Have you tried to convert your data series into signals so that you could work in the frequency domain for instance, using Fourier or wavelets. It won't tell you what triggers what, but it will allow you to have an idea if your dataset are related and at which scale (the idea with the wavelet analysis, allowing you to remove let say a general trend or noise...etc in your signal).

It is an observational study, not experimental so you cannot make sure such changes are due to land use change. Ideally you want to look at a number of reservoirs, with variations in land use changes.

You could use a modified landuse regression model. These are commonly used in air pollution studies to model the air pollution concentrations using the surrounding land uses. The hypothesis is that surrounding landuses act as drivers for air pollution concentrations.

However, you could use this as a descriptive model to explore the relationship between different land uses and the water quality.

The independent data for these models are measurements of the amount of surrounding land use by type. The dependent data would be your water quality measures.

You could also determine the change in area and then model this with change to the water quality.

A simple example, if you believe increased industrial land use is decreasing water quality, Regress the water quality measure against the annual change in % industrial land use (within a buffer of the reservoir). Multiple distance buffers can be applied to see how this changes the results.

I have attached powerpoint slides from a recent presentation that might help explain my idea.

Multivariate Statistical methods such as Principal component analysis and factor analysis are useful in differentiating the natural and anthropogenic influences on water quality.

Different land uses such as urban areas, agricultural fields, industrial zones etc may be demarcated by their characteristic ionic composition. By doing a factor analysis, the different sources can be grouped. And later plotting these factor as spatial variation maps in comparison with land use maps, you can easily show the corresponding influence on water quality.

For you reference a paper published in contaminant hydrology is attached.

Have you tried to convert your data series into signals so that you could work in the frequency domain for instance, using Fourier or wavelets. It won't tell you what triggers what, but it will allow you to have an idea if your dataset are related and at which scale (the idea with the wavelet analysis, allowing you to remove let say a general trend or noise...etc in your signal).

Gomez solution seems realistic to me as well.

First of all, thank you very much for all those responding to my question. This forum is very useful for me. I am a new comer in remote sensing and I am trying to learn a lots of new concepts and analytical tools. Comparing may reservoirs from different types of watershed is a challenging idea for me due to time and financial constraints, and that is why if I can come one or two reservoirs from different watershed in maximum, I will take this option. I am planning to apply wavelet to reduce noise for my time series data with the purpose of identifying land use pattern. For the regression model and PCA, would that be possible since I will have two sets of time series data? Let's say, I will have land use map in 5 year interval for 15 years, and I will have also modeled water quality map of the reservoir (i.e, TSS and Chl-a).

You may also want to refine what water quality problem you want to focus on, the size of your actual reservoir, and the spatial and temporal dimensions of your research problem. Different water quality problems may have different spatial and temporal scales/dimensions,that may lead you to different approaches to spatial and temporal stratification/sampling, sensor selection, etc.

You may also want to make a broader literature review of methods used in reservoirs, but, also, in landscape/lucc analysis, to have a broad "menu" of methods to choose from.

in general, landscape and land use change analyses may offer good references for you; here, spatial temporal dimensions have been dealt with in many many papers that may inspire your work. Other than Remote Sensing of Environment, I.J.R.S., and the like, you may also want to makre a critical review of articles from Land Use Policy, Landscape Ecology, and journals like these, which have editorial policies covering as much scientific questions as methodological investigations..

Principal component analysis is a recommended statistical analysis that will highlight particular land use variables in relation to incidences in water quality change per time.