00:26.065 --> 00:28.925
And there, there of course,

00:29.335 --> 00:33.805
there is related issues of s increase in groundwater.

00:34.365 --> 00:35.605
'cause groundwater availability is also

00:35.605 --> 00:37.005
because some of the, most of the water

00:37.265 --> 00:39.565
for the agriculture use was coming from down water.

00:40.345 --> 00:43.565
And, and, and the real problem with droughts is

00:43.565 --> 00:46.365
that you don't really know you are into a drought,

00:46.435 --> 00:48.085
till you are already in a drought.

00:48.705 --> 00:50.765
So, so, so that, that is the,

00:50.765 --> 00:53.125
and then I looked at some of those, you know, models

00:53.665 --> 00:56.245
and this models were like, some, some

00:56.245 --> 00:59.005
of the index indices, you know, drought indices like,

00:59.265 --> 01:00.965
Palmer, drought index.

01:01.225 --> 01:02.565
Uh, there is something called,

01:02.805 --> 01:04.325
standardized precipitation index.

01:04.985 --> 01:07.005
And they look at a slightly larger scale.

01:07.625 --> 01:10.245
And you know, and also they're basically related

01:10.265 --> 01:13.845
to precipitation, that the rainfall and the try to model

01:13.865 --> 01:17.285
and, you know, the, the British, the rainfall levels

01:17.385 --> 01:18.605
or precipitation levels

01:19.185 --> 01:21.005
and like Palmer, the drought index

01:21.525 --> 01:25.005
includes some soil moisture, information also.

01:25.005 --> 01:26.965
But basically they're looking at a larger picture

01:27.505 --> 01:31.365
and not specifically trying to, predict if what happens

01:31.465 --> 01:32.525
to maybe the dam levels

01:33.485 --> 01:36.105
and how much of where the groundwater level will be.

01:37.045 --> 01:38.265
So, and this, this,

01:38.265 --> 01:40.985
all these indices al statistical analysis,

01:41.565 --> 01:43.545
and they look at the distribution, they try

01:43.545 --> 01:45.305
to fit a probability distribution

01:45.565 --> 01:48.025
and then they try to say, okay, what is the mean value

01:48.025 --> 01:49.985
or the actually standardized it.

01:50.295 --> 01:52.545
Some, some called tric, non-parametric mixture.

01:53.005 --> 01:54.545
And I thought that was not what we wanted.

01:54.605 --> 01:57.705
We wanted to see, whether you can predict in some way

01:57.705 --> 01:59.345
that what is happening in a smaller scale,

01:59.795 --> 02:01.545
maybe the city scale or regional scale.

02:02.165 --> 02:04.225
And then of course I had been working with,

02:04.565 --> 02:05.865
for some time signal processing.

02:05.865 --> 02:06.865
And although I'm a civil engineer,

02:06.965 --> 02:10.225
but I did work for a long time when I, I was in US

02:10.285 --> 02:11.305
for a long time also,

02:11.665 --> 02:13.785
I was very interested in the signal processing

02:14.085 --> 02:16.105
and its application because, so I thought

02:16.105 --> 02:17.785
that there is some signals in the nature

02:18.755 --> 02:20.065
which we might be able

02:20.065 --> 02:23.625
to recognize something from the signals that based on that,

02:24.175 --> 02:25.265
that the drought is coming

02:25.645 --> 02:28.105
and we might not know it, we may not physically know it,

02:28.165 --> 02:31.745
but then maybe this hidden the data and,

02:31.925 --> 02:34.265
and the question is, what type of data then?

02:34.415 --> 02:36.345
Then I thought maybe what can be relevant data?

02:36.345 --> 02:37.545
And then we did not know really.

02:37.685 --> 02:41.385
So what will be, so we know the drought is occurred, so,

02:41.525 --> 02:42.705
or maybe in the past also

02:42.705 --> 02:43.865
there are some records of droughts.

02:43.885 --> 02:45.305
The meteorological data is there

02:45.525 --> 02:48.585
and also the, the city council have got data on the dam

02:48.585 --> 02:49.665
levels, thing.

02:49.685 --> 02:53.315
So, so I thought, let us look at that.

02:53.465 --> 02:56.875
What are the parameters which affect this in the

02:56.875 --> 02:58.795
past has

02:59.085 --> 03:00.085
Affected the drought condition. So

03:00.085 --> 03:02.005
there, or there was something in the data that, you know,

03:02.005 --> 03:03.085
from which I can recognize

03:03.085 --> 03:05.765
or we can recognize, that actually is hidden,

03:05.865 --> 03:07.445
but we cannot see it for, you know,

03:07.505 --> 03:11.325
we cannot analyze ourselves humanly, but maybe lot of data

03:11.385 --> 03:12.925
and digging and our thing data mining

03:13.025 --> 03:16.365
and whether we can actually assert these patterns.

03:16.865 --> 03:20.285
And then I thought of course artificial neural network is,

03:20.665 --> 03:22.445
has been used quite a bit here

03:22.445 --> 03:25.445
and there it has been, there are improved portion of it,

03:25.625 --> 03:26.725
but that is one of the tools.

03:26.945 --> 03:30.885
So, but, but the idea is to extract as much information

03:30.985 --> 03:32.285
as we can from this data

03:32.585 --> 03:35.285
and then find out which parameters actually

03:35.285 --> 03:36.725
influence the drought conditions.

03:37.065 --> 03:40.085
But that we learned from the past that, you know, so,

03:40.385 --> 03:43.285
and then we are trying to various, various parameters,

03:43.285 --> 03:44.285
you know, and see that,

03:44.915 --> 03:46.485
that whether they're hidden in the data

03:46.505 --> 03:48.965
and then it could have in the past it could have

03:49.325 --> 03:50.685
actually predicted the drought.

03:50.795 --> 03:52.085
Okay. The also, I'll say

03:52.085 --> 03:54.445
that when I look at the literature there, I,

03:54.485 --> 03:57.085
I can mention some of the, some of the publications, others,

03:57.145 --> 03:59.445
and they looked at the, standardized,

03:59.645 --> 04:01.565
precipitation index, palmers index

04:01.985 --> 04:03.925
and also their various drought indexes.

04:04.465 --> 04:07.925
So, and there some are useful, some are not useful,

04:08.165 --> 04:10.205
somewhat accurate predictors not so,

04:10.745 --> 04:12.165
but then as a, that is a, that's

04:12.895 --> 04:16.325
takes a slightly more larger scale picture on a global scale

04:16.905 --> 04:18.925
and maybe the the,

04:19.015 --> 04:21.325
maybe the global circulation model might be a part of it.

04:22.025 --> 04:24.445
The southern OS oscillation index might be a part of it

04:25.065 --> 04:27.885
and it is maybe EL effects and so,

04:27.885 --> 04:29.765
and other things might be a part of it.

04:30.665 --> 04:32.125
I'm not saying that you can ignore them,

04:32.225 --> 04:34.445
but then they are slightly looking at a larger scale.

04:34.785 --> 04:37.325
So I, I thought that maybe first thing is that

04:37.885 --> 04:40.205
whatever data we have from the precipitation records

04:40.505 --> 04:42.125
and maybe the soil moisture levels

04:43.675 --> 04:45.445
localized in a very localized scale,

04:45.445 --> 04:47.685
which is available from maybe I think,

04:47.685 --> 04:50.325
meteorological, you know, grew

04:51.025 --> 04:52.525
and also geoscience in Australia.

04:52.985 --> 04:55.125
And some of this of course from the city council data

04:55.145 --> 04:57.605
for the dams, whether we can use this data

04:57.705 --> 05:01.725
and see what has happened in the past, how

05:02.805 --> 05:05.875
maybe everything might not be important.

05:06.165 --> 05:08.435
Maybe the precipitation six months ago might not be

05:08.435 --> 05:10.795
important, but maybe last three months might be important.

05:10.965 --> 05:13.675
Maybe the soil, but soil moisture level is not

05:13.675 --> 05:15.275
that important because that is something maybe

05:15.295 --> 05:16.395
result and not a cause.

05:16.495 --> 05:18.275
And then so, so those type of things.

05:18.535 --> 05:21.275
And of course in addition we also said maybe the sudden

05:21.555 --> 05:23.755
oxidation the index, I can include that.

05:23.945 --> 05:25.995
What is the predicted value or the existing value

05:26.255 --> 05:28.395
or in so values or something.

05:28.395 --> 05:32.035
These are some of the parameters whether we can use this all

05:32.285 --> 05:33.435
input, all this information.

05:33.535 --> 05:36.795
See in the past from the record in the past whether we could

05:36.795 --> 05:39.235
have actually predicted the TR drought conditions,

05:39.695 --> 05:42.195
but with the leak time of maybe three months to six months,

05:42.595 --> 05:44.115
I knew that we, we are doing it

05:44.115 --> 05:46.715
for more than six months probably it erroneous.

05:46.935 --> 05:49.395
But if you can know if drought is coming even six months

05:49.865 --> 05:52.195
earlier, you can plan for a lot of things

05:52.195 --> 05:55.435
because there a, there is is a supply side

05:55.495 --> 05:58.205
and also there's a, you know, demand site.

05:58.545 --> 06:00.765
So we can manage a demand site if you can predict it

06:00.765 --> 06:02.365
earlier, especially in urban settings,

06:02.365 --> 06:03.365
in agriculture settings.

06:03.625 --> 06:05.245
So I thought that that is very useful.

06:05.425 --> 06:06.805
But surprisingly what we found,

06:06.825 --> 06:10.205
and I actually used one of some of our B honor students

06:10.225 --> 06:11.605
and I things, so let us look at as a,

06:12.465 --> 06:15.845
I'm aware the some some application of A N was there

06:15.845 --> 06:19.285
before, but there, you know, so I think Desai

06:19.285 --> 06:20.605
and Misra, Misra and DE

06:20.665 --> 06:23.565
and all this environmental ecology, ecology modeling, some,

06:23.565 --> 06:25.245
some, some of the work has been done,

06:25.665 --> 06:28.805
but they're basically looking at reproducing the time series

06:29.065 --> 06:31.205
of the, some of this

06:31.635 --> 06:34.275
indices using this.

06:34.275 --> 06:35.675
So basically the learning from the data

06:35.695 --> 06:38.675
by looking at the time series of, you know, maybe SPI

06:38.675 --> 06:40.195
or something, a standardized

06:40.435 --> 06:41.715
specification index or something like that.

06:41.855 --> 06:45.155
But that was not what we wanted though it might look, appear

06:45.155 --> 06:46.195
to a similar in concept.

06:46.295 --> 06:48.675
But this was saying we look at, because Aiden

06:48.675 --> 06:50.755
and other things are other this you know,

06:51.315 --> 06:52.995
artificial email networks, they can actually have the

06:53.195 --> 06:54.275
ability to find out

06:54.275 --> 06:55.875
what are the parameters which are relevant,

06:55.875 --> 06:57.635
which are not relevant, what is important

06:57.655 --> 06:58.675
and what is not important.

06:58.935 --> 07:00.395
And then probably that based on

07:00.395 --> 07:01.955
that we can learn from the past

07:02.215 --> 07:03.515
and then use that in the future.

07:03.705 --> 07:07.035
That was the whole idea. And we found the soil master was a

07:07.035 --> 07:09.555
very surprising thing that once you into into the soil

07:09.555 --> 07:14.035
master, you know, palace from you know, the metallurgy

07:14.095 --> 07:16.955
and meteorology, the prediction was very good

07:17.375 --> 07:19.115
and we are surprised to see that that is

07:19.115 --> 07:20.315
what affected quite a bit

07:20.315 --> 07:22.995
because that was something hidden something in that data

07:23.415 --> 07:25.835
and that that gig lot of, lot of, so that is,

07:25.935 --> 07:29.235
so it was very successful in in my opinion,

07:29.415 --> 07:32.075
so in predicting not only the soil moisture level

07:32.095 --> 07:35.715
but also SPI also is we try to use this, this input,

07:35.775 --> 07:39.115
so predict the SPI also this this standardized per

07:40.035 --> 07:42.715
precipitation index and also the soil MAs level.

07:42.785 --> 07:45.715
Then you went to groundwater levels, you went to the,

07:45.715 --> 07:47.475
even the solidity because in coastal, like

07:47.475 --> 07:48.635
for solidity is important.

07:48.775 --> 07:50.395
If the groundwater uses more,

07:51.015 --> 07:53.395
the gradients comes towards from the sea to the,

07:53.495 --> 07:54.675
and the sun increases

07:55.135 --> 07:57.075
and based on the data whether we can fit the data.

07:57.575 --> 08:01.615
And then if you're successful, then also we have

08:01.615 --> 08:02.695
to look at the output.

08:02.695 --> 08:04.015
Also another output was the DAM level.

08:04.015 --> 08:06.535
So there's a dam nearby which is also dam features

08:06.535 --> 08:07.575
like tedious conditioning.

08:07.575 --> 08:11.695
The droughts is like 15% or 10%, which is no water to supply

08:12.235 --> 08:15.815
and whether, whether you can predict the dam level also.

08:16.395 --> 08:17.495
So it is the past data

08:17.595 --> 08:20.215
to check whether it is actually able to do it.

08:20.475 --> 08:22.015
But we did not assume that we assure

08:22.015 --> 08:24.215
that maybe we have the data for even when we're checking the

08:24.215 --> 08:25.695
past data, we check

08:25.695 --> 08:28.455
that we have the data only from six months below before that

08:29.665 --> 08:31.925
and you know then whether you can predict with the leak time

08:31.925 --> 08:33.285
of six months or three months, whatever.

08:33.285 --> 08:35.725
And again, I use the data which is the lead time

08:35.725 --> 08:36.925
of six months or three months

08:37.225 --> 08:38.725
and then see whether it can be predicted.

08:38.725 --> 08:40.765
So it is not that we are, you are learning from the data,

08:40.765 --> 08:43.125
but we are checking in a the real time, okay,

08:43.225 --> 08:47.525
the major findings is that it this the a n

08:47.625 --> 08:50.165
and also maybe it's improvement improved, you know, some

08:50.165 --> 08:51.885
of the also because now a lot

08:51.885 --> 08:53.965
of advancements are going on is able

08:53.965 --> 08:55.845
to extract the information, the pattern

08:55.875 --> 08:56.885
that exists in the nature.

08:56.995 --> 08:59.725
That is what I feel though it is in a digital form.

08:59.785 --> 09:02.405
So it can actually detect some

09:02.405 --> 09:03.445
of the patterns which might be

09:03.445 --> 09:04.565
useful in predicting the future.

09:05.225 --> 09:08.725
And that is what and ally and you know, in terms of numbers

09:08.825 --> 09:10.685
and these are the, these are hidden patterns that there

09:11.345 --> 09:14.325
and what we cannot do as human beings though it is,

09:14.325 --> 09:15.805
you know, we can use intuition or something,

09:16.265 --> 09:18.605
but a n can do that in much more effective way.

09:18.785 --> 09:21.165
But a n is not the only, I mean it can be something else

09:21.165 --> 09:23.365
because the concept come is a signal coming

09:23.625 --> 09:26.725
and what, what are the inputs that affecting the signals

09:27.065 --> 09:28.405
and how they actually modified

09:28.545 --> 09:30.525
and how do you can predict what will be the next.

09:30.905 --> 09:34.445
So we looked at lead time of three months, six months,

09:34.445 --> 09:35.885
six months, six months and see what they tested.

09:36.265 --> 09:38.045
And then with the independent set of data,

09:38.105 --> 09:40.285
we said whether it is actually able to predict the drought.

09:40.355 --> 09:43.085
That is what it is. And not only that one, we are very,

09:43.195 --> 09:47.645
very moral or is very satisfactory prediction

09:47.645 --> 09:50.085
of groundwater levels, which is very important.

09:50.645 --> 09:54.005
A you went to the groundwater ty levels

09:54.665 --> 09:58.965
during the droughts and suddenly SPI was very good.

09:59.105 --> 10:01.685
So all this, but we also use sudden O observation index

10:01.685 --> 10:03.165
and other things are many factors,

10:03.505 --> 10:07.125
but also you, we tried to break the SPI that is the,

10:07.545 --> 10:10.125
the the standardized precipitation index which looks,

10:10.125 --> 10:12.645
looks at the rainfall statistic, but I use that.

10:12.645 --> 10:14.485
Then once we have break that we use that

10:14.485 --> 10:16.805
as a input to these models.

10:16.955 --> 10:20.525
Okay. So that information was actually, actually created

10:20.665 --> 10:22.165
and then use input and so

10:22.165 --> 10:24.365
therefore our prediction in my, in very,

10:24.365 --> 10:27.125
very good better than anything that we have body,

10:27.345 --> 10:28.485
but also the dam levels.

10:28.865 --> 10:30.405
But again, the dam level prediction is

10:30.405 --> 10:32.285
that some operation policy comes with the picture.

10:32.945 --> 10:36.685
So, but there is a black box element there

10:36.685 --> 10:38.125
because the operation policy

10:38.125 --> 10:40.125
otherwise might be actually distorting it somewhere,

10:40.545 --> 10:44.485
but you droughts in other where basically you try to

10:44.485 --> 10:45.765
release how much is available.

10:46.585 --> 10:49.765
So, so they're kind of, you know, related and quite close.

10:50.225 --> 10:52.085
So therefore we are able to actually pay the,

10:52.085 --> 10:53.445
the dime levels also.

10:53.785 --> 10:55.805
So the whole idea is that it'll improve

10:55.825 --> 10:58.565
and also we can use it in some other regions also.

10:58.565 --> 10:59.645
We can develop it for other region,

10:59.665 --> 11:00.725
not only in these localities,

11:00.725 --> 11:02.645
some are similar ones, but we have to study that.

11:03.025 --> 11:05.165
And also in mining areas we can use it for,

11:05.165 --> 11:06.205
because mine has a lot of,

11:06.205 --> 11:07.565
because they can actually, you can

11:07.665 --> 11:09.565
and some industries which they can predict

11:09.745 --> 11:11.005
how much water might be available

11:11.465 --> 11:12.765
or there might drought coming in.

11:12.845 --> 11:14.205
I think it's very, very, some

11:14.205 --> 11:16.205
of these mining people also picked up some of these,

11:16.745 --> 11:19.005
you know, work and then they thought it's very useful.

11:19.345 --> 11:21.485
But I am very satisfied with the results.

11:21.905 --> 11:25.885
Oh this our, in my opinion what we did was preliminary,

11:26.925 --> 11:28.505
we have have the proof of concept.

11:28.935 --> 11:29.985
That is what we've developed.

11:29.985 --> 11:31.545
So this application with proof of concept

11:31.685 --> 11:35.105
and I think that is satisfactory, quite satisfactory.

11:35.125 --> 11:37.665
It can do the job but it has to be extended

11:38.285 --> 11:41.065
and maybe some other parameters have to, you know, utilized,

11:41.065 --> 11:43.265
maybe it should be applied in some other regions

11:43.265 --> 11:44.505
with other geographic regions,

11:44.505 --> 11:46.745
with other climatic climatic conditions.

11:47.525 --> 11:50.745
And also basically I want to see an application where the,

11:50.805 --> 11:52.825
for agriculture, you know how to plan

11:52.825 --> 11:54.865
The irrigation when you have some

11:54.965 --> 11:56.945
or in urban areas,

11:57.205 --> 11:59.945
how do you risk put the restrictions earlier so

11:59.945 --> 12:01.425
that you can control the demand sites.

12:02.045 --> 12:03.825
But even, even, you know, so mining

12:03.845 --> 12:04.945
and is of course this is,

12:05.005 --> 12:07.105
and there even hydrogen, you know,

12:07.345 --> 12:08.585
nowadays energy production is there.

12:08.585 --> 12:10.505
So a lot of water demand

12:10.965 --> 12:12.865
and we need to be actually affected.

12:12.885 --> 12:14.785
If your drought is impending drought is there.

12:15.285 --> 12:19.385
So, so I think there are very various applications possible,

12:19.765 --> 12:21.225
but the study is not complete.

12:21.685 --> 12:23.425
Uh, of course I want to carry it forward.

12:23.905 --> 12:25.065
I hope I get a chance to do that.

Predicting Drought Impacts in Queensland Agriculture

00:00:00 - There was a severe drought in Northern Queensland, sometimes in 2013, 2014. And there was a severe effect on the ban water supply, the water availability for, the Townsville city. Then also this a very large agricultural area nearby Baran area where, you know, a lot of sugar cane farming. It is international known for that. And that requires a lot of water. And there, there of course, there is related issues of s increase in groundwater.
00:00:35 - because some of the, most of the water for the agriculture use was coming from down water. And, and, and the real problem with droughts is that you don't really know you are into a drought, till you are already in a drought. So, so, so that, that is the, and then I looked at some of those, you know, models and this models were like, some, some of the index indices, you know, drought indices like, Palmer, drought index. there is something called, standardized precipitation index.
00:01:07 - And you know, and also they're basically related to precipitation, that the rainfall and the try to model and, you know, the, the British, the rainfall levels or precipitation levels and like Palmer, the drought index includes some soil moisture, information also. But basically they're looking at a larger picture and not specifically trying to, predict if what happens to maybe the dam levels and how much of where the groundwater level will be.
00:01:38 - all these indices al statistical analysis, and they look at the distribution, they try to fit a probability distribution and then they try to say, okay, what is the mean value or the actually standardized it. Some, some called tric, non-parametric mixture. And I thought that was not what we wanted. We wanted to see, whether you can predict in some way that what is happening in a smaller scale, maybe the city scale or regional scale. And then of course I had been working with, for some time signal processing. And although I'm a civil engineer, but I did work for a long time when I, I was in US
00:02:11 - I was very interested in the signal processing and its application because, so I thought that there is some signals in the nature which we might be able to recognize something from the signals that based on that, that the drought is coming and we might not know it, we may not physically know it, but then maybe this hidden the data and, and the question is, what type of data then? Then I thought maybe what can be relevant data? And then we did not know really. So what will be, so we know the drought is occurred, so,
00:02:42 - there are some records of droughts. The meteorological data is there and also the, the city council have got data on the dam levels, thing. So, so I thought, let us look at that. What are the parameters which affect this in the past has Affected the drought condition. So there, or there was something in the data that, you know, from which I can recognize or we can recognize, that actually is hidden, but we cannot see it for, you know, we cannot analyze ourselves humanly, but maybe lot of data and digging and our thing data mining

Predicting Drought Conditions Using Neural Networks

00:03:16 - And then I thought of course artificial neural network is, has been used quite a bit here and there it has been, there are improved portion of it, but that is one of the tools. So, but, but the idea is to extract as much information as we can from this data and then find out which parameters actually influence the drought conditions. But that we learned from the past that, you know, so, and then we are trying to various, various parameters,
00:03:44 - that whether they're hidden in the data and then it could have in the past it could have actually predicted the drought. Okay. The also, I'll say that when I look at the literature there, I, I can mention some of the, some of the publications, others, and they looked at the, standardized, precipitation index, palmers index and also their various drought indexes. So, and there some are useful, some are not useful, somewhat accurate predictors not so, but then as a, that is a, that's takes a slightly more larger scale picture on a global scale
00:04:19 - maybe the global circulation model might be a part of it. The southern OS oscillation index might be a part of it and it is maybe EL effects and so, and other things might be a part of it. I'm not saying that you can ignore them, but then they are slightly looking at a larger scale. So I, I thought that maybe first thing is that whatever data we have from the precipitation records and maybe the soil moisture levels localized in a very localized scale, which is available from maybe I think,
00:04:51 - and also geoscience in Australia. And some of this of course from the city council data for the dams, whether we can use this data and see what has happened in the past, how maybe everything might not be important. Maybe the precipitation six months ago might not be important, but maybe last three months might be important. Maybe the soil, but soil moisture level is not that important because that is something maybe result and not a cause. And then so, so those type of things.
00:05:21 - oxidation the index, I can include that. What is the predicted value or the existing value or in so values or something. These are some of the parameters whether we can use this all input, all this information. See in the past from the record in the past whether we could have actually predicted the TR drought conditions, but with the leak time of maybe three months to six months, I knew that we, we are doing it for more than six months probably it erroneous. But if you can know if drought is coming even six months
00:05:52 - because there a, there is is a supply side and also there's a, you know, demand site. So we can manage a demand site if you can predict it earlier, especially in urban settings, in agriculture settings. So I thought that that is very useful. But surprisingly what we found, and I actually used one of some of our B honor students and I things, so let us look at as a, I'm aware the some some application of A N was there before, but there, you know, so I think Desai and Misra, Misra and DE

Predicting Drought Through Data Analysis Techniques

00:06:23 - some, some of the work has been done, but they're basically looking at reproducing the time series of the, some of this indices using this. So basically the learning from the data by looking at the time series of, you know, maybe SPI or something, a standardized specification index or something like that. But that was not what we wanted though it might look, appear to a similar in concept. But this was saying we look at, because Aiden and other things are other this you know,
00:06:53 - ability to find out what are the parameters which are relevant, which are not relevant, what is important and what is not important. And then probably that based on that we can learn from the past and then use that in the future. That was the whole idea. And we found the soil master was a very surprising thing that once you into into the soil master, you know, palace from you know, the metallurgy and meteorology, the prediction was very good and we are surprised to see that that is what affected quite a bit because that was something hidden something in that data
00:07:25 - so it was very successful in in my opinion, so in predicting not only the soil moisture level but also SPI also is we try to use this, this input, so predict the SPI also this this standardized per precipitation index and also the soil MAs level. Then you went to groundwater levels, you went to the, even the solidity because in coastal, like for solidity is important. If the groundwater uses more, the gradients comes towards from the sea to the,
00:07:55 - and based on the data whether we can fit the data. And then if you're successful, then also we have to look at the output. Also another output was the DAM level. So there's a dam nearby which is also dam features like tedious conditioning. The droughts is like 15% or 10%, which is no water to supply and whether, whether you can predict the dam level also. So it is the past data to check whether it is actually able to do it. But we did not assume that we assure that maybe we have the data for even when we're checking the
00:08:25 - that we have the data only from six months below before that and you know then whether you can predict with the leak time of six months or three months, whatever. And again, I use the data which is the lead time of six months or three months and then see whether it can be predicted. So it is not that we are, you are learning from the data, but we are checking in a the real time, okay, the major findings is that it this the a n and also maybe it's improvement improved, you know, some of the also because now a lot of advancements are going on is able to extract the information, the pattern
00:08:56 - That is what I feel though it is in a digital form. So it can actually detect some of the patterns which might be useful in predicting the future. And that is what and ally and you know, in terms of numbers and these are the, these are hidden patterns that there and what we cannot do as human beings though it is, you know, we can use intuition or something, but a n can do that in much more effective way. But a n is not the only, I mean it can be something else because the concept come is a signal coming and what, what are the inputs that affecting the signals

Predicting Drought and Groundwater Levels Efficiently

00:09:28 - and how do you can predict what will be the next. So we looked at lead time of three months, six months, six months, six months and see what they tested. And then with the independent set of data, we said whether it is actually able to predict the drought. That is what it is. And not only that one, we are very, very moral or is very satisfactory prediction of groundwater levels, which is very important. A you went to the groundwater ty levels during the droughts and suddenly SPI was very good.
00:10:01 - and other things are many factors, but also you, we tried to break the SPI that is the, the the standardized precipitation index which looks, looks at the rainfall statistic, but I use that. Then once we have break that we use that as a input to these models. Okay. So that information was actually, actually created and then use input and so therefore our prediction in my, in very, very good better than anything that we have body, but also the dam levels. But again, the dam level prediction is
00:10:32 - So, but there is a black box element there because the operation policy otherwise might be actually distorting it somewhere, but you droughts in other where basically you try to release how much is available. So, so they're kind of, you know, related and quite close. So therefore we are able to actually pay the, the dime levels also. So the whole idea is that it'll improve and also we can use it in some other regions also. We can develop it for other region, not only in these localities,
00:11:03 - And also in mining areas we can use it for, because mine has a lot of, because they can actually, you can and some industries which they can predict how much water might be available or there might drought coming in. I think it's very, very, some of these mining people also picked up some of these, you know, work and then they thought it's very useful. But I am very satisfied with the results. this our, in my opinion what we did was preliminary, we have have the proof of concept. That is what we've developed. So this application with proof of concept
00:11:35 - It can do the job but it has to be extended and maybe some other parameters have to, you know, utilized, maybe it should be applied in some other regions with other geographic regions, with other climatic climatic conditions. And also basically I want to see an application where the, for agriculture, you know how to plan The irrigation when you have some or in urban areas, how do you risk put the restrictions earlier so that you can control the demand sites.
00:12:23 - and is of course this is, and there even hydrogen, you know, nowadays energy production is there. So a lot of water demand and we need to be actually affected. If your drought is impending drought is there. So, so I think there are very various applications possible, but the study is not complete. of course I want to carry it forward. I hope I get a chance to do that.

Faculti AI can make mistakes

Research

Policy

Analysis